* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Climate Change Impacts On Rainfed Corn Production In
Heaven and Earth (book) wikipedia , lookup
Global warming controversy wikipedia , lookup
ExxonMobil climate change controversy wikipedia , lookup
Soon and Baliunas controversy wikipedia , lookup
Global warming hiatus wikipedia , lookup
Michael E. Mann wikipedia , lookup
Fred Singer wikipedia , lookup
Climatic Research Unit email controversy wikipedia , lookup
Climate resilience wikipedia , lookup
Climate change denial wikipedia , lookup
Climate engineering wikipedia , lookup
Atmospheric model wikipedia , lookup
Politics of global warming wikipedia , lookup
Global warming wikipedia , lookup
Citizens' Climate Lobby wikipedia , lookup
Climate governance wikipedia , lookup
Climate change feedback wikipedia , lookup
Climate change in Australia wikipedia , lookup
Solar radiation management wikipedia , lookup
Carbon Pollution Reduction Scheme wikipedia , lookup
Economics of global warming wikipedia , lookup
Climate change adaptation wikipedia , lookup
Climatic Research Unit documents wikipedia , lookup
Climate change in Tuvalu wikipedia , lookup
Effects of global warming on human health wikipedia , lookup
Climate sensitivity wikipedia , lookup
Media coverage of global warming wikipedia , lookup
Scientific opinion on climate change wikipedia , lookup
Public opinion on global warming wikipedia , lookup
Instrumental temperature record wikipedia , lookup
Attribution of recent climate change wikipedia , lookup
Climate change in Saskatchewan wikipedia , lookup
Climate change in the United States wikipedia , lookup
Effects of global warming wikipedia , lookup
Years of Living Dangerously wikipedia , lookup
Climate change and poverty wikipedia , lookup
Surveys of scientists' views on climate change wikipedia , lookup
Effects of global warming on humans wikipedia , lookup
Climate change and agriculture wikipedia , lookup
Climate change, industry and society wikipedia , lookup
University of Central Florida Electronic Theses and Dissertations Masters Thesis (Open Access) Climate Change Impacts On Rainfed Corn Production In Malawi 2013 Kondwani Msowoya University of Central Florida Find similar works at: http://stars.library.ucf.edu/etd University of Central Florida Libraries http://library.ucf.edu Part of the Engineering Commons, and the Water Resource Management Commons STARS Citation Msowoya, Kondwani, "Climate Change Impacts On Rainfed Corn Production In Malawi" (2013). Electronic Theses and Dissertations. Paper 2771. This Masters Thesis (Open Access) is brought to you for free and open access by STARS. It has been accepted for inclusion in Electronic Theses and Dissertations by an authorized administrator of STARS. For more information, please contact [email protected]. CLIMATE CHANGE IMPACTS ON RAINFED CORN PRODUCTION IN MALAWI by KONDWANI MSOWOYA Bachelor of Science in Civil Engineering, 2007 University of Malawi, the Polytechnic A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in the Department of Civil, Environmental and Construction Engineering in the College of Engineering and Computer Science at the University of Central Florida, Orlando Florida Fall Term 2013 Major Professor: Kaveh Madani ABSTRACT Agriculture is the mainstay of the economy in Malawi and accounts for 40% of the Gross Domestic Product (GDP) and 90% of the export revenues. Corn (maize) is the major cereal crop grown as staple food under rainfed conditions, covers over 92% of the total agricultural area, and contributes 54% of the caloric intake. Corn production is the principle occupation and major source of income for over 85% of the total population in Malawi. Issues of hunger and food insecurity for the entire nation are associated with corn scarcity and low production. Global warming is expected to cause climate change in Malawi, including changes in temperature and precipitation amounts and patterns. These climate changes are expected to affect corn production in Malawi. This study evaluates the impacts of climate change on rainfed corn production in Malawi. Lilongwe District, with about 1,045 square miles of agriculture area, has been selected as a representative area. First, outputs of 15 General Circulation Models (GCMs) under different emission scenarios are statistically downscaled. For this purpose, a weather generator (LARSWG) is calibrated and validated for the study area and daily precipitation as well as minimum and maximum temperature are projected for 15 GCMs for three time horizons of 2020s, 2050s and 2090s. Probability assessment of bounded range with known distributions is used to deal with the uncertainties of GCMs’ outputs. These GCMs outputs are weighted by considering the ability of each model to simulate historical records. AquaCrop, a new model developed by FAO that simulates the crop yield response to water deficit conditions, is employed to assess potential rainfed corn production in the study area with and without climate change. Study results indicate an average temperature increase of 0.52 to 0.94oC, 1.26 to 2.20oC and 1.78 to 3.58oC in the nearterm (2020s), mid-term (2050s) and long-term (2090s) future, respectively. The expected changes in precipitation during these periods are -17 to 11%, -26 to 0%, and -29 to -3%. Corn ii yields are expected to change by -8.11 to 0.53%, -7.25 to -14.33%, and -13.19 to -31.86%, during the same time periods. The study concludes with suggestion of some adaptation strategies that the Government of Malawi could consider to improve national food security under climate change. iii First, this study is dedicated to my beloved mother Maggie and my father Wyson for making me who I am today. I also dedicate this work to the love of my life Ruth and my daughter Vania for their perpetual love, support and for believing in me. I love you so much. iv ACKNOWLEDGMENTS My sincere gratitude goes to my thesis advisor Dr. Kaveh Madani for his perpetual support and for entrusting me with the responsibility to carry out this research. I am thankful to my MS committee members Dr. Dingbao Wang and Dr. Petros Xanthopolous for their constructive advice rendered during the course of the study. I also acknowledge profound comments from Jeffrey Elledge, which contributed to the perfection of this work. I am greatly indebted to Rahman Davtalab for his tireless support throughout the study. He was always there for me whenever I needed his assistance. Thanks also to Andres Vargas, Milad Hooshyer, Ahmed Hamed, Marcial Mota, Tyler Pierce, Mousa Maimoun, Juan Gonzalez, Amir Zarvichi, Saeed Hadian, Chris Rowney, and all HEESA members at UCF for their precious comments in preparation of this work. Many thanks go to my beautiful and wonderful friend and wife Ruth, for her encouragement and support throughout the course of my studies. Special recognition and appreciation to Mr. Adams Chavula, the Principal Agrometeorologist in the Department of Climate Change and Meteorological Services in Malawi, for providing me with all the required historical observed weather data. My sincere gratitude also goes to my financiers, the Fulbright Scholarship Board for according me the opportunity to advance my knowledge and skills. Family and friends too numerous to mention from back home in Malawi, UK, US, Australia, Netherlands, Japan and all parts of the world, I say thank you for being there for me. I love you all, May the Almighty God always bless you! v TABLE OF CONTENTS LIST OF TABLES ......................................................................................................................... ix LIST OF FIGURES ........................................................................................................................ x CHAPTER ONE: INTRODUCTION ............................................................................................. 1 1.1 Background ........................................................................................................................... 1 1.2 Problem Statement and Study Justification .......................................................................... 4 1.3 Research Objectives .............................................................................................................. 6 1.4 Method Summary.................................................................................................................. 6 CHAPTER TWO: LITERATURE REVIEW ................................................................................. 8 2.1 Potential Climate Change Impacts on Crop Production ....................................................... 8 2.2 Models Used to Predict Climate Change (GCMs) .............................................................. 11 2.3 Climate Change Emission Scenarios .................................................................................. 13 2.4 Selection of Baseline Period ............................................................................................... 15 2.5 Downscaling of GCM Outputs ........................................................................................... 16 2.6 Crop Model ......................................................................................................................... 19 CHAPTER THREE: BRIEF OVERVIEW OF THE STUDY AREA.......................................... 22 CHAPTER FOUR: RESEARCH METHODOLOGY.................................................................. 26 4.1 Introduction ......................................................................................................................... 26 4.2 Long Ashton Research Station Weather Generator (LARS-WG) ...................................... 28 4.2.1 Input data for LARS-WG............................................................................................. 28 4.2.2 Outline of the Process of Stochastic Weather Generation ........................................... 29 4.2.2.1 Calibration of the Model ....................................................................................... 29 4.2.2.2 Model Validation .................................................................................................. 31 4.2.2.3 Generation of Synthetic Weather Data ................................................................. 32 4.3 GCM’s Emission Scenarios ................................................................................................ 34 4.4 Uncertainty Analysis of GCMs ........................................................................................... 36 4.4.1 Step 1: Weighting of GCMs ........................................................................................ 36 4.4.2 Step 2: Generation of Probability Distribution Functions (PDFs) ............................... 37 4.4.3 Step 3: Generation of Cumulative Distribution Functions (CDFs) ............................. 38 4.5 Crop Yield Estimation with AquaCrop............................................................................... 38 4.5.1 Introduction .................................................................................................................. 38 vi 4.5.2 Description of the Model ............................................................................................. 39 4.5.2.1 The Atmosphere: Required Input Climate Data ................................................... 41 4.5.2.2 The Crop ............................................................................................................... 42 4.5.2.3 The Soil ................................................................................................................. 43 4.5.2.4 The Management .................................................................................................. 44 4.5.3 Algorithm and Calculation Procedure of AquaCrop ................................................... 44 4.5.3.1 Drainage ................................................................................................................ 45 4.5.3.2 Surface Runoff and Infiltration ............................................................................. 46 4.5.3.3 Soil Evaporation.................................................................................................... 47 4.5.3.4 Crop Transpiration ................................................................................................ 48 4.5.3.5 Biomass Production .............................................................................................. 49 4.5.3.6 Response of Harvest Index (HI) to Water Stress .................................................. 50 4.5.4 Calibration of AquaCrop.............................................................................................. 51 4.5.5 Validation of AquaCrop ............................................................................................... 52 CHAPTER FIVE: RESULTS AND DISCUSSION ..................................................................... 55 5.1 Introduction ......................................................................................................................... 55 5.2 Estimation of Future Climate Variables ............................................................................. 55 5.2.1 Calibration and Validation of LARS-WG ................................................................... 55 5.2.2 Estimating Future Climate Variables ........................................................................... 58 5.2.3 Uncertainty Analysis of GCMs .................................................................................... 62 4.2.3.1 Weighting of GCMs .............................................................................................. 62 4.2.3.2 Probability Distribution Functions (PDFs) ........................................................... 62 4.2.3.3 Cumulative Distribution Functions (CDFs) .......................................................... 66 4.2.3.4 Generation of Probability Percentiles ................................................................... 69 5.3 Corn Yield Production with Aquacrop ............................................................................... 76 5.3.1 Model Calibration ........................................................................................................ 76 5.3.2 Model Validation ......................................................................................................... 77 5.3.3 Simulating Current and Future Corn Yields with Climate Change ............................. 78 5.3.4 Correlation Analysis between Corn Yields and Rainfall and Temperature Changes .. 81 CHAPTER SIX: CONCLUSIONS AND RECOMMENDATIONS ........................................... 84 6.1 Study Results ...................................................................................................................... 84 vii 6.2 Suggestions to Climate Change Adaptation Strategies ....................................................... 86 6.3 Study Limitations ................................................................................................................ 88 6.4 Future Work ........................................................................................................................ 89 6.4.1 Introduce Supplementary Irrigation Corn Farming ..................................................... 89 6.4.2 Improving the Water Retaining Capabilities of the Soils ............................................ 88 6.4.3 Intra Annual and Inter Seasonal Variations ................................................................. 89 6.4.4 Using the Up to Date Emission Scenarios (AR5) ........................................................ 89 REFERENCES ............................................................................................................................. 90 viii LIST OF TABLES Table 1. A Comparison between NWP Models and GCMs ........................................................ 12 Table 2. The GCMs used in this study.......................................................................................... 12 Table 3. Driving forces for emission scenarios for 2020, 2050 and 2100 timelines .................... 14 Table 4. Characteristics of the major emission scenarios ............................................................. 14 Table 5. Description of the two emission scenarios used in this study......................................... 15 Table 6. Comparative summary of the relative merits and demerits of statistical and dynamical downscaling techniques ................................................................................................. 18 Table 7. Summary of crop models used in climate change impact studies .................................. 20 Table 8. CO2 concentrations in parts per million (ppm) for the four main climate scenarios (SRES) ........................................................................................................................... 35 Table 9. Input crop parameters used in calibration of AquaCrop ................................................. 52 Table 10. Statistical comparison of synthetic and observed weather data using Q test ................ 56 Table 11. Calculated weight of each GCM in simulating future rainfall...................................... 63 Table 12. Calculated weight of each GCM in simulating future temperature .............................. 63 Table 13. Summary of mean annual and overall ranges of estimated climate variables in each time period, at different scenarios and probability percentiles ...................................... 70 Table 14. AquaCrop validation results including prediction error statistics for corn yields ........ 78 Table 15. Corn production (tons/hectare) at different probability percentiles.............................. 79 Table 16. Summary of values of correlation coefficient for rainfall and maximum temperature 83 ix LIST OF FIGURES Figure 1. Malawi’s agricultural value-added and total GDP (Billion USD), 1970-2008 ............... 2 Figure 2. Corn production and price in Malawi from 1980-2012 .................................................. 4 Figure 3. Major emission scenarios .............................................................................................. 15 Figures 4a and 4b. Map of Malawi and location and soil types in Lilongwe District. ................. 22 Figure 5. Corn production and area under cultivation in the Lilongwe District (2004-2013) ...... 24 Figure 6. Topography of Lilongwe District ................................................................................. 24 Figure 7. Land use and vegetative cover in the Lilongwe District ............................................... 25 Figure 8. Methods that have been used in this study .................................................................... 27 Figure 9. Main components of soil-plant-atmosphere continuum in AquaCrop .......................... 40 Figure 10. Plot of P-values for X2 and t tests for rainfall and temperature................................... 57 Figure 11: Box plots showing the simulated mean monthly rainfall for 15 GCMs under scenarios A and B for three different time steps. ........................................................................ 59 Figure 12. Box plots showing the simulated mean monthly minimum temperature for 15 GCMs under scenarios A and B for three different time periods. .......................................... 60 Figure 13. Box plots showing the simulated mean monthly maximum temperature for 15 GCMs under scenarios A and B for three different time steps.. ............................................. 61 Figure 14. Sample Discrete PDFs outlining the relationship between weights of 15 GCMs and monthly changes in relative rainfall.. .......................................................................... 64 Figure 15. Sample Discrete PDFs outlining the relationship between weights of 15 GCMs and monthly changes in minimum temperature................................................................. 65 Figure 16. Sample Discrete PDFs outlining the relationship between weights of 15 GCMs and monthly changes in maximum temperature.. .............................................................. 66 Figure 17. Sample CDFs for rainfall based on PDFs in Figure 14. .............................................. 67 Figure 18. Sample CDFs for minimum temperature based on PDFs in Figure 15. ...................... 68 Figure 19. Sample CDFs for maximum temperature based on PDFs in Figure 16. ..................... 69 Figure 20. The estimated future changes in relative rainfall at three probability percentiles. The horizontal dash-dot line indicates no changes in rainfall.. .......................................... 71 Figure 21. Estimated future changes in minimum temperature at three probability percentiles.. 72 Figure 22. Estimated future changes in maximum temperature at three probability percentiles . 73 Figure 23. Plots of changes in rainfall, minimum temperature and maximum temperature under each scenario and risk level ........................................................................................ 75 Figure 24. Results of AquaCrop validation using the 2000-2004 simulated and recorded corn yields ........................................................................................................................... 77 Figure 25. Changes in corn production for 2015-2035 relative to baseline period of 1971-2000.80 Figure 26. Change in corn production for 2046-2065 relative to baseline period of 1971-2000. 80 Figure 27. Change in corn production for 2080-2099 relative to baseline period of 1971-2000. 80 Figure 28. Correlation analysis between change in corn yields and change in precipitation and maximum temperature respectively ............................................................................ 82 x LIST OF ABBREVIATIONS B : Biomass CC : Canopy Cover CDFs : Cumulative Distribution Functions CPT : Climate Predictability Tool CSI : Critical Success Index Dr : Root zone depletion DSSAT : Decision Support System for Agro-technology Transfer ETo : Reference evapotranspiration FAO : Food and Agriculture Organization of the United Nations FAOSTAT : Food and Agriculture Organization Statistics GCM : Global Circulation Model GCM : Global Circulation Models GDD : Growing Degree Day GDP : Gross Domestic Product GoM : Government of Malawi HI : Harvest Index IPCC : Intergovernmental Panel for Climate Change JICA : Japan International Cooperation Agency Kcb : Crop transpiration coefficient Ke : Crop evaporation Kr : Reduction coefficient Ks : Stress coefficient KS : Kolmogorov-Smirnov KW : Kruskal- Wallis LAI : Leaf Area Index Lars-WG : Long Ashton Research Station Weather Generator LEPS : Linear Error in Probability Space MAE : Mean Absolute Error MAGM : Master of Science in Agricultural Meteorology MAPE : Mean Absolute Percentage Error xi MK : Malawi Kwacha NSO : National Statistics office NWP : Numerical Weather Prediction PDFs : Probability Distribution Functions RMSE : Root Mean Square Error SEPLD : Social Economic Plan for Lilongwe District SRA : Scenario A SRES : Special Report on Emission Scenarios UDS : United States Dollar USDA : US Department of Agriculture USGCRP : United States Global Change Research Program xii CHAPTER ONE: INTRODUCTION 1.1 Background Agriculture is the mainstay of economy in Malawi. It employs 85% of the labor force, standing for 40% of the Gross Domestic Product (GDP) and 90% of the foreign exchange earnings (Msukwa, 1994; GoM, 2007b; Mucavele, 2009). With a subtropical climate, Malawi relies on rainfall for its agricultural production. Since time immemorial, corn has been the main staple food crop for the entire nation. Of the 9.4 million hectares of total land under cultivation between 1960 and 2000, only 7,404 hectares of this land was irrigated, leaving 99.99% of the total agricultural land for rainfed farming. Corn covers more than 92% of the total agricultural area (GOM, 2007b) and contributes over 54% of the caloric intake in Malawi (NSO, 2005). Most importantly, agriculture is the principle occupation for over 85% of the total population in Malawi (Kachule, 1994; GoM 1995). Other food crops grown on a smaller scale include cassava, potatoes, rice, coffee, tea, sorghum, beans, Irish potatoes, bananas, groundnuts and sugarcane. Figure 1 shows the contribution of agriculture to the overall GDP for the past four decades, clearly indicating that the rate of economic growth in Malawi is greatly influenced by agriculture (Tchale, 2009). Over the past few decades, corn production has been on the decline due to erratic rainfall and temperature changes, with rains either falling for shorter periods and in smaller amounts than usual or due to serious extreme events like droughts and flash floods. Heavy droughts experienced in 1991-92, 1994-95 and 2004-2005, were followed by flash floods in the subsequent years. During these periods, corn production significantly declined leaving a wide margin between production and consumer satisfaction (GoM, 2007b). Since 2005, corn production has fluctuated from one year to another, with an average production of 0.7-1.5 tons 1 per hectare (t/ha). Corn production below 1.5 t/ha indicates famine for that particular period (JICA, 2008). Previous reports in Malawi indicate that low corn production in some parts of the nation due to low rainfall has resulted in farmers living on an empty stomach for as long as five months until the next growing season (Kanyama-Phiri et al., 1994). Hunger crisis experienced in the aforementioned periods led to declarations of national state of emergency (GoM, 2012). National GDP (Billion USD) 14 GDP (Billion USD) Agriculture Value Added (Billion USD) 12 10 8 6 4 2 2008 2006 2004 2002 2000 1998 1996 1994 1992 1990 1988 1986 1984 1982 1980 1978 1976 1974 1972 1970 0 Year Figure 1. Malawi’s agricultural value-added and total GDP (Billion USD), 1970-2008 (Reproduced from Tchale, 2009). Previous research on climate change impacts on crop production in different parts of the world agrees with the assertion that climate change has major implications for agricultural production in general and for corn yield in particular. In the United States and other developed nations, extensive studies on the impacts of climate change on agricultural production have been carried out (Cai et al., 2009; Adams, 1989; Rosenzweig, 1989; Mendelson et al., 1994). There has been relatively little research in developing nations although recently, a few papers have been published (Schmidhuber and Tubiello, 2007; Ringler et al., 2010; Gohari et al., 2013). Yet, 2 developing countries are the ones which could suffer more from the effects of climate change, due to their limited economic capability and adaptive capacity. Being one of the least developed countries, Malawi’s corn production can be affected by climate change impacts. Nevertheless, little is known about such impacts to date. Population increase in the region has been blamed as a contributing factor to low crop production. This is because population increase has resulted in loss of agricultural land to settlement, continuous farming and deforestation, eventually leading to soil erosion and infertility. In consequence, the market value of agricultural grains has increased dramatically (Ng’ong’ola et al., 1992). Figure 2 shows the changes in corn production and average price in Malawi for the 19802011 period. It is clear from this figure that corn production has indeed not been stable; however, corn prices have been steadily on the rise over the past decades. In 1980’s, an average national growth rate of 3% in corn production was observed, followed by a growth rate decline of 2 percent per annum from 1990 to 1994. Heavy drought periods experienced from 1991 to 1992, exacerbated corn production during this period and during the following years. A series of positive and negative growth rates were experienced between 1994 and 2004, resulting in a tip over of 2.2% per annum in 2004 (Chirwa et al., 2006). Steady increase in corn production was realized from 2005-2008 when the Government of Malawi (GoM) introduced its subsidy program giving farmers the opportunity to buy farm inputs like fertilizer and seeds, at subsidized rates. However, low rainfall amounts and poor rainfall distribution in most parts of the country have since resulted in low corn yields. Climate change impacts on rainfed corn production in Malawi have not been adequately investigated to date. Given the significance of this staple crop for the nation’s labor force and 3 GDP, this study intends to bridge the knowledge gap on the effects of climate change on rainfed corn production in Malawi. 3.5 0.4 Production Price 0.35 Million Tons 3 0.3 2.5 0.25 2 0.2 1.5 0.15 1 0.1 0.5 0 1980 0.05 Price in thousand USD per Ton 4 0 1985 1990 1995 Year 2000 2005 2010 Figure 2. Corn production and price in Malawi from 1980-2012 (Chirwa et al., 2006; GoM, 2013) 1.2 Problem Statement and Study Justification The fourth report by the Intergovernmental Panel on Climate Change (IPCC) indicates that on average, global temperature has increased by 0.74oC (0.56 to 0.92oC) from 1906-2005, and a warming of 0.2oC per decade is estimated for the next 2 decades (SRES, 2000; IPCC, 2007). Extensive studies have shown that the changes in global temperature will have a significant effect on agricultural production (Thornton et al., 2006; Mendelsohn, 2008; USGCRP, 2009; Lee et al., 2012). These studies have helped to comprehend and appreciate the extent by which crops would be vulnerable to varying climatic conditions. For instance, studies have investigated the effects of climate change on crop yield variability (Carlson et al., 1996; Phillips et al., 1998; Podestá et al., 1999; Rubas et al., 2006) and crop yield (Kaiser et al., 1993; Rosenzweig and Parry 1994; Riha et al., 1996; Reilly et al., 2003; Parry et al., 2004). In 4 summary, these studies have provided evidence that climate change will have impacts on agricultural production. Poor spatial rainfall distribution and pattern often results in crop production losses (Rockström, 2000). Shortage of corn production in Malawi results in significant rise in market prices of corn (GoM, 2006). This has a direct impact on the local population of which 85% have an income of less than $1 per day. A market information report on food crops by the Malawian Ministry of Food Security (2006) indicates that in 2005 and 2006, the average national corn prices were $0.14/kg and $0.37/kg, respectively, indicating an increase of 160% just in a year. This rise was as a result of the drought experienced within this period. However, as of June 2013, the national average corn price was $0.53/kg, representing an increase of approximately 276% from 2005 (GoM, 2006; 2013). This phenomenon simply shows that despite an increase in production (about 150% increase in corn production), corn prices have also significantly increased, perhaps due to the increased population and demand. Amongst all sectors (energy, agriculture, water, fisheries, livestock et cetera), agriculture is the most sensitive and vulnerable to climate change (IPCC, 1990). The second IPCC report (1996) predicted that tropical and subtropical regions would experience higher losses in crop production, while temperate climates might gain in productivity with climate warming. Issues of hunger and famine in Malawi are associated with low corn production as a result of poor rainfall. It is apparent therefore, that as a subtropical nation that wholly relies on rainfed farming, Malawi’s food production could be greatly impacted on by changes in climate. Therefore, studying the potential impacts of climate change on rainfed corn production in Malawi would be necessary to take timely mitigating actions to minimize the undesirable impacts. 5 1.3 Research Objectives The main objective of this study is to evaluate the effects of climate change on rainfed corn production in Malawi. Specifically, the study aims to achieve the following objectives: i. Quantifying future changes in temperature and precipitation relative to the baseline period of 1971-2000; ii. Estimating potential current and future corn production in the study area in various future time periods under different climate change scenarios; and iii. Recommending some adaptation strategies to minimize the undesirable impacts of climate change on food production in Malawi. 1.4 Method Summary Lilongwe District, the largest corn growing/producing District in Malawi, has been selected as representative study area to estimate the potential effects of climate change on corn production in the nation. Relative to the baseline period 1971-2000 for which historical climate data were available, future changes in climate variables in 2015-2035 (2020s), 2046-2065 (2050s) and 2080-2099 (2090s) are estimated using General Circulation Models (GCMs). Following the method proposed by Gohari et al. (2013) to cope with uncertainties of GCMs projections, a probability analysis method with bounded distribution functions is used, which results in generation of mean monthly PDFs and CDFs from which risk levels at 25%, 50%, 75% are extracted. The LARS-WG stochastic weather generator (Semenov and Barrow, 2010) is used to downscale GCMs outputs to local area. Statistical indicators are used to investigate the significance and reliability of LARS-WG to predict future climate data, based on historical climate data. ETo Calculator (Allen et al., 1998), a FAO simulation model, is used to estimate evapotranspiration based on Penman-Monteith algorithm. AquaCrop (Steduto et al., 2009), a 6 new model developed by FAO, is used to estimate crop production under different climate scenarios, with and without climate change. The study concludes with suggestion of some adaptation strategies that the Government of Malawi could consider to improve national food security associated with climate change. The rest of this thesis is structured as follows: Chapter 2 reviews literature on the potential impacts of climate change on crop production and provides basic information on climate change models and scenarios as well as the downscaling technique used in this study. This chapter also provides background information on the available crop yield simulation models. Chapter 3 provides a brief background of the study area while Chapter 4 outlines the overall methodology used in order to meet the study objectives. Chapter 5 presents and discusses the results and major findings of the study. Chapter 6 suggests some adaptation techniques to local decision makers to minimize the cost of climate change for Malawi, outlines the limitations of the study and concludes the study. 7 CHAPTER TWO: LITERATURE REVIEW 2.1 Potential Climate Change Impacts on Crop Production There is overwhelming evidence that the earth’s planet is warming as a result of greenhouse gasses (IPCC, 2007). Burning of fossil fuels and changes in land use are the major sources of greenhouse gasses; and if nothing is done to curb these practices, the concentrations of greenhouse gasses in the atmosphere will grow substantially by the next century (IPCC, 2007; Mendelsohn, 2008). Climate change is expected to cause various effects; however, the largest impact of global climate change will be on agriculture (Nordhaus, 1991; Cline, 2007). The United States Global Change Research Program (USGCRP, 2009) report indicates that despite an increase in crop yields as a result of technological improvements, extreme weather changes have negatively affected production in some years. The report argues that warmer temperatures not only shorten growth periods, they also lead to lower crop yields. Shorter growth periods may be appropriate in areas with soils of low moisture content; however this leads to accelerated growth in some varieties of crops like corn by shortening period of seed germination, crop growth and maturity. Ultimately, accelerated growth of crops results to a significant decrease in crop yields for a given amount of land. Tubiello et al., (2002) agrees with the notion that climate change greatly affects agricultural crop production, albeit the magnitude will vary from a place to another and from one crop to another due to different natural and anthropogenic factors that contribute to climate change. Several studies indicate that increase in temperature negatively affects crop growth and hence crop yields, however, the impact varies between crops and across regions since different crops respond differently to climate variability (Thornton et al., 2006; Mendelsohn, 2008; USGCRP, 2009; Lee et al., 2012). Mendelshon (2008) reviewed different studies that measure 8 the magnitude of the impact of warming on farms in developing countries. His review affirmed the hypothesis that “tropical and subtropical agriculture in developing countries is more sensitive to climate than temperate agriculture’’. His study concluded that even marginal changes in global warming will negatively affect crop growth and hence production in Africa and Latin America. Ringler et al. (2010) assessed Climate Change Impacts on food security in Sub-Saharan Africa using a comprehensive scenario which is based on ensembles of 17 GCMs. The study results depicted high temperatures and mixed changes in rainfall by the year 2050. The study also found that climate change would result in a decrease in crop yields excepting for millet and sorghum; with wheat exhibiting the largest decrease of -22% by year 2050. The study also determined that with climate change, prices of corn, rice and wheat would increase by 4%, 7% and 15% respectively by 2050 relative to the baseline period. Other predicted 2050 impacts by 2050 include: changes in crop production and increase in growth area, less affordability of food due to price increase, reduction in calorie intake and increased childhood malnutrition for the SubSaharan region Kimball (1983) conducted a study in the U.S. and discovered that increased amounts of carbon dioxide (CO2) can increase crop yields. However, this can be counteracted by a rise in temperature, leading to a reduction in crop yield overall. Cai et al. (2009) assessed climate change impacts on rainfed corn production in central Illinois, U.S. Study results indicate that under rainfed conditions, corn yields will decline by 23 to 34% in 2055 in central Illinois. In addition, if no adaptation measures are implemented, the study estimated a probability of 32 to 70% of not meeting 50% of the potential yield by the year 2055. You et al. (2009) carried out research in China and found out that an increase in temperature of 1 °C during the growing period may lead to wheat production reduction of 3–10% 9 and a decrease in winter wheat production of about 5%-35% (Ozdogan, 2011). Tao and Zhang (2010) determined that with higher temperatures, corn yields are likely to decrease by 2.4-45.6% in northern China plain. An average decrease in crop production of 15% may be incurred with higher temperatures and evapotranspiration rates despite constant precipitation, as a result of shorter crop growth periods which negatively affects time to crop maturity (Schlenker and Lobell, 2010). A similar study, conducted in northwestern Turkey, indicates that winter wheat yields may decline by more than 20% due to climate change causing shorter growth periods and reduced precipitation (Ozdogan, 2011). Similar studies within the same region determined that higher temperatures have a positive effect on crop yield. For instance, in northern China plain, corn, rice, potato, and winter wheat yields could increases with increasing temperature and precipitation under climate change (Chaves et al., 2009). A recent study evaluated the agricultural responses to climate change in Iran’s ZayandehRud River Basin for a period of 2015-2035. The study considered four crops; wheat, barley, rice and corn, and determined that for an average monthly temperature change of 1.1oC to 1.5oC and precipitation decrease of 11% to 31% within the basin, crop production would decrease as follows: 2.5% to 20.7% for wheat, 1.4% to 17.2% for barley, 2.1% to 9.5% for rice, and 5.7% to 19.1% for corn (Gohari et al., 2013). Schmidhuber and Tubiello (2007) reviewed the potential impacts of climate change on global food security. They determined that climate change will adversely impact all four aspects of food security namely: availability, stability, utilization and access. However, the effects will differ across regions and nations depending on the socio-economic policies embraced by each region or nation in response to climate change impacts. The study further states that climate change will exacerbate the dependence on imports by developing nations, with Sub-Saharan 10 Africa being more vulnerable compared to South Asia which will be affected but to a lesser extent. The paper concludes that climate change will mostly affect poor people. 2.2 Models Used to Predict Climate Change (GCMs) General Circulation Models (GCMs) are “computer based version of earth’s system that mathematically simulates the climate system and the interaction between the system components” (Hassan, 2012). They simulate historical, present and future climate scenarios taking into account the level of greenhouse gases and aerosols under different future projections. The process is achieved by dividing the oceans and atmosphere into a horizontal grid with a horizontal resolution of 2o- 48o with 10-20 layers aligned vertically. This enables predictions of climate change for the next 100 years using a coarse grid scale (IPCC, 2001 and Dibike and Coulibaly, 2005). In general, most GCMs are capable of simulating global and continental scale processes in detail and provide a reliable representation of the average planetary climate (Hassan, 2012). GCMs use the same equations of motion as a numerical weather prediction (NWP) model. The major difference is that NWP models are used to simulate future weather conditions for a short and medium period of time (say 1-3 days and 4-10 days, respectively). In contrast, GCMs predict weather conditions for much longer periods (multiple decades), besides providing statistical interpretation of the results (e.g. means, medians, variance and standard deviations) (Geerts and Linacre, 1998). NWP models and GCMs are compared in Table 1, while Table 2 lists the 15 GCMs used in this study. 11 Table 1. A Comparison between NWP Models and GCMs (Geerts & Linacre, 1998) CONTRASTS Goal Spatial coverage Temporal range Spatial resolution Relevance of initial conditions Relevance of clouds, radiation Surface (land, ice, ocean) Relevance of ocean dynamics Relevance of model stability Time dimension SIMILARITIES Physics NWP to predict weather regional or global days variable (20-100 km) high low low low low essential Method Finite difference expression of continuous equations, or spectral representation; run prognostically Variables and motion of the atmosphere in 3 dimensions Controlled by spatial resolution (CFL condition) Output Maximum time step GCM to predict climate global years usually coarse low high high high high ignored Use equations of motion (plus radioactive transfer equations, water conservation equations.) Table 2. The GCMs used in this study (Semenov, 1992; 2012) Center Beijing Climate Centre Canadian Centre for Climate Modeling and Analysis Centre National de Recherches Australia's Commonwealth Scientific and Industrial Research Organization Institute of Atmospheric Physics Goddard Institute for Space Studies Geophysical Fluid Dynamics Laboratory Hadley Centre for Climate Prediction and Research/UK Met. Office Hadley Centre for Climate Prediction and Research/UK Met. Office Institute for Numerical Mathematics Institute Pierre Simon Laplace Meteorological Research Institute, Japan Max-Planck Institute for Meteorology National Centre for Atmospheric Research National Centre for Atmospheric Research Center Acronym BCC China Global Climate Model BCC-CM1 Grid Resolution 1.9o x 1.9o CCCma Canada CGCM3 (T47) 2.8o x 2.8o CNRM France CNRM-CM3 1.9o x 1.9o CSIRO Australia CSIRO-MK3.0 1.9o x 1.9o LASG GISS GFDL China USA USA 2.8o x 2.8o 3o x 4o 2.0o x 2.5o UKMO UK FGOALS-g1.0 GISS-AOM GFDL-CM2.1 HadCM3 UKMO UK HadGEM1 1.3o x 1.9o INM IPSL NIES MPI-M NCAR Russia France Japan Germany USA INM-CM3.0 IPSL-CM4 MIROC3.2 (hires) ECHAM5-OM CCSM3 4o x 5o 2.5o x3.75o 1.1o x 1.1o 1.9o x 1.9o 1.4o x 1.4o NCAR USA PCM 2.8o x 2.8o 12 Country 2.5o x3.75o 2.3 Climate Change Emission Scenarios Climate models project future climate based on different emission scenarios, each based on a specific set of assumptions (Shaka, 2008). These assumptions include, future trends in energy demand, emitted greenhouse gases, changes in land use and behavior of climate system over a long period of time, etc. (Houghton et al., 2001). Figure 3 presents a schematic representation of the four major emission scenarios as per the Special Report on Emission Scenarios (SRES) (IPCC, 2007). These scenarios (A1, A2, B1 and B2) make different assumptions about future population increases, economic and technological developments, energy and land use, and the global approaches to sustainability which differently affects emissions of greenhouse gasses. In general, scenario A prioritizes economic development, whereas scenario B is more concerned with environmental sustainability. Numbers “1” and “2” represent different technological advancements within storylines, with “1” representing a much faster and diverse development and “2” representing a slower and regional development (IPCC, 2007). To put it another away, a future that is more concerned with addressing global problems and sustainable environment is represented by scenario B1. On the other hand, a future in which nations prioritize their regional development, leading to unsustainable and unequal economic growth is represented by scenario A2 (Anderson et al., 2008). Table 3 shows a summary of the driving forces for the 2020, 2050 and 2100 timelines leading to the development of A1, A2, B1 and B2 scenarios. Table 4 gives a brief description of the major characteristics of each scenario. 13 Table 3. Driving forces for emission scenarios for 2020, 2050 and 2100 timelines (Reproduced from SRES) Scenario Group A1 A2 B1 B2 Population (billion)(1990=5.3) 2020 7.6 8.2 7.6 7.6 2050 8.7 11.3 8.7 9.3 2100 7 15.1 7 10.4 12 World GDP (10 USD/year (1990=21) 2020 57 41 53 51 2050 187 82 136 110 2100 555 243 326 235 Per capita income ratio: developed countries and economies in transition (Kyoto Treaty Annex 1) to developed countries (Kyoto Treaty non-Annex 1) (1990=16.1) 2020 6.2 9.4 8.4 7.7 2050 2.8 6.6 3.6 4 2100 1.6 4.2 1.8 3 Table 4. Characteristics of the major emission scenarios (IPCC SRES, 2000) Scenario Group Population Growth GDP Growth Energy Use Land Use Change Oil/Gas Resource Availability Technological Change Change Favoring A1B Low Very High Very High Low Medium A2 Low Very High High Low Medium B1 Low High Low High Low B2 Medium Medium Medium Medium Medium Rapid Rapid Medium Medium Balanced Non-Fossil Fuel Efficiency and Dematerialization "dynamics as usual" Two commonly used emission scenarios, i.e. SRA1 and SRB1, are considered here in the analysis of the impacts of future climate change on corn yield. For simplicity, these two scenarios are referred to as SRA and SRB respectively. Table 5 provides a description of these two scenarios. 14 Figure 3. Major emission scenarios (Reproduced from Anderson et al., 2008) Table 5. Description of the two emission scenarios used in this study (IPCC- 4th Assessment Report) Scenario SRA1 Description A future world affected by a rapid economic growth, global population that peaks in mid-century and declines thereafter, and rapid introduction of new and more efficient technologies, with the development balanced across energy sources SRB1 A convergent world with the same global population as in the A1 storyline but with rapid changes in economic structures toward a service and information economy, with reductions in material intensity, and the introduction of clean and resourceefficient technologies 2.4 Selection of Baseline Period The baseline period is the reference period on which calculation of future climate changes is based. Definition of the baseline period is important in order to select the observed climate dataset that combines with climate change information to generate climate change projections (Houghton et al., 2001). Carter et al. (1994), Mohammed (2009) and Hassan (2012) outline four criteria that are commonly used in selection of the baseline period. First, the baseline period must truly represent the current or recent averages of climate conditions within 15 the area. Second, the baseline period must be sufficiently long and cover a wide range of climate variations, including extreme weather conditions. Third, the suitable baseline period is the one for which the major climatic data like rainfall (precipitation), temperature, sunshine and relative humidity are readily available, easily accessible and adequately distributed over space. Fourth, the baseline period should have high quality climate data (with few missing data, if any). Based on the aforementioned selection criteria, a baseline period of 30 years from 19712000 was selected for the purpose of this study. 2.5 Downscaling GCMs Outputs Global Circulation Models (GCMs) have been developed to predict average large-scale phenomenon of atmospheric weather (Holton, 1992). GCMs are quite capable of simulating global climatic data at continental and hemispherical scales, but have limited capacity in simulating local weather features and dynamics at sub-grid scale (Wigley et al., 1990; Carter et al., 1994). Outputs from climate models cannot be used directly in climate studies at local level due to the existence of a mismatch in spatial resolution between GCM outputs and regional hydrological/statistical models, in addition to the large coarse resolution of GCMs, compared to hydrologic/statistical models that use very fine spatial resolutions (Dibike and Coulibaly, 2005). Even if future GCMs runs on high resolution, there will still be a need to downscale the outputs so that they match the local sites before being further used in climate change studies (DOE, 1996). Downscaling therefore converts the coarse spatial resolution of any GCM outputs to a fine spatial resolution by generating local data from GCM outputs. This is done by linking global scale predictions to regional dynamics in order to generate climate variables specific to a particular region (Flint and Allan, 2012). Downscaling can be done using different approaches, 16 including: (i) nesting a regional climate model into an existing GCM, (ii) statistical regressions and (iii) using stochastic weather generators. The first method of downscaling entails identifying a specific area and applying some driving factors from a GCM to the regional climate models. This method is also referred to as dynamic downscaling because regional climate models are dynamic models like GCMs, albeit they exist in three layers. The first layer operates through a GCM; the second layer uses the local data for the specified area; and the third layer mathematically generates results based on the previous two layers. The output is moderately local data driven by both global models and regional specifics. The complexity of the models used makes the process challenging and less attractive. The main advantage of this technique is that it better simulates extreme climate events like floods and droughts and climate anomalies at regional scale (Fowler et al., 2007; Harun et al., 2008). Statistical regression involves using different regression methods to link local climate data to specific drivers in GCMs. First, a relationship between large scale variables (driving factors of GCMs) and climate variables at local scale is established. The established relationship is then used to predict future climate variables under various situations as stipulated by GCMs. In other words, regression models entail establishing “linear or nonlinear relationships between sub grid scale (e.g. single-site) parameters and coarse-resolution (grid-scale) predictor variables” (Hassan, 2012). Stochastic weather generators (WGs) are also statistically driven. WGs generate statistical linkages within climate variables in order to simulate present and future weather conditions of a specified station or area, using the historical observed climate variables of that station/area. The generated data are expected to have similar statistical properties with historical 17 observed data and the output data are complete records of daily time series unlike observed records which may have some missing data (Kevin et al., 2005). WGs provide a wide range of statistical interpretation of the observed and simulated daily weather variables for a particular site (Harun et al., 2008). The WGs are said to underestimate inter-annual variability (Buishand, 1978; Johnson et al., 1996; Chenet al., 2009). Table 6 provides a general comparison of the dynamic and statistical downscaling methods. Table 6. Comparative summary of the relative merits and demerits of statistical and dynamical downscaling techniques (Wilby and Wigley, 1999; Hassan, 2012) Statistical Downscaling Dynamic Downscaling Comparatively cheap and computationally efficient Produces responses based on physically consistent processes Produces finer resolution Information from GCM-scale output that can resolve atmospheric processes on a smaller scale Can provide point-scale Climatic variables from GCMscale output Advantages Disadvantages Can be used to derive variables not available from RCMs Easily transferable to other regions Based on standard and accepted statistical procedures Able to directly incorporate observations into method Require long and reliable observed historical data series for calibration Dependent upon choice of predictors Non-stationarity in the predictorpredict and relationship Climate system feedbacks not included Dependent on GCM boundary forcing; affected by biases in underlying GCM Domain size, climatic region and season affects downscaling skill 18 Computationally intensive Limited number of scenario ensembles available Strongly dependent on GCM boundary forcing This study employs the Long Ashton Research Station Weather Generator (LARS-WG), (Semenov and Barrow, 2002: 2010) a well-known weather generator that can estimate current and future climate variables for a specific area. LARS-WG is preferred for its ability to fill in missing data and interpolate/extrapolate data to a defined period of time using statistical properties of the historical observed data (Dibike and Coulibaly, 2005). LARS-WG has been tested in varying climatic conditions and has demonstrated a good performance in reproducing various weather statistics including extreme weather events in most situations (Semenov et al., 1998, Semenov 2008a). Results from the Lars-WG are in the form of daily time-series climate variables like rainfall (precipitation), maximum and minimum temperature, and solar radiation (Semenov and Barrow, 2002; Shaka, 2008). 2.6 Crop Model Several studies have been conducted to evaluate the impacts of climate change on crop production using crop simulation models. Examples of crop models that have been frequently applied thus far include: CERES-Wheat (Crop Environment Resource Synthesis) (Godwin et al., 1989; Jones et al., 2003), DSSAT (Hoogenboom et al., 1994) SWAP (soil–water–atmosphere– plant model) (van Dam et al., 1997), CERES-Maize (corn) (Jones & Kiniry, 1986), Crop Syst (Stöckle, 2003), CropWat (Allen et at., (1998), and InFoCrop (Singh et al., 2005a). Of these models, three are said to have a strong emphasis on interactions between crop, water and climate change, namely (i) CropWat, (ii) AquaCrop and (iii) SWAP (Aerts and Droogers, 2004). These three models have been widely used and tested in different climate change impact studies worldwide, have a user-friendly interface, and are openly accessible. Table 7 gives a summary of different crop models that have been used in previous climate change impacts assessment studies. 19 Table 7. Summary of crop models used in climate change impact studies (Kang et al., 2009) Crop model CERES-Maize CERES-Wheat Crop Syst Ceres Rice SWAP InFocrop IBSNAT-ICASA GLAM GLYCIM SWAT AquaCrop Target crop Maize Maize Maize Maize Wheat Wheat Wheat Rice Rice Rice Rice and Wheat Cereal/soybean Peanut Soybean Maize Teff Cotton Maize Area of Study Dry matter Sustainable production Planting date and different weather Precise and deficit irrigation CO2 levels CO2 levels Rainfall and warming CO2 levels CO2 levels Elevated CO2 and temperature Climate change Climate Change Climate uncertainty Rainfall & CO2 concentration Climate vulnerability Yield response to water Deficit irrigation optimization Yield response to water References Cuculeannu et al. (2002) Walker et al. (2006) Tojo et al. (2007) Popova et al. 2005) Eitzinger et al. (2003) Luo et al. (2003) Anwar et al. (2007) Yao et al. (2007) Droogers et al. (2004) Krishman et al. (2007) Aggerwal et al. (2006) Parry et al. (1999) Challinor (2008) Xie et al. (2004) Reddy et al. (2000) Araya et al. (2010) Garcia Villa et al. (2009) Zinyengere et al. (2011) In this study, the FAO (Food and Agriculture Organization) AquaCrop Version 4 (Steduto et al., 2009) is used to estimate corn production in the study area under different climate scenarios. AquaCrop provides an improved and powerful approach for assessing crop yield response to water deficit. This model has been developed and recommended by the FAO for several reasons: requires less climate data than most crop models, has a user-friendly interface, has a strong focus on relationships amongst climate change, carbon dioxide, water and crop yields, and it has a large number of users globally (Steduto et al., 2009). AquaCrop is free software and easily accessible at http://www.fao.org/nr/water/aquacrop.html. Numerous validation tests for AquaCrop have been carried out in different parts of the world. These tests have approved the reliability of AquaCrop when the input information is 20 clear-cut (Mainuddin et al., 2010; Todorovic et al., 2009; Heng et al., 2009; Farahani et al., 2009). 21 CHAPTER THREE: BRIEF OVERVIEW OF THE STUDY AREA Lilongwe District contains the capital city of Malawi. It is located at the center of Malawi at latitude and longitude 13.50 south and 33.420 east respectively, and at an elevation of 1133m above sea level. Figure 4a shows the map of Malawi and Figure 4b shows the location and types of soil in the Lilongwe District. Note that sandy loam soil is the prevalent soil type in the area. a b Figures 4a and 4b. Map of Malawi and location and soil types in Lilongwe District (SEPLD, 2006). The 2008 population and housing census revealed that the Lilongwe District has a total population of 1.9 million people, of which 35% live in urban areas whilst 65% persons live in rural areas. With the intercensal growth rates of 4.3 and 3.1 for Lilongwe’s urban and rural areas, respectively, the current (2013) total population for the District is estimated at 3.2 million people (NSO, 2008). During which The Lilongwe District has a humid subtropical climate that is relatively dry and strongly seasonal. The warm and wet season is from November to April, during which the District 22 receives over 95% of the total annual rainfall. The annual average amount of rainfall for Lilongwe is 900 mm, with the month of February receiving the highest monthly rainfall amount of about 211 mm. A cool dry winter stretches from May to August. During this period, mean temperature ranges from 17 oC to 27 oC, and minimum temperatures range from -3 oC to 10 oC, with the coldest temperatures in the months of June and July. A hot and dry season runs from September to October, when temperatures go as high as 36 oC and humidity ranges from 50% to 80%. The month of July is normally very dry (Malawian Department of Climate Change and Meteorological Services, 2013) Like in all parts of the nation, corn is by far the most important staple food for the people of the Lilongwe District. Average per capita corn consumption is approximately 133 kg per month. Corn also accounts for about 74% of the caloric intake per households and is cultivated by 97% of farmers. Corn is considered to be a source of life and wealth by smallholder farmers and occupies 54% of the land cultivated by small scale farmers (Peter and Harrera, 1989; IFPRI, 2002; NSO, 2005; FAO 2009a). It is therefore not surprising that the focus on food security in Malawi and in the District is on corn production. Figure 5 shows trends in corn production and the area under cultivation in the Lilongwe District during the 2004 to 2013 period. This figure shows that corn yields steadily increased from the 2004 growing season followed by a decline between the 2006 to 2008 growing seasons. Like in many parts of the nation, the decline is attributed to drought conditions that the nation experienced during these seasons. With the introduction of subsidized fertilizers, corn yields increased after the 2007 growing season. However, even with fertilizer subsidy, corn yields have not been as high as it was expected, despite continuous expansion in the cultivated area. Low rainfall amounts and shorter rainfall periods are considered to be the factors contributing to the 23 decline in corn yields (GoM, 2013). Figure 6 shows the topography of the Lilongwe District and Figure 7 shows land use and vegetative cover in the Lilongwe District. Note the coverage of rainfed farming in the District in Figure 7. 3.5 250 200 Yield (Tons) 2.5 150 2.0 1.5 Yield 100 Area 1.0 50 0.5 0.0 2002 2004 2006 2008 Year 2010 2012 0 2014 Figure 5. Corn production and area under cultivation in the Lilongwe District (2004-2013) Figure 6. Topography of Lilongwe District (SEPLD, 2006) 24 Area (1000 Hectares) 3.0 Figure 7. Land use and vegetative cover in the Lilongwe District (SEPLD, 2006) 25 CHAPTER FOUR: RESEARCH METHODOLOGY 4.1 Introduction This chapter first outlines the Long Ashton Research Station Weather Generator (LARSWG) model, a statistical model for downscaling climate variables. A description of model calibration and validation to determine its potential to simulate future climate data is then illustrated. Fifteen (15) GCMs each under scenarios A and B are also discussed. Second, due to high uncertainties associated with GCMs, the chapter provides a description of probability analysis for dealing with uncertainty of the models using weighting method. This process leads to generation of mean monthly probability distribution functions (PDFs) and cumulative distribution functions (CDFs), under each scenario for each time period. Consequently, climate variables at different probability percentiles (25%, 50% and 75%) are estimated. Third, we describe how AquapCrop, a crop simulation model estimates potential present and future corn production using climate variables generated in the previous section. Figure 8 is a step by step description of the methods used in this work. The historical observed metrological data used in this study was obtained from the Department of Climate Change and Meteorological Services in the Ministry of Natural Resources, Energy and Environment in Malawi. The data is from the Chitedze Research Station, the main meteorological station in Lilongwe District. The following climate data was obtained for a baseline period of 30 years (1971-2000): (i) Daily minimum and maximum temperature, (ii) Daily precipitation (rainfall) (iii) Mean daily relative humidity (iv) Daily hours of sunshine and 26 (v) Daily wind speed Figure 8. Methods that have been used in this study GCMs simulated data for mean monthly precipitation, minimum and maximum temperature, relative humidity and wind speed for the baseline period (1971–2000), were extracted from the Canadian Climate Change Scenario Network Site. The site is accessible at: (http://www.cccsn.ec.gc.ca/?page=dd-gcm ). The monthly differences between observed and GCMs simulated climate data for the baseline period of 1971-2000, were calculated and used for analysis in subsequent chapters. Three time periods are considered for future climate projections: the near future (20152035), the medium future (2046-2099) and the long term future (2080-2099); abbreviated to 2020s, 2050s and 2090s respectively. The estimated climate data in each of the three time 27 periods can give an idea of future changes in precipitation and temperature, and with the use of crop models, future changes in food production can be estimated. 4.2 Long Ashton Research Station Weather Generator (LARS-WG) This study utilizes Long Ashton Research Station Weather Generator (LARS-WG Version 5.0) to simulate future weather data for Malawi, with Lilongwe District as a case study. LARS-WG is implemented in C++ with full Windows interface and is freely accessible at http://www.rothamsted.ac.uk/mas-models/larswg.php. The model simulates weather data at a single site under current and future conditions (Racsko et al., 1991; Semenov et al., 1998; Semenov and Brooks, 1999). LARS-WG Version 5 also includes fifteen (15) General Climate Models (GCMs) which have been used in the IPCC 4th Assessment Report (2007). Each of the models comprises at least three of the following scenarios: SRA1B, SRA2 and SRB1, for three time periods (2020s, 2055s and 2090s). The simulated data from the model are in form of daily time-series for the following climate variables: maximum and minimum temperature (°C), precipitation (mm) and solar radiation in Mega joule per square meter day (MJm-2day-1) (Semenov and Barrow, 2002). Based on the model developer (Semenov and Barrow, 2002: 2010), a brief description of the model procedure in simulating weather data is provided in the subsequent section. 4.2.1 Input Data for LARS-WG The following weather data (in time series) are required as inputs to the model; (i) Minimum temperature (°C) (ii) Maximum temperature (°C) (iii)Precipitation (rainfall) (mm) (iv) Solar radiation (MJm-2day-1) 28 In the absence of solar radiation, the model accommodates the use of sunshine hours. LARS-WG automatically converts the sunshine hours to solar radiation using an algorithm which was described by Rietveld (1978). The Model will either work with precipitation (rainfall) only, or precipitation and any combination of the aforementioned climate variables. Other important input information to be specified are the name of the station from which historical observed data was obtained and the location of the weather station in form of latitude, longitude and altitude. The ability of LARS-WG to simulate reliable data depends on the availability of observed data. The model simulates future weather data based on as little as a single year of observed weather data. However, the more observed daily weather data used, the closer to a true value are the simulated results. Semenov and Barrow (2002) recommend the use of daily weather data of at least between 20-30 years for better results. Weather data for long periods are significant in a way that they capture some of the less frequent events like droughts and floods. In this study, daily historical observed weather data for a period from 1971-2000 was used as baseline data. 4.2.2 Outline of the Process of Stochastic Weather Generation According to Semenov and Barrow (2002), the process of generating synthetic weather data can be grouped into three distinct steps: (i) model calibration, (ii) model validation and (iii) generation of synthetic weather data. Each of these processes is described herein. 4.2.2.1 Calibration of the Model Calibration is the first step that is executed by the model in order to generate synthetic weather data. Calibration of LARS-WG is carried out by a function on the main menu called “Site Analysis”. The process is done so as to determine statistical characteristics and site 29 parameters of the observed weather data. In our study, site analysis was performed on observed data for a period of 30 years (1971-2000). Once the program encounters “illegal data” during execution, an “error” is displayed. “Illegal” data includes the value of minimum temperature being greater than maximum temperature and non-negative precipitation values. Examples of statistical characteristics of weather data in the output file include the following; (i) Empirical distribution characteristics for the length of wet and dry series of days in the observed data, (ii) The mean and standard deviation, by month, of wet and dry series length, (iii) Rainfall distribution, maximum, minimum, number of observed years (months or days), mean and standard deviation. (iv) Maximum, minimum, number of observed years (months or days), mean and standard deviation of temperature. (v) Periods of cool and warm weather (vi) Maximum, minimum, number of observed years (months or days), mean and standard deviation of solar radiation. Examples of weather generated site parameters in the output file include the following; (i) Histogram intervals and frequency of events in the intervals for precipitation amount by month from January to December. (ii) Fourier coefficients for means and standard deviations of minimum and maximum temperature on wet and dry days 30 (iii) Average autocorrelation value for minimum and maximum temperature and solar radiation. (iv) Solar radiation amount (MJm-2day-1) on wet and dry days by month from January to December. 4.2.2.2 Model Validation Once LARS-WG has been calibrated, its ability to simulate future weather data in the representative study site is assessed. Validation is a process that is used to determine how well a model can simulate potential future climate variables. The process involves comparing and analyzing the statistical characteristics of the observed and synthetic weather data in order to determine the existence of any statistically-significant differences between them. Validation of the model can be conducted in two different ways: (i) using the GENERATOR option to synthesize daily weather data based on the information in the site parameter files and then undertake comparisons between the observed and synthetic data ‘off-line’, or (ii) using the Q-test option that executes statistical comparisons between climate parameters derived from observed weather data and synthetic weather data generated using LARS-WG. The Q-test achieves this using two options: “Test” and “Compare”, both of which compares the probability distributions for the synthetic and observed data using the Chi-square goodness-of-fit test and the means and standard deviations using the t and F tests (Semonov and Barrow, 2002). Using the “Test” option, the user generates synthetic data for any number of years; 300 years is set as the default value. The generated synthetic data are then analyzed and parameter files produced. The parameter files contain information about probability distribution, mean and standard deviations of the synthetic data. Results from “Compare option” compares the statistical characteristics of the observed data with those of simulated data generated from the 31 parameter files calculated from the observed data. The output parameters from this function are (i) the degrees of freedom, Chi-squared values and probabilities for wet and dry series and monthly distribution of precipitation, and (ii) observed/weather generated mean and standard deviation, t- and F- values and probabilities for maximum and minimum temperature, rainfall and solar radiation both in daily and monthly formats (Semonov and Barrow, 2002). In this study, the Q-test was used in validation of the model due to its simplicity in data interpretation. 4.2.2.3 Generation of Synthetic Weather Data Once LARS-WG has been calibrated (Site Analysis) and the performance of the weather generator has been verified (Q-test), the “Generator” option then generates synthetic weather data. Synthetic weather data generated from this option have the same characteristics as observed weather data. This option also enables one to generate synthetic weather data analogous to a scenario of climate change. The following output data is obtained from this function; (i) Relative change in monthly mean rainfall (ii) Relative change in duration of wet and dry spell (iii) Absolute change in monthly mean minimum temperature (iv) Absolute change in monthly mean maximum temperature (v) Relative changes in daily temperature variability (vi) Absolute change in monthly mean radiation To approximate the probability distributions of wet and dry series, and variables like daily precipitation, maximum and minimum temperature and solar radiation, LARS-WG uses a semi-empirical distribution (SED) also known as probability distribution function (PDF). Due to an increase in number of intervals of this version of LARS-WG (from 10 in the previous version to 23), there is greater probability of obtaining output data that is accurate. For each climatic 32 variable v, there exists a value for climate variable vi that corresponds to probability of occurrence pi, and can be estimated as follows (Semenov and Barrow, 2010); where P ( ) signifies the probability based on observed climate variable denoted as vobs. Each climate variable has 2 fixed values of probability, po and pn, where po = 0 and pn = 1, and corresponding values of vo = min {vobs} while vn = max {vobs}. To precisely estimate the extreme values of climate variable, extremely low and high values of the variable are assigned a probability value pi that is close to 0 and 1 respectively. The remaining values are evenly distributed on a probability scale. LARS-WG version 5 enables one to easily calculate the SED for climate variables. Some climate variables are believed to follow an annual circle. For instance, higher values of variables like temperature and radiation are reported to be higher in summer and lower in winter particularly in the Northern Hemisphere. A smooth seasonal curve for daily and maximum or minimum temperature and radiation is produced by calculating the SED. To estimate the SED for a given day, the model interpolates SEDs between two months. For instance, if we want to calculate SED, denoted Dk for any day within the year denoted k. If the middle day of the month m is denoted km, and assuming [k-km] <=15. Distributions for 3 months denoted Mm-1, Dm and Dm+1 are used to calculate Dk. To preserve the universality, a reasonable assumption that km-1< k< = km is made. The resulting distribution Dk, for a day k, is a weighted sum of 2 distributions namely Dm-1 and Dm and is written as: ( where ) N (m) is the number of days in a month m. α is meant to reduce the difference between mean value Em of distribution Dm, and the mean value Ēm, for the 33 month m, calculated from values generated from interpolated distributions Dd for each day of the month. Ēm is thus given in equation 3 below. ∑ ⁄ α is selected to satisfy the following conditions; | | where Av is an absolute acceptable error for variable v1 and Rv is a relative error as a proportion to Ēm. Having calculated Ēm, from equations (1) through (4), equation (4) can be rearranged as follows; | | where ∆m- = Em- Em-1 and ∆m+ = Em+1-Em and α is limited as follows: 1 ≤ α ≤ 4. When Em is a local minimum or local maximum, meaning both ∆m- and ∆m+ have different signs, we use an unmodified distribution Dm, for all days k, where | | ≤ 15 (Semenov and Barrow, 2010). 4.3 GCM’s Emission Scenarios In chapter 2, we saw that GCMs are significant in estimation of future climate data. To estimate future climate data, GCMs require emission scenarios. Each scenario is created based on assumption of future concentration of greenhouse gasses in the atmosphere, particularly carbon dioxide (CO2). By selecting a model and a scenario, LARS-WG generates daily climate variables for the site under study. The magnitude of the output variables depends on the selected model-scenario. Output data from this process includes daily maximum and minimum temperature (oC), daily rainfall (mm) and daily radiation (MJm-2day-1). In this work, we use the outputs from 15 GCMs under two emission scenarios: SRA1 and SRB1, as recommended by IPCC Special Report on Emission Scenarios (SRES) (Houghton et al., 2001). Under SRA1, SRES assumes a future world of very rapid economic growth, global 34 population that peaks in mid-century and declines thereafter, and rapid introduction of new and more technologies. B1 scenario assumes a convergent world with the same global population as in A1 storyline, but with rapid changes in economic structure towards a service and information economy, and reduction in materials intensity and the introduction of clean and resources efficient technologies (IPCC-SRES, 2007). Table 8 shows the key assumptions of the SRES emissions scenarios and corresponding increases in CO2 concentrations for three different time periods. Table 8. CO2 concentrations in parts per million (ppm) for the four main climate scenarios (SRES). CO2 for base line scenario (1960-1990) is 334ppm (Nakicenovic and Swart, 2000) CO2 concentration, ppm Scenario Key assumptions A1B “the rich word” A world characterized by very rapid economic growth (3%/yr), low population growth (0.27%/yr) and rapid introduction of new and efficient technologies. There is global economic and cultural convergence and capacity building. Cultural identities separate different regions making the world more heterogeneous and less cooperative. Emphasis is put on family values, local traditions and high population growth (0.83% per year) with little emphasis on economic growth (1.65% per year) and material wealth. Rapid change in economic structure, dematerialization, improved equity more global concern regarding environmental and social sustainability, with introduction of clean technologies. Global population reaches 7 billion A heterogeneous society with emphasis on local solutions to economic, social and environmental sustainability as opposed to global solutions. High priority is given to human welfare, equality and environmental protection. A2 “The separated world” B1 “The sustainable world” B2 “the world of technological inequalities 35 20152035 20462065 20812100 418 541 674 414 545 754 410 492 538 406 486 581 4.4 Uncertainty Analysis of GCMs We have discussed thus far, that due to uncertainties associated with GCMs, it is necessary that the uncertainties are dealt with accordingly. This procedure is significant because the uncertainties affect the credibility of the outputs, and the subsequent impact assessment studies. Outputs from GCMs are often an extreme range of low and high values of climate variables, rendering the results impracticable for further assessment studies (Reilly et al., 2003; Cai et al., 2009). Three major factors contribute to uncertainty of GCMs: (i) differences in emissions scenarios derived from projections of economic activity, population growth, and technological development; (ii) sensitivity of global climate to greenhouse gas forcing; (iii) regional variability, which occurs between models of different regional responses (RRs) and within models of chaotic behaviors and modes of climate variability (Huntingford et al., 2005; Cai et al., 2009). Numerous methods for dealing with uncertainties of GCMs have been proposed, some of which include (i) using a central prediction with error bars, (ii) expressing the results as a central prediction, (iii) using a bounded range with known probability distribution and (iv) by using a bounded range with larger range of unknown probabilities (OECD, 2003; Gohari et al., 2013). In our study, we use a bounded range with known probability distribution (Gohari et al., 2013) to deal with the uncertainties of fifteen (15) GCMs because it is the most used technique. This technique requires the execution of three steps which are described below. 4.4.1 Step 1: Weighting of GCMs The first step of this technique involves weighting each of the 15 GCMs used in the study based on the Mean Observed Temperature-Precipitation (MOTP) method (Equation 6) (MassahBavani and Morid, 2005; Gohari et al., 2013). In order to weight each GCM, the ability of the 36 model to project weather data is considered. In other words, the method considers the monthly average difference between observed and simulated climate variables (precipitation and minimum and maximum temperature). ⁄| ( ⁄| | |) where Wij is the weight of GCM j in month i; and | | is the absolute difference between the average precipitation (rainfall) or temperature between observed value and the value simulated by GCM j in month i. 4.4.2 Step 2: Generation of Probability Distribution Functions (PDFs) This step entails generation of PDFs of changes in climate variables based on the calculated weights. The PDFs outline the relationship between the weight of each GCM and the average changes in monthly precipitation, minimum temperature and maximum temperature. With 15 GCMs and 2 emission scenarios used in this work, 30 PDFs are thus constructed for each month. The generated discrete PDFs of the main variables are ultimately converted to cumulative probability functions (CDFs). Several studies have identified the use of Gamma distribution function as an important tool for analysis of climate data (Ines and Hansen, 2006; Block et al., 2009; Piani et al., 2010; Gohari e al., 2013). Based on similar studies that were carried by Pindyck (2012), Teutschbein and Seibert (2012) and Gohari et al. (2013), the Gamma function has been selected for generation of cumulative distribution functions as follows; where α and β are shape and scale parameters of the Gamma distribution function respectively, -x 37 is the climate variable (temperature or precipitation), and is the incomplete Gamma function as given in Equation (8). ∫ By changing values of α and β, we obtain the best fit based on maximum likelihood model. The summation of squared error (Equation 9) has been used to show how best the Gamma function fits the data. ∑ where yi is the data point; ¥i is the estimation of Gamma function and n is the number of data points. For this study, n = 15. 4.4.3 Step 3: Generation of Cumulative Distribution Functions (CDFs) In this step, the PDFs generated in the previous step are converted to CDFs for each of the 12 months (January-December). Next, values of climate change variables at three probability percentiles are extracted from the generated CDFs at the following risk levels: 25th, 50th, and 75th probability percentile. The calculated climate variables at three different percentiles are then used as input data for modeling of potential corn production with AquaCrop. The 25th probability percentile indicates a scenario of high changes in precipitation and low temperature changes. The 75th probability percentile represents a scenario with low changes in precipitation but high temperature changes. The 50th probability percentile is the median probability percentile for both precipitation and temperature. 4.5 Crop Yield Estimation with AquaCrop 4.5.1 Introduction The complexities associated with quantification of crop responses to water deficit led to the introduction of empirical production functions as a realistic approach to assessing crop 38 response to water availability (Steduto et al. 2009). In the FAO Irrigation and Drainage Paper no. 33, Doorenbos and Kassam (1979), presents an empirical function approach for assessing the yield response of tree crops and field crops using the following equation: where Yx and Ya are maximum and actual yield, ETx and ETa represent maximum and actual evapotranspiration and Ky is the proportionality constant between relative yield loss and relative reduction in evapotranspiration. The FAO Paper number 33 was later revised to a framework that differentiates field crops from tree crops. This led to the development of a new field crop model called AquaCrop which evolved from Equation (10), and was designed for planning, management and scenario simulations of field crops. The model’s balance between accuracy, simplicity and robustness renders it preferable over other models (Steduto et al., 2009). In this section of the study, we discuss the conceptual framework, design, structure and unique features of AquaCrop version 4.0. 4.5.2 Description of the Model As stated in 3.5.1 above, AquaCrop originated from an empirical function approach as depicted in Equation (10) by separating actual evapotranspiration (ETa) into soil evaporation (Es) and crop transpiration (Ta); and the final yield (Y) into biomass (B) and harvest index (HI). The separation of ETa to Es and Ta is important particularly during incomplete ground cover where Es is not applicable. On the other hand, separating Y into B and HI is necessary as it enables one to distinguish the functional relationships that exist between B and environment and HI and environment. These changes led to the development of a linear equation (Equation 11), that is referred to as the “AquaCrop growth engine” (Steduto et al., 2009). ∑ 39 where WP is the parameter for water productivity (kg of biomass per m2 per mm of water transpired over the period when production of biomass occurred) and Ta is crop transpiration (mm). Steduto et al. (2009) also explains that the transition from Equation (10) to Equation (11) improved the robustness of the model and time scale. The former is used for long periods like seasons normally in months, whilst the latter is used for daily time steps, a period that is essential for assessing crop response to water deficits. Figure 9 is a flowchart showing the functional relationships of the soil-plant-atmosphere continuum of AquaCrop. Figure 9. Main components of soil-plant-atmosphere continuum in AquaCrop (Steduto et al., 2009) As shown in Figure 9 above, the structure of AquaCrop encompasses a complex relationship amongst soil, plant and atmospheric continuum. The model comprises the soil and its water balance; the plant, which includes crop development, growth and yield processes; and 40 the atmosphere that includes temperature, rainfall, evaporative demand and carbon dioxide concentration. AquaCrop also incorporates some management aspects including irrigation, fertilizer application and control of field runoff through ridging and mulching (Steduto et al., 2009). These relationships are discussed in more detail in the following sub-sections. 4.5.2.1 The Atmosphere: Required Input Climate Data To run AquaCrop, five key meteorological input variables are required: (i) daily rainfall, (ii) daily minimum air temperature (iii) daily maximum air temperature (iv) daily evaporative demand of the atmosphere expressed as reference evapotranspiration (ETo), and (v) mean annual carbon dioxide (CO2) concentration in the bulk atmosphere. The first four parameters are obtained from a meteorological station; while the mean CO2 used is from Mauna Loa observatory records in Hawaii (Raes et al., 2012). Apart from daily time series, the first four parameters may also be provided at 10 day or monthly time scales. At run time, AquaCrop down scales the 10 day or monthly time scales into daily values (Gommes, 1983). In addition, AquaCrop does not include an algorithm for calculation of ETo, thus a separate software, FAO ETo Calculator (version 3.2), is used to calculate ETo (Raes et al., 2008; FAO, 2009). The procedure followed for ETo calculation is identical to that outlined in the FAO Irrigation and Drainage Paper 56, using the FAO Penman-Monteith equation (Allen et al., 1998) which is given by: where ETo reference evapotranspiration (mm day-1) Rn net radiation at the crop surface (MJm-2day-1) G soil heat flux density (MJ m-2 day-1) T mean daily air temperature at 2 m height (oC) 41 u2 wind speed at 2 m height (ms-1) es saturation vapor pressure (kPa) ea actual vapor pressure (kPa) es-ea saturation vapor pressure deficit (kPa) ∆ slope vapor pressure curve (kPa oC-1) γ psychometric constant (kPa oC-1) Figure 9 indicates that temperature influences crop phenology, rainfall and ETo are significant inputs into the water balance model, whilst CO2 concentration of the bulk atmosphere has an influence on water productivity (WP) and crop growth rate (Steduto et al., 2009). 4.5.2.2 The Crop Figure 9 outlines five major components associated with the crop system namely: phenology, aerial canopy, rooting depth, biomass production and harvestable yield. As the crop grows, its canopy expands and its roots are simultaneously deepened, thereby establishing main development stages. The canopy represents the source of actual transpiration which is used to determine the amount of biomass produced from the water productivity (WP). Using the harvest index (HI), the actual harvestable part of biomass, the yield, is calculated using Equation (13) (Steduto et al., 2009). where, Y is the yield, B is biomass and HI is harvest index. The harvest index is essentially a percentage of biomass that is translated into yield. Steduto et al. (2009) states that canopy is the most significant component of AquaCrop, because the amount of water transpired from the crop determines biomass production and hence 42 the yield. Its expansion is expressed as a percentage of green canopy ground-cover (CC), such that where there is no water stress for the crop, canopy expansion from emergency to full development follows exponential growth and exponential decay during the first and second halves of full development respectively as shown in Equations 14 and 15. where CC is canopy cover at time t, CCo is the canopy cover at t=0, CGC is the canopy growth coefficient in fraction per day or per degree day, CCx is the maximum canopy cover and t is the time in days or degree days. The growing degree days (GDD) which is the number of temperature degrees that determine a proportional crop growth and development are calculated using Equation 16: where Tmax and Tmin are maximum and minimum temperatures and Tbase is the temperature at which crop growth stops. 4.5.2.3 The Soil AquaCrop allows configuration of up to 5 layers of soil of different texture in its profile. All the soil texture classes found in USDA triangle are set as default in the model; however, the user has an option to input their own values applicable to the soil under study. Each soil texture class is associated with specific hydraulic properties, and the model estimates the hydraulic characteristics of the soils entered by the user using pedotransfer functions. The user can also specify values of hydraulic conductivity of the soil(s). The model requires the following hydraulic characteristics: drainage coefficient (τ), the hydraulic conductivity at saturation (ksat), 43 the volumetric water content at saturation (θsat), field capacity (θFC), and wilting point (θWP) (Steduto et al., 2009). 4.5.2.4 The Management The management section of the model comprises two parts: (i) a more general field management and (ii) a more specific water management. Field management considers practices that could be applied to a farm with an aim of realizing the desired harvest/yield. Examples of field management include using field surface practices like mulching to reduce soil evaporation or using soil bunds to reduce surface runoff, both of which subsequently enhance infiltration. Field management also involves managing the fertility levels of the soil. This option considers three fertility levels namely: non-limiting, medium, and poor fertility. These fertility levels affect the water productivity (WP), fraction of canopy growth coefficient (CGC), maximum canopy cover (CCx), and rate of decline of canopy during senescence (Steduto et al., 2009). The options considered in water management include either rainfed agriculture (with no irrigation) or irrigation (in which case the user selects method of irrigation i.e sprinkler, drip, surface or furrow). The irrigation part of the model is used to simulate the crop response under supplemental or deficit irrigation. 4.5.3 Algorithm and Calculation Procedure of AquaCrop The algorithm and calculation procedures of AquaCrop place more emphasis on the soil water balance (Steduto et al., 2009). The model uses a set of equations to simulate soil water movement and root water uptake. The equations describe the dependent variable θ (soil water content), using dimensionless drainage coefficient τ, which is obtained from Ksat (hydraulic conductivity at saturation) for vertical flow of water in the soil profile (Raes et al., 2006). 44 The soil profiles in AquaCrop are grouped into smaller fractions and this is done in order to nearly accurately describe the movement, retention and water uptake of the soil during the growing season. In total, the model comprises 12 soil compartments each of which measure 0.1m thick (Az). Depending on the crop and soil type, the size of each compartment is adjusted to cover the entire root zone. Each of these soil compartments has different hydraulic characteristics depending on the type of soil associated with a particular compartment (Steduto et al., 2009). In the following subsections, we discuss soil drainage, surface runoff and infiltration, soil evaporation, crop transpiration and biomass production in context of AquaCrop model. 4.5.3.1 Drainage Drainage of water in each soil compartment occurs when water content θ, is above field capacity, FC. The amount of water that drains from compartment i, at time step ∆t, is expressed by exponential drainage function, which considers water content θi and the drainage properties of the soil (Equation 17) (Raes, 1982; Raes et al., 1988; 2006) is the decrease in soil water content θ during ∆t (m3m-3d-1), τ is the drainage where coefficient, θsat is volumetric water content at saturation, and θFC is the field capacity. As the value of τ becomes close to 1, a fully saturated soil will drain the fastest (approximately between 1 to 2 days); conversely, smaller values of τ (τ <1) decreases water drainage capabilities of the soil. Numerous studies indicate that AquaCrop perfectly mimics the infiltration and drainage patterns that occur in in situ field observations (Descheemaeker, 2006; Raes et al., 2006; Geerts 45 et al., 2008). To simulate drainage in a soil profile containing many and different compartments, the drainage ability of each compartment is considered. An assumption is made that the total drainage occurring in soil compartments at the top, will pass through lower compartments so long the compartments at the bottom have greater or equal drainage characteristics with compartments at the top. If a compartment at the bottom has less drainage capabilities than the one at the top, excess water is store within the compartment thereby increasing its water content and subsequently its drainage ability. 4.5.3.2 Surface Runoff and Infiltration To simulate surface run off, AquaCrop employs the Curve Number Method proposed by USDA (1964), Rallison (1980) and Steenhuis et al. (1995). The Curve Number, abbreviated as CNII, is adjusted according to the wetness of the top layer of the soil. Smedema and Rycroft (1983) uses some derived relationships for curve number values for “antecedent moisture class” (AMC) I (dry), II and III (wet), in which a soil depth of about 0.3m is considered for determining AMC. For simplicity, surface runoff resulting from crop irrigation is considered to be zero. If the surface runoff through irrigation is significant, the model requires that the user inputs the net water application. On the other hand, water infiltration (through irrigation or rainfall) into the soil compartments is limited by the saturated hydraulic conductivity (Ksat) of the soil in the top layer. Depending on the drainage characteristics of the soil compartments, storage of water occurs simultaneously with drainage, from top to bottom compartments (Steduto et al., 2009). 46 4.5.3.3 Soil Evaporation In AquaCrop, soil evaporation occurs in two stages: an energy limiting stage (Stage I) and a falling rate stage (Stage II) (Philip, 1957; Ritchie, 1972). In stage I, the readily evaporable water (REW) is estimated using Equation 18. is the soil water content (m3m-3) when air where θFC is water content at field capacity, dried, and is the thickness (m) of the evaporating surface. The U value of Ritchie (1972) that binds the relationship between REW and cummulative evaporation for Stage I, was determined by assuming that the value of point θPWP, and equal half the value of permanent wilting equals 40mm. Thus evaporation is limited to the energy limiting stage, so long REW remains in surface layer, and the rate of evaporation is the maximum rate, Ex. Howerer, when REW is depleted in Stage I, then evaporation occurs in the falling rate stage (Stage II). This results in flow of water from bottom to surface layer of the soil (Ritche, 1972, Raes, 2009). Soil evaporation for both energy limiting and falling rate stages is determined using Equation 19. where, Ex is the maximum rate of evaporation, is evaporation coefficient for fully wet and bare soil surface (default value set at 1.10), ETo is reference evapotranspiration, and Kr is the dimensionless evaporation reduction coefficient. For stage I, Kr is set at 1, but decreases to values lower than 1 as evaporation switches from stage I to stage II. As soil water content diminishes, so does its hydraulic conductivity. To account for the decline in hydraulic conductivity, the model uses an exponential function as shown in Equation 20 (Allen et al., 1998). 47 where fK is the decline factor, and Wrel is the relative water content of the soil layer. Ritchie (1972) discovered that when fK = 4, there exists a “good fit between the square root of time approach and the soil water content approach used by AquaCrop in the simulation of stage II evaporation”. Kr varies sturdily with Wrel and is estimated using Equation 20 above. Canopy cover impacts on the magnitude of evaporation. For a soil that is covered by crop canopy, the rate of evaporation (Ex) in the energy limiting stage (stage I) is estimated using Equation 21. where CC* is the portion of soil surface covered and adjusted for micro-advective effects, and is estimated using Equation 22 (Adams et al., 1976; Villalobos and Fereres (1990). As canopy cover (CC) senesces (grows old) due to phenology or water stress as a result of Ex, Ex is estimated using Equation 23. ( ) where fcc is the adjustment factor expressing the sheltering effect of the dead canopy cover, and CCtop is the canopy cover before the senescence. At maximum canopy cover, CCtop = CCx, the maximum canopy cover. 4.5.3.4 Crop Transpiration For a well-watered field, crop transpiration is proportional to the canopy cover and is estimated using Equation 24. 48 where CC* is the adjusted canopy cover as in Equation 22, and Kctr,x is the coefficient for maximum crop transpiration. After maximum canopy cover (CCx) has been attained, the crop begins to age slowly, consequently transpiration declines. To account for this phenomenon, an adjustment factor fage is multiplied to , leading to the adjusted coefficient of transpiration . As senescence continues, crop photosynthesis and transpiration remarkably drops, and another adjustment factor that accounts for scenescence fsen, is applied to (fsen =1 at start of scenescence (when CC=CCx) and 0 where no canopy exists). These relatioships are shown in Equations 25 to 27. where ( ) and, where α is a program parameter and is used to decrease (α<1) or increase (α>1) the reduction efficiency of transpiration and photosynthesis of the aging canopy (Raes et al., 2009). 4.5.3.5 Biomass Production According to Steduto et al. (2009), biomass (B) above the ground is derived from transpiration through crop water productivity (WP) normalized for reference evapotranspiration (ETo) and carbon dioxide (CO2). Considering effects of low temperatures, chemical properties of harvestable parts of the plant, and modifications of dimensionless factors, the daily above ground biomass (m) is estimated using Equation 28. ( ) 49 where Ksb is the stress coefficient accounting for effects of low temperature on biomass production, fWP is the adjustment factor (fWP ≤1) accounting for differences in chemical composition of biomass and harvestable parts, Tr and ETo are crop transpiration and reference evapotranspiration respectively. 4.5.3.6 Response of Harvest Index (HI) to Water Stress As discussed in 3.5.2.2, final crop yield (Y) is the product of final biomass (B) and harvest index (HI). Steduto et al. (2009) outlines the effects of water stress on HI. As flowering or formation of tuber commences, HI is programmed to increase until plant maturity. The value of HI at plant maturity under no stress conditions is referred to as reference harvest index (HIo) for a specific crop type. The impact of water stress on HI depends on its severity and time of occurrence, with the reproductive stage being the most crucial phase. However, if crop growth continues, water stress may lead to a reduction in leaf growth thereby increasing the HI. Equation 29 indicates that the rate of increase (dHI/dt) of harvest index (HI) is expedited when the maximum canopy cover is reached and the value of Ksexp is less than 1. ( where, ( ) ( ) ) is the rate of increase of HI for non stress conditions, is the adjustable crop parameter, and Ksexp,t is the adjustment factor for the competition between vegetative and reproductive growth after flowering begins at time t. Even though growth of canopy cover is complete, Equation 29 is still applicable so long as crop growth continues. Crop response to water stress includes reduction in leaf growth and closure of stomata. This slows down the increase in HI whose rate of increase is described in Equation 30. √ ( ) ( ) 50 where b is the adjustable crop parameter, and is the adjustment factor for stress that leads to stomata closure and reduction in photosynthesis. When water stress is so severe to allow pollination to occur, harvest index is represented by the relationship shown in Equation 31 below. [∑( where, )] is the (31) adjusted for the reduction in pollination caused by stress, Kspol is the water stress coefficient for pollination on a given day, F is the fraction of the total number of potentially successful flowers going through anthesis on that day, and is a factor allowing for the effects of excessive sinks (Raes et al., 2009). 4.5.4 Calibration of AquaCrop Calibration of AuaCrop was achieved by comparing the recorded/actual corn production (as provided by the Department of Agriculture in Ministry of Agriculture and Food Security in Malawi) with the corn yield simulated by the model. Input data detailing various crop parameters used in the model were obtained from Chitedze research station. By comparing the actual and simulated corn yields, we were able to adjust crop parameters, field management and soil characteristics through trial and error, until a closest match between recorded and simulated corn yield was achieved. Data (recorded corn yields, rainfall, minimum and maximum temperature, wind speed at 2m above ground, mean relative humidity and daily number number of sunshine hours) for Chitedze station for year 2005 was used to calibrated the model. Table 9 presents some of the most important crop input parameters that were used in calibration of the model (Abedinpour et al., 2012). 51 Table 9. Input crop parameters used in calibration of AquaCrop Description of parameter Base temperature Cut-off temperature Canopy growth coefficient (CGC) Canopy decline coefficient (CDC) at senescence Leaf growth threshold (Pupper) Value used 10 40 20.9 0.948 0.16 Leaf growth threshold (Plower) Leaf growth stress coefficient curve shape Expansion stress coefficient (Pupper) Expansion stress coefficient (PLower) Expansion stress coefficient curve shape Stomatal conductance threshold (Pupper) Stomatal stress coefficient curve shape Senescence stress coefficient curve shape Senescence stress coefficient (Pupper) 0.81 2.8 Coefficient, inhibition of leaf growth on HI 8.0 Coefficient, inhabitation of stomata on HI 4.5 Maximum basal crop coefficient (Kcb) Time from sowing to emergence Time to maximum canopy cover Time from sowing to start flowering Time from sowing to start senescence Time from sowing to maturity Length of the flowering stage Length of building up HI 1.25 8 52 70 98 127 10 52 0.18 0.63 1.2 0.65 1.9 1.6 0.46 Unit of parameter o C o C % day−1 % day−1 % of TAW (fraction of total available water (TAW)) % of TAW Unit less (moderately convex curve) % of TAW % of TAW % of TAW Unit less Unit less (high convex curve) Unit less (moderately convex curve) Unit less (Initiation of canopy senescence) Unit less (HI increased by inhibition of leaf growth at anthesis) Unit less (HI increased by inhibition of stomata at anthesis) Unit less Days Days Days Days Days Days Days 4.5.5 Validation of AquaCrop Having calibrated AquaCrop, it was significant that the model be validated in order to evalute its performance in simulating crop yields. Model validation is important in order to determine if the model has the ability to replicate the data, to analyse the effectiveness of model calibration and compare synthetic data with those done in previous studies (Krause et al., 2005). 52 Loague and Green (1991) indicates that there are numerous statistical indicators for evaluating perfomance of AquaCrop, nonetheless, Willmott (1984) argues that each of the statisital indicators have their own weaknesses and strengths. To effectively evaluate the performance of the model, the use of ensemble statistical indicators is appropriate (Willmott, 1984). In our analysis of performance of AquaCrop, the following list of statistical indicators were used: prediction error (Pe), coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE) and model efficiency (E). E and R2 indicate the predictive power of the model whilst Pe, MAE and RMSE is used to signify the amount of error associated with the model prediction (Abedinpour et al., 2012). Calibrated crop paramaters, management practices and soil types and conditions remained constant. Corn yields for years 2000 to 2004 were used to compare with those simulated by the model within the same time period. Equations 32-36 show the statistical relationships that were used to evaluate the performance of the model based on its predictive prowess and amount of error associated with it (Nash and Sutcliffe, 1970). Equation 32 is a relationship used to estimate, R2, which is the squared values of Pearson colleration coefficient that ilustrates squared ratio between covariance and the multiplied standard deviations of the observed and estimated corn yields. ∑ [ √∑ ̅ ̅ ̅ ∑ ̅ ] 53 ∑ ∑ ( ) where, Si and Oi are synthetic and actual (observed) production, ̅ i is the mean value of Oi and N is the number of observtions. √ ∑ √∑ The model is said to perform better when values of E and R2 approaches one and when values of Pe, MAE and RMSE approaches zero (Moriasi et al., 2007). 54 CHAPTER FIVE: RESULTS AND DISCUSSION 5.1 Introduction This chapter presents the results of the study in order to meet the stated objectives. First, the results of the estimated future climate variables using GCMs, and statistical downscaling of the outputs using LARS-WG, are presented. Results from statistical comparison between historically observed and simulated weather data and those derived from model calibration are also presented. Third, outcomes of uncertainty analysis of GCMs in climate projections are discussed. Ultimately, we discuss the calibration and validation procedure of the crop model (AquaCrop) and present the results of future simulated corn yields relative to the baseline period. 5.2 Estimation of Future Climate Variables 5.2.1 Calibration and Validation of LARS-WG The Q test function was used to determine the ability of Lars-WG to rationally estimate future climate variables. This was achieved using two statistical tests; chi-square test (X2) and ttest. The chi-square test was used to determine the existence of any significant difference between the expected and observed frequencies in the meteorological data. Additionally, a t-test was used to check the existence of any reliable difference between the means of the estimated and observed data sets. Large X2 and t values indicates the existence of real difference between observed and estimate climate variables. Conversely, smaller X2 and t values indicate that there is less difference between observed and estimated data sets. Each X2 and t value has a corresponding p-value, which is the probability that the pattern of data in the sample could be produced by random variables. A p-value of 0.05 simply means there is a probability of 5% that there is no difference between observed and estimate data. P-values below the set significance 55 level indicate that the simulated climate variables are far from the true climate values. For the purpose of this study, a p-value was set at 0.05 which is commonly used in climate studies. Table 10 shows the results of the p-values for both X2 and t-tests for rainfall, minimum and maximum temperature from January to December. Figure 10 plots the monthly p-values of X2 and t-tests for each of the three variables shown in Table 10. The results indicate that p-values in all months for both rainfall and temperature are larger than the selected significance level of 0.05. Thus the model can be said to satisfactorily simulate future climate data. Table 10. Statistical comparison of synthetic and observed weather data using Q test (Significant p-value = 0.05) Rainfall Month Minimum Temperature Maximum Temperature X2 pvalue t pvalue X2 pvalue t pvalue X2 pvalue t pvalue Jan 0.06 1.00 0.30 0.77 0.05 1.00 0.05 0.96 0.05 1.00 0.75 0.46 Feb 0.07 1.00 0.48 0.63 0.05 1.00 0.14 0.97 0.05 1.00 0.26 0.79 Mar 0.06 1.00 0.68 0.50 0.16 0.91 2.21 0.08 0.05 1.00 1.25 0.22 Apr 0.08 1.00 0.25 0.81 0.11 1.00 1.60 0.11 0.11 1.00 0.39 0.70 May 0.22 0.57 0.26 0.80 0.05 1.00 0.43 0.67 0.16 0.91 2.04 0.75 Jun 0.31 0.19 0.33 0.74 0.11 1.00 0.86 0.39 0.11 1.00 0.28 0.78 Jul 0.22 0.60 0.25 0.80 0.05 1.00 0.02 0.98 0.05 1.00 0.94 0.35 Aug 0.39 0.14 0.96 0.34 0.11 1.00 1.31 0.19 0.05 1.00 0.07 0.95 Sep 0.13 0.98 0.13 0.90 0.05 1.00 0.09 0.93 0.11 1.00 2.48 0.36 Oct 0.07 1.00 0.13 0.90 0.11 1.00 1.31 0.19 0.11 1.00 1.78 0.18 Nov 0.06 1.00 1.15 0.25 0.16 0.91 1.85 0.07 0.05 1.00 0.44 0.66 Dec 0.06 1.00 0.39 0.70 0.11 1.00 0.34 0.74 0.05 1.00 0.22 0.83 56 1.2 chi square t-tes sig value P-value (Rainfall) 1 0.8 0.6 0.4 0.2 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month 1.2 P-value (max temp) chi square t-tes sig value 1 0.8 0.6 0.4 0.2 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month 1.2 P-value (Min temp) chi square t-tes sig value 1 0.8 0.6 0.4 0.2 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month Figure 10. Plot of P-values for X2 and t tests for rainfall and temperature 57 5.2.2 Estimating Future Climate Variables Fifteen (15) GCMs were used to project future rainfall and temperature under two emission scenarios (A and B), in three different time horizons (2020s, 2050s and 2090s) relative to a baseline period of 1971-2000. To avoid a mismatch between global and local climate data due to low resolution of GCMs, outputs of GCMs were downscaled using Lars-WG, a stochastic weather generator. Figures 11 to 13 show the simulated mean monthly values of rainfall, minimum and maximum temperature in each time horizon, under each scenario. In general, Figures 11 to 13 indicate a wide range in synthetic climate variables as predicted by each of the 15 GCMs in each emission scenario. For instance, consider the estimated rainfall amount in the month of January estimated under emission scenario B in 2020s (SRB-2020s). Simulated rainfall amount by all GCMs for this month ranges approximately between 210-270mm. This range in magnitude of output data simply confirms the notion that output weather variables from GCMs are associated with uncertainties. This phenomenon recurs in the rest of the months, scenarios and in all time periods both in simulated rainfall and minimum and maximum temperature data. It is therefore significant that such uncertainties are accounted for before outputs of GCMs are used in climate change assessment studies. 58 250 300 50% 75% Observed SRA-2020's 200 150 100 50 Simulated rainfall (mm) Simulated rainfall (mm) 300 0 200 150 100 50 0 Month Month 300 50% 75% Observed SRA-2050's 250 200 150 100 50 Simulated rainfall (mm) Simulated rainfall (mm) 300 0 50% 75% Observed SRB-2050's 250 200 150 100 50 0 Month Month 300 50% 75% Observed SRA-2090's 250 200 150 100 50 Simulated rainfall (mm) 300 Simulated rainfall (mm) 50% 75% Observed SRB-2020's 250 0 50% 75% Observed SRB-2090's 250 200 150 100 50 0 Month Month Figure 11: Box plots showing the simulated mean monthly rainfall for 15 GCMs under scenarios A and B for three different time steps. The box plots indicate the lower (50%) and upper (75%) quartiles, the line between them is the median, and the lower and upper whiskers represent the mean ± standard deviation. The dotted black line shows the historical observed mean monthly rainfall. 59 25 50% 75% Observed SRA-2020's 20 Simulated min temp (oC) Simulated min temp(oC) 25 15 10 20 15 10 5 5 Month Month 25 50% 75% Observed SRA-2050's 20 Simulated min temp (oC) Simulated min temp (oC) 25 15 10 5 50% 75% Observed SRB-2050's 20 15 10 5 Month Month 25 25 SRA-2090's 50% 75% Observed 20 Simulated min temp (oC) Simulated min temp (oC) 50% 75% Observed SRB-2020's 15 10 50% 75% Observed SRB-2090's 20 15 10 5 5 Month Month Figure 12. Box plots showing the simulated mean monthly minimum temperature for 15 GCMs under scenarios A and B for three different time periods. The box plots indicate the lower (50%) and upper quartiles (75%), the line between them is the median, and the lower and upper whiskers represent the mean ± standard deviation. The dotted black line shows the historical observed mean monthly minimum temperature. 60 35 50% 75% Observed Simulated maxi temp (oC) Simulated max temp (oC) 35 SRA-2020's 30 25 20 25 20 35 Simulated max temp (oC) Simulated max temp (oC) 50% 75% Observed Month SRA-2050's 30 25 20 SRB-2050's 25 20 50% 75% Observed Month 35 SRA-2090's Simulated max temp (oC) Simulated max temp (oC) 50% 75% Observed 30 Month 35 SRB-2020's 30 Month 35 50% 75% Observed 30 25 20 50% 75% Observed SRB-2090's 30 25 20 Month Month Figure 13. Box plots showing the simulated mean monthly maximum temperature for 15 GCMs under scenarios A and B for three different time steps. The box plots indicate the lower (50%) and upper quartiles (75%), the line between them is the median, and the lower and upper whiskers represent the mean ± standard deviation. The dotted black line shows the historical observed mean monthly maximum temperature. 61 5.2.3 Uncertainty Analysis of GCMs As indicated in 4.2.2, high uncertainties associated with GCM’s output greatly affect the confidence levels of the results of any climate assessment studies. Several methods have been proposed for dealing with uncertainties of GCMs. This study employs the use of probability assessment of bounded range with known probability distribution in order to deal with uncertainties of the 15 GCMs. This process first involves determination of individual weights of each model generation of probability distribution functions (PDFs) and construction of cumulative distribution functions (CDFs) based on the generated PDFs. 5.2.3.1 Weighting of GCMs Each GCM is weighted according to its potential to simulate climate variables. Table 11 shows the weight of each GCM in simulating future changes rainfall and Table 12 shows the weight of each GCM in simulating future changes temperature in each month. 5.2.3.2 Probability Distribution Functions (PDFs) This is the second process of uncertainty analysis of GCMs. It involves generating monthly PDFs of rainfall and minimum and maximum temperature. Figures 14-16 show sample monthly discrete PDFs for the 15 GMs under scenarios A (SRA) and B (SRB). 62 Table 11. Calculated weight of each GCM in simulating future rainfall Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec BCM2 CGMR CNCM3 CSMK3 FGOALS GFCM21 GIAOM HADCM3 HADGEM INCM3 IPCM4 MIHR MPEH5 NCCCSM NCPCM 0.04 0.00 0.01 0.00 0.01 0.04 0.01 0.05 0.08 0.03 0.00 0.01 0.05 0.01 0.09 0.15 0.03 0.02 0.00 0.03 0.03 0.05 0.27 0.04 0.03 0.00 0.03 0.01 0.02 0.12 0.05 0.11 0.03 0.02 0.01 0.01 0.06 0.01 0.03 0.04 0.06 0.08 0.01 0.05 0.06 0.07 0.02 0.02 0.01 0.00 0.13 0.33 0.03 0.03 0.00 0.06 0.03 0.07 0.02 0.00 0.01 0.00 0.03 0.01 0.04 0.04 0.01 0.11 0.43 0.03 0.02 0.00 0.05 0.03 0.42 0.01 0.01 0.01 0.00 0.02 0.01 0.01 0.02 0.14 0.28 0.03 0.02 0.14 0.15 0.06 0.01 0.08 0.03 0.10 0.32 0.01 0.02 0.00 0.03 0.07 0.05 0.06 0.01 0.03 0.02 0.03 0.08 0.01 0.15 0.00 0.07 0.05 0.12 0.07 0.01 0.05 0.05 0.06 0.02 0.03 0.12 0.84 0.05 0.06 0.30 0.06 0.01 0.09 0.13 0.04 0.01 0.01 0.06 0.01 0.03 0.01 0.01 0.01 0.00 0.01 0.01 0.06 0.14 0.03 0.04 0.01 0.03 0.05 0.04 0.32 0.86 0.23 0.05 0.02 0.02 0.51 0.05 0.05 0.02 0.03 0.11 0.06 0.01 0.05 0.03 0.04 0.03 0.01 0.03 0.01 0.02 0.03 0.01 0.01 0.00 0.02 0.02 0.39 0.03 0.16 Table 12. Calculated weight of each GCM in simulating future temperature BCM2 CGMR CNCM3 CSMK3 FGOALS GFCM21 GIAOM HADCM3 HADGEM INCM3 IPCM4 MIHR MPEH5 NCCCSM NCPCM Jan 0.04 0.01 0.06 0.20 0.01 0.01 0.09 0.01 0.05 0.00 0.00 0.01 0.01 0.02 0.49 Feb Mar Apr 0.17 0.03 0.08 0.02 0.01 0.03 0.07 0.30 0.06 0.13 0.04 0.23 0.04 0.02 0.09 0.03 0.00 0.03 0.08 0.02 0.13 0.03 0.01 0.03 0.24 0.01 0.08 0.02 0.01 0.05 0.01 0.00 0.01 0.02 0.01 0.07 0.03 0.01 0.02 0.06 0.02 0.04 0.04 0.52 0.04 May Jun 0.05 0.09 0.03 0.02 0.04 0.02 0.13 0.11 0.08 0.04 0.05 0.33 0.06 0.03 0.08 0.03 0.22 0.13 0.03 0.02 0.02 0.02 0.12 0.11 0.03 0.02 0.04 0.02 0.02 0.02 July Aug 0.06 0.03 0.02 0.01 0.02 0.02 0.09 0.67 0.04 0.01 0.07 0.02 0.03 0.01 0.02 0.01 0.16 0.04 0.02 0.01 0.03 0.01 0.37 0.13 0.02 0.01 0.03 0.01 0.02 0.02 Sep Oct 0.10 0.06 0.15 0.06 0.06 0.04 0.05 0.04 0.02 0.02 0.05 0.05 0.03 0.05 0.02 0.02 0.07 0.12 0.11 0.18 0.03 0.03 0.08 0.12 0.03 0.02 0.04 0.04 0.17 0.14 Nov 0.13 0.03 0.07 0.04 0.01 0.03 0.12 0.01 0.10 0.24 0.01 0.11 0.02 0.03 0.05 Dec 0.12 0.04 0.06 0.06 0.01 0.04 0.24 0.01 0.08 0.02 0.01 0.02 0.01 0.15 0.14 63 0.30 0.25 Weights of GCMs Weights of GCMs 0.30 2020s-SRA 0.20 0.15 0.10 0.05 2050s-SRA 0.20 0.15 0.10 0.05 - 0.9 0.95 1 1.05 Relative rainfall change 0.8 1.1 0.9 1 1.1 Relative rainfall change 1.2 1.00 1.00 2050s-SRB 0.80 Weights of GCMs Weights of GCMs 0.25 0.60 0.40 0.20 2090s-SRB 0.80 0.60 0.40 0.20 - 0.8 0.9 1 1.1 Relative rainfall change 0.8 1.2 0.9 1 1.1 Relative rainfall change 1.2 Figure 14. Sample Discrete PDFs outlining the relationship between weights of 15 GCMs and monthly changes in relative rainfall. The top two graphs indicate relative rainfall changes in January for scenario A for 2020s and 2050s. The bottom 2 graphs indicate relative rainfall changes in February for scenario A for 2050s and 2090s. 64 0.60 2020s-SRA 0.50 Weights of GCMs Weights of GCMs 0.60 0.40 0.30 0.20 0.10 0.40 0.30 0.20 0.10 - 0.4 0.9 Min temperature change (oC) 1.4 0.4 2.4 4.4 Min temperature change (oC) 6.4 0.6 0.6 2020s-SRB 0.5 Weights of GCMs Weights of GCMs 2090s-SRA 0.50 0.4 0.3 0.2 0.4 0.3 0.2 0.1 0.1 0 0 0.4 0.9 1.4 1.9 Min temperature change (oC) 2050s-SRB 0.5 0.4 2.4 0.6 0.8 1 Min temperature change (oC) 1.2 Figure 15. Sample Discrete PDFs outlining the relationship between weights of 15 GCMs and monthly changes in minimum temperature. The top two graphs indicate temperature changes in March for scenario A for 2020s and 2090s. The bottom 2 graphs indicate temperature changes in April for scenario B for 2020s and 2050s. 65 0.3 2090s-SRA 2050s-SRA Weights of GCMs Weights of GCMs 0.3 0.2 0.1 0.1 0 0 0.4 1.4 2.4 Max temperature change (oC) 0.4 3.4 2.4 4.4 Max temperature change (oC) 6.4 0.4 0.4 2020s-SRB 0.3 Weights of GCMs Weights of GCMs 0.2 0.2 0.1 0 2050s-SRB 0.3 0.2 0.1 0 0 0.25 0.5 0.75 1 Max temperature change (oC) 1.25 0 0.5 1 1.5 2 2.5 Max temperature change (oC) 3 Figure 16. Sample Discrete PDFs outlining the relationship between weights of 15 GCMs and monthly changes in maximum temperature. The top two graphs indicate temperature changes in April for scenario A for 2050s and 2090s. The bottom 2 graphs indicate temperature changes in April for scenario B for 2020s and 2050s. 5.2.3.3 Cumulative Distribution Functions (CDFs) The generated PDFs were converted to CDFs using the gamma distribution function whose parameters alpha (α) and beta (β) were determined from a simple code written in Matlab programming language. We then extracted climate variables at 25%, 50% and 75% probability percentiles from the developed CDFs. Climate values at these three percentiles were then used as input data for corn yield impact assessment with AquaCrop. 66 Figures 17-19 indicate sample monthly CDFs from the generated PDFs, for two scenarios at three different time steps (2020s, 2050, and 2090s). 1 1 2050s SRA Exceedence Probability Exceedence Probability 2020s SRA 0.8 0.6 0.4 0.2 0 0.6 0.4 0.2 0 0.9 1 0.95 1 1.05 Relative rainfall change 1.1 0.7 0.8 0.9 1 1.1 1.2 Relative rainfall change 1 2050s SRB 2090s SRB Exceedence probability Exceedence probability 0.8 0.8 0.6 0.4 0.2 0.8 0.6 0.4 0.2 0 0 0.8 0.9 1 1.1 Relative rainfall change 1.2 0.8 Figure 17. Sample CDFs for rainfall based on PDFs in Figure 14. 67 0.9 1 1.1 Relative rainfall change 1.2 1 1 2090s SRA Exceedence probability Exceedence probability 2050s SRA 0.8 0.6 0.4 0.2 0 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 2 2.5 Change in Tmin(oC) 3 0 6 1 1 2090s SRB Exceedence probability 2020s SRB Exceedence probability 1 2 3 4 5 Change in Tmin(oC) 0.8 0.6 0.4 0.2 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 2 Change in Tmin (0C) 2.5 0 2 4 6 Change in Tmin (oC) Figure 18. Sample CDFs for minimum temperature based on PDFs in Figure 15. 68 8 1 2050s SRA 0.8 Exceedence probability Exceedence probability 1 0.6 0.4 0.2 0 0 0.5 1 1.5 2 2.5 3 Chang in Tmax (oC) 2090s SRA 0.8 0.6 0.4 0.2 0 3.5 0 1 6 1 2020s SRB 2050s SRB 0.8 Exceedence probability Exceedence probability 1 2 3 4 5 Change in Tmax (oC) 0.6 0.4 0.2 0 0.8 0.6 0.4 0.2 0 0 0.5 1 Change in Tmax (oC) 1.5 0 1 2 3 Change in Tmax (OC) 4 Figure 19. Sample CDFs for maximum temperature based on PDFs in Figure 16. 5.2.3.4 Generation of Probability Percentiles The magnitude of the expected changes in rainfall, minimum and maximum temperature at three different probability percentiles (25%, 50% and 75%), were determined from the synthetic CDFs in 4.2.2.3 for both scenarios (SRA and SRB) and in three time steps (2020s, 2050s and 2090s). The results are plotted as shown in Figures 20-22 below. Figure 20 shows the expected changes in future rainfall amounts under two scenarios. The simulated results depict that higher rainfall changes are expected under scenario A than 69 scenario B in all three time periods. The expected changes in monthly rainfall for each time period varies between risk levels. Results indicate increase in rainfall in some months and decrease in rainfall in same months at different risk levels. For example, rainfall is predicted to decrease between may and August ( cool and dry winter), excepting for SRA in 2020s where it is expected to increase. On average, annual rainfall amounts are predicted to decrease by 2% to 17%, 0% to 26% and 3% to 29% in 2020s, 2050s and 2090s respectively. Average annual and overall ranges of estimated rainfall in each time period, under each scenario and risk level, are shown in Table 13. Table 13. Summary of mean annual and overall ranges of estimated climate variables in each time period, at different scenarios and probability percentiles Climate Variable Scenario Percentile 25 A 50 Rainfall Change (%) 25 B 50 75 75 2020s 2050s 2090s 11 -5 -6 -1 -16 -19 -11 -26 -29 -2 0 -3 -9 -12 -17 -17 -23 -29 -17 -26 -29 11 0 -3 Tmin Change (oC) 2020s 2050s 2090s 0.65 1.59 2.64 0.79 1.86 3.03 0.94 2.15 3.46 0.59 1.31 1.84 0.72 1.61 2.32 0.88 1.94 2.87 0.59 1.31 1.84 0.94 2.15 3.46 Tmax Change (oC) 2020s 2050s 2090s 0.58 1.56 2.53 0.75 1.86 3.02 0.94 2.2 3.58 0.52 1.26 1.78 0.68 1.61 2.3 0.88 2.01 2.87 0.52 1.26 1.78 0.94 2.2 3.58 70 Range 1.8 2020s-SRA 2 25% probability percentile 50% probability percentile 75% probability percentile Relative precipitation change Relative precipitation change 2 1.6 1.4 1.2 1 0.8 0.6 0.4 1.8 2020s-SRB 1.6 1.4 1.2 1 0.8 0.6 0.4 Month 1.4 2050s-SRA 1.2 25% probability percentile 50% probability percentile 75% probability percentile Relative precipitation change Relative precipitation change Month 1.4 1 0.8 0.6 0.4 0.2 1.2 25% probability percentile 50% probability percentile 75% probability percentile 0.8 0.6 0.4 0.2 Month 1.4 1.4 2090s-SRA 1.2 Relative precipitation change Relative precipitation change 2050s-SRB 1 Month 1 0.8 0.6 0.4 0.2 25% probability percentile 50% probability percentile 75% probability percentile 25% probability percentile 50% probability percentile 75% probability percentile 0 1.2 2090s-SRB 1 0.8 0.6 0.4 0.2 25% probability percentile 50% probability percentile 75% probability percentile 0 Month Month Figure 20. The estimated future changes in relative rainfall at three probability percentiles. The horizontal dash-dot line indicates no changes in rainfall. The top two plots indicate relative rainfall changes for the 71 near future (2020s), the middle plots indicate relative rainfall changes for the middle period (2050s) and the bottom plots indicate the relative rainfall changes in the long term period (2090s), all under SRA and SRB. 1.6 1.4 Change in min temp (oC) Change in min temp (oC) 1.6 25% probability percentile 50 probability percentile 75 probability percentile 2020s-SRA 1.2 1 0.8 0.6 0.4 1.4 2020s-SRB 1.2 1 0.8 0.6 0.4 Month Month 4.5 3.5 2050s-SRA 25% probability percentile 50% probability percentile 75% probability percentile Changes in min temp (oC) Changes in min temp (oC) 4.5 4 3 2.5 2 1.5 4 2050s-SRB 3.5 3 2.5 2 1 Month Month 4.5 25% probability percentile 50% probability percentile 75% probability percentile 4.5 2090s-SRA Change in min temp (oC) Change in min temp (oC) 25% probability percentile 50% probability percentile 75% probability percentile 1.5 1 4 25% probability percentile 50% probability percentile 75% probability percentile 3.5 3 2.5 2 4 2090s-SRB 25% probability percentile 50% probability percentile 75% probability percentile 3.5 3 2.5 2 1.5 1 Month Month Figure 21. Estimated future changes in minimum temperature at three probability percentiles. The top two plots indicate minimum temperature changes for the near future (2020s), the middle plots indicate 72 1.4 25% probability percentile 50% probability percentile 75% probability percentile 2020s-SRA 1.2 1 0.8 0.6 Change in max temp (oC) Change in max temp (oC) minimum temperature changes for the middle period (2050s) and the bottom plots indicate the minimum temperature changes in the long term period (2090s), under SRA and SRB. 0.4 1.4 2020s-SRB 1.2 25% probability percentile 50% probability percentile 75% probability percentile 1 0.8 0.6 0.4 0.2 Month Month 2050s-SRA 2.7 25% probability percentile 50% probability percentile 75% probability percentile Change in max temp(oC) Change in max temp (oC) 3.2 3.2 2.2 1.7 1.2 0.7 0.2 2.7 2050s-SRB 2.2 1.7 1.2 0.7 0.2 Month Month 2090s-SRA 25% Probability percentile 50% probability percentile 75% probability percentile Change in max temp (oC) Change in max temp (oC) 5.5 5 4.5 4 3.5 3 2.5 2 1.5 1 25% probability percentile 50% probability percentile 75% probability percentile 5.5 5 4.5 4 3.5 3 2.5 2 1.5 1 2090s-SRB 25% probability percentile 50% probability percentile 75% probability percentile Month Month Figure 22. Estimated future changes in maximum temperature at three probability percentiles. The top two plots indicate maximum temperature changes for the near future (2020s), the middle plots indicate maximum temperature changes for the middle period (2050s) and the bottom plots indicate the maximum temperature changes in the long term period (2090s), all under SRA and SRB. 73 Figures 21 and 22 are plots of minimum and maximum temperatures for scenarios A and B under three percentiles at three time periods. Overall, temperature is expected to increase in all cases, with higher temperature changes expected under scenario A than in scenario B. Absolute values for changes in minimum temperature for 2020s, 2050s and 2080s are 0.59oC to 0.94oC, 1.31oC to 2.15oC and 1.84oC to 3.46oC respectively. Similarly, changes in maximum temperature for the same time periods were estimated as follows; 0.59oC to 0.94oC, 1.31oC to 2.15oC and 1.84oC to 3.46oC respectively. A summary of average changes in minimum and maximum temperature for each scenario and time period is shown in Table 13. Figure 23 shows plots of percent changes in rainfall and absolute changes in minimum and maximum temperature under each scenario and risk level. 74 Change in Rainfall (%) 15 10 5 0 -5 -10 -15 -20 -25 -30 -35 2020s A25 A50 2050s A75 B25 2090s B50 B75 Change in min temp (oC) Scenario Percentile 4 2020s 3.5 2050s 2090s 3 2.5 2 1.5 1 0.5 0 A25 A50 A75 B25 B50 B75 Scenario percentile Change in max temp (oC) 4 2020s 3.5 2050s 2090s 3 2.5 2 1.5 1 0.5 0 A25 A50 A75 B25 B50 B75 Scenario percentile Figure 23. Plots of changes in rainfall, minimum temperature and maximum temperature under each scenario and risk level 75 Simulated rainfall and temperature results compare well with those obtained in the previous studies within the same area. Saka et al. (2003) used four GCMs (CSIRO, ECHAM, CGCMI and HadCM2) to estimate future changes in weather data for years 2020, 2075 and 2100 using baseline weather data of 1961-1990. They estimated that mean temperature will increarse by 1oC by the year 2020, 2oC by the year 2075 and 4oC by the year 2100. These estimates are consistent with results obtained in this study which are summarised in Table 13. In terms of rainfall, they also discovered that its pattern was not so certain with increase and decrease in different months in a year. However, they predicted a decrease in annual rainfall amounts ranging from 2% to 8% relative to the baseline period which is less than the values obtained in this study. Simulated weather data for Malawi reflects what scholars have defined as an ‘Aridification Scenario’, or the ‘dry’ and ‘core’ scenarios, meaning a reduction in rainfall amount and an increase in temperatures (Hulme, 2000; Saka et al., 2003). 5.3 Corn Yield Estimation with Aquacrop 5.3.1 Model Calibration Calibration of AquaCrop was carried out in order to set a benchmark for future simulation of corn yields at various probability percentiles and different time periods. By inputting the 2005 weather data into the model, the model grossly overestimated corn yields when compared with recorded data. It was therefore essential that adjustment of important and sensitive parameters in the model be carried out. By adjusting crop, management and soil properties in the main menu of AquaCrop, we were able to derive an output yield close in value to the recorded yield in 2005. Table 9 in 4.5.4 shows the final calibrated values of crop parameters which were vital in determining the final Biomass, Harvest Index and hence Yield of 76 corn using AquaCrop. Before calibrating the model, first we estimated the potential evapotranspiration (ETo) using the FAO ETo Calculator which uses the Penman-Monteith equation. Recorded data for corn production in year 2005 for Lilongwe District was .92 tons per hectare. After varying the model parameters, the closest simulated production was 0.926 tons per hectare. Once model calibration was complete, the model was then validated to determine its potential to simulate corn yields. 5.3.2 Model Validation The FAO AquaCrop model was validated in order to determine its potential to simulate future corn yields at various probability percentiles and time periods. Historical recorded corn yields for the District from 2000-2004 were used to compare with those simulated by AquaCrop. Figure 24 shows a plot of the results obtained from the model validation versus the recorded yields for the case study. Simulated Yield (ton/ha) 2 1.8 Yield = 0.9012x + 0.1505 R² = 0.911 1.6 1.4 1.2 Corn Yield 1 Linear (Corn Yield) 0.8 0.8 1 1.2 1.4 1.6 1.8 2 Observed Yield (ton/ha) Figure 24. Results of AquaCrop validation using the 2000-2004 simulated and recorded corn yields From Figure 24, a coefficient of determination (R2) of 0.911 was obtained. Other statistical parameters like prediction error (Pe), model efficiency (E), root mean square error 77 (RMSE) and mean absolute error (MAE) were estimated using Equations 33-36. Table 14 presents a summary of the results of the statistical analysis used in validation of AquaCrop. Table 14. AquaCrop validation results including prediction error statistics for corn yields Year Yields (ton/ha) Pe E Observed Simulated (±%) 2000 1.87 1.92 2.67 2001 1.62 1.46 9.88 2002 1.83 1.81 1.09 2003 1.04 1.12 7.69 2004 1.38 1.42 2.75 R2 RMSE MAE (ton/ha) 0.92 0.91 0.09 0.00 Note that the minimum and maximum prediction error (Pe) for corn yields for five years growing seasons (2000-2004) ranges from 1.09 to 9.88%, and RMSE and MAE values were 0.09 and 0.00 respectively. The three statistical parameters are either equal or close to zero, an indication of the suitability of the model to simulate future corn yields. Estimated values of model efficiency (E) and coefficient of determination (R2) of 0.92 and 0.91 respectively are both close to 1, which further increases confidence in the model’s potential to simulate future corn yields. The validation results indicate that the model can be used to simulate future corn yields within acceptable deviations from the true values. The values of simulated corn yields exhibited very little departure from the historical harvested yields and this was substantiated with the statistical analysis that was carried out in the process of model validation. 5.3.3 Simulating Current and Future Corn Yields with Climate Change Simulating future potential rainfed corn yields with climate change is the main focus of this study. To achieve this objective, we used the calibrated and validated crop model (AquaCrop) to estimate both average historical corn yields for the baseline period and future 78 potential corn production at various probability percentiles. Climate data (rainfall, minimum and maximum temperature, sunshine hours, relative humidity) for the 30 years baseline period (19712000) were broken down into monthly averages in each of the 12 months. The results from the model run indicated an average baseline period corn production of 3.41 ton/ha. This value is considered as corn production without climate change. Average values of future potential corn yields at various probability percentiles were calculated in a similar manner. Table 15 summarizes the simulated corn yields as a percentage, relative to the baseline period. Table 15. Corn production (tons/hectare) at different probability percentiles Period Percentile 2015-2035 (2020s) 2046-2065 (2050s) 2080-2099 (2090s) 25 50 75 25 50 75 25 50 75 Production change (%) Scenario A Scenario B -6.58 -4.70 -5.14 -8.19 -9.84 -14.33 -13.19 -14.86 -16.24 0.53 -8.11 -4.85 -7.25 -11.72 -13.54 -14.01 -15.39 -31.86 Range (%) 0.53 -8.11 -7.25 -14.33 -13.19 -31.86 Figures 25-27 show the changes in corn production relative to the baseline period and are generated from the results shown in Table 15. In general, the results suggest a decrease in corn yields for all the three time steps, scenarios and percentiles excepting for 2015-2035 period where a positive increase in corn production of 0.53% is expected under scenario B at 25% probability percentile. Note that SRA25 means change in corn production for scenario A at 25% probability percentile. 79 Change in Production (%) 2.00 0.00 -2.00 -4.00 -6.00 -8.00 -10.00 % Change SRA25 -6.58 SRB25 0.53 SNRA50 -4.70 SNRB50 -8.11 SNRA75 -5.14 SNRB75 -4.85 Change in Production (%) Figure 25. Change in corn production for 2015-2035 relative to baseline period of 1971-2000. 0.00 -2.00 -4.00 -6.00 -8.00 -10.00 -12.00 -14.00 -16.00 % change SRA25 -8.19 SRB25 -7.25 SNRA50 -9.84 SNRB50 -11.72 SNRA75 -14.33 SNRB75 -13.54 Change in Production (%) Figure 26. Change in corn production for 2046-2065 relative to baseline period of 1971-2000. 0.00 -5.00 -10.00 -15.00 -20.00 -25.00 -30.00 -35.00 % change SRA25 -13.19 SRB25 -14.01 SNRA50 -14.86 SNRB50 -15.39 SNRA75 -16.24 SNRB75 -32.67 Figure 27. Change in corn production for 2080-2099 relative to baseline period of 1971-2000. 80 The expected percentage changes in corn yields for 2015-2035, 2046-2065, and 20802099 are 0.53% to -8.11%, -0.73% to -14.33%, and -13.19% to -31.86% respectively. Saka et al. (2003) used CERES Maize to simulate corn yield for a 1961-1990 baseline period within the same area and discovered that there was a strong correlation between observed and simulated corn yields for the same period (with simulated yields being slightly higher than observed production). They then concluded that for future periods, corn yields would also be higher relative to the baseline period. However, this assumption appears to be ambiguous because first, in their study, they directly used outputs of GCMs into the simulation model without downscaling them. We discussed earlier on that due to low resolution of GCMs, their outputs need to be downscaled before carrying out further climate change assessment studies to avoid a mismatch between global and local climate data. Second, we also discussed that owing to high uncertainty levels associated with GCMs, it is necessary that an analysis to account for the uncertainties of the models is carried out. This procedure was also not done in their study. Third, the fact that historical recorded corn yields matches with historical simulated yields does not guarantee future recurrence of similar phenomenon. The 1961-1990 baseline period used in their study may be considered as pre-change period of climate change, therefore, production in the post-change period (2020, 2075, 2210), where changes in atmospheric greenhouse gasses is considered, must be estimated before major conclusions are drawn. 5.3.4 Correlation Analysis between Corn Yields and Rainfall and Temperature Changes A correlation analysis was carried out to examine the existence of a linear association between changes in predicted corn yields and changes in rainfall and temperature, as shown in figure 28 (a and b) below. A correlation coefficient is determined by taking the square root of the coefficient of determination (R2), with the sign of the correlation coefficient being the same as 81 the sign of the coefficient of x in the equation. Values of correlation coefficient range between -1 and +1, with each of the extremes indicating that the two variables are perfectly related in a negative and positive linear sense, respectively. A correlation coefficient of 0 indicates that there is no linear relationship between the two variables. 5 y = 0.0079x - 4.7687 R² = 0.0007 0 ∆ Corn Yields (%) -5 y = 0.2678x - 7.1527 R² = 0.8886 -10 -15 2020s -20 2050s y = 0.4077x - 10.593 R² = 0.4046 -25 -30 2090s a -35 -40 -30 -20 -10 ∆P (%) 0 10 20 5 ∆ Corn Yields (%) 2020s y = -5.6145x - 0.7362 R² = 0.1006 0 -5 2050s 2090s y = -7.4392x + 2.2061 R² = 0.7781 -10 -15 -20 y = -2.7287x - 10.279 R² = 0.0577 -25 -30 b -35 0 0.5 1 1.5 2 2.5 o ∆Tmax ( C) 3 3.5 4 - Figure 28. Correlation analysis between change in corn yields and change in precipitation and maximum temperature respectively 82 Figure 28a shows a positive linear relationship between changes in corn yields and rainfall changes for 2050s and 2090s, and no linear relationship in 2020s. The figure shows that a decrease in precipitation results in decrease in corn yields. Negative values of correlation coefficient in figure 28b indicate that an increase in temperature results in decrease in corn yields. Correlation coefficients results in each time period are summarized in Table 16. In general, excepting for 2020s, absolute values of correlation coefficients associated with changes in corn yields and rainfall changes are higher than those for changes in corn yields and maximum temperature changes. This signifies that future corn yields could be more sensitive to rainfall changes compared with changes in temperature. Table 16. Summary of values of correlation coefficient for rainfall and maximum temperature Correlation coefficients Period Rainfall Tmax 2020s 0.03 (-) 0.32 2050s 0.94 (-) 0.88 2090s 0.64 (-) 0.24 83 CHAPTER SIX: CONCLUSIONS AND RECOMMENDATIONS 6.1 Study Results The main objective of this study was to evaluate climate change impacts on rainfed corn production in Malawi. 15 GCMs were used to estimate future climate variables in three different time periods (2020s, 2050s and 2090s) using two scenarios (SRA and SRB). Historical weather data (1971-2000) for Chitedze Meteorological Station in Lilongwe District was utilized for future climate projections under climate change. Lilongwe District was selected as a representative site of the study. Owing to low resolution of GCMs that result in a mismatch between simulated global and local data, it was significant to downscale the outputs of GCMs (rainfall and temperature data). Downscaling of GCMs simulated weather data was achieved using LARS-WG model, a stochastic weather generator that has been extensively used in climate change studies. The model was calibrated and validated in order to determine its potential to estimate significant and reliable future climate variables. High p-values of chi-square and t tests signified high similarity and reliability between observed and simulated climate data. GCMs are also associated with high uncertainty levels which negatively impact on the credibility of the output data. This study employed the use of probability analysis in which a bounded range with known probability distribution was used to account for the uncertainties of fifteen (15) GCMs. The procedure involved (i) weighting of GCMs, (ii) generating monthly probability distribution functions (PDFs) and (iii) constructing monthly cumulative probability functions (CDFs) from which weather data at three risk levels (25%, 50% and 75%) in three time periods and two scenarios, were extracted and used as input data for the crop model. 84 AquaCrop Version 4, a new model developed by FAO that simulates the yield response to water deficit conditions, was used to assess potential corn production under rainfed farming with and without climate change. The model was first calibrated and validated using historical data to determine its reliability to simulate corn yields and statistical indicators like R2, E, RMSE, MAE and Pe were used to assess the suitability of the model to simulate potential corn yields. AquaCrop does not include an algorithm for calculating ETo, therefore, the FAO ETo Calculator was used to estimate evapotranspiration for potential crop production in AquaCrop. Results indicate mean annual rainfall changes of 11 to -17%, 0 to -26% and -3 to -29% in 2020s, 2050s and 2090s respectively. However, mean monthly rainfall patterns depicted both increase and decrease behavior in different months and time period. An overall increase in mean monthly temperature in all three time periods was observed. Average minimum temperature changes for 2020s, 2050s and 2090s were 0.59oC to 0.94oC, 1.31oC to 2.15oC and 1.84oC to 3.46oC respectively. Similarly, changes in maximum temperature for the same time steps were as follows; 0.52oC to 0.94oC, 1.26oC to 2.20oC and 1.78oC to 3.58oC respectively. Simulated temperature results compare well with the 2007 report of the Intergovernmental Panel on Climate Change (IPCC) which indicates that in sub-Saharan Africa, temperatures will rise by over 3 °C in 21st century. Results of previous study by Saka et al. (2003) in the same region also concurred with the findings of this study, particularly those of temperature changes. The results from the crop model show an overall decrease in corn yields in all three time periods. The expected range of changes in corn production for 2020s, 2050s, and 2090s are as follows; 0.53% to -0.81%, -0.73% to -14.33%, and -13.19% to -31.86% respectively. 85 6.2 Suggestions to Climate Change Adaptation Strategies We discussed earlier on that in the past, Malawi has experienced some extreme and devastating weather events like droughts and floods that caused severe food shortage in the country. The results of this study suggest that if no adaptation measures are implemented, climate change will exacerbate food shortage in Malawi. Recent news from Malawi’s Ministry of Agriculture indicates that in the 2013/2014 financial year, over 2 million people in Malawi are reported to be food insecure due to poor harvest in the just ended growing season (GoM, 2013). In view of the aforementioned, this paper suggests some adaptation strategies that the Government of Malawi could consider to ensure food security under climate change. Rockstrom (2003) argues that water harvesting could be an answer to future droughts and may supplement the existing rainfed farming. Water harvesting may be a viable solution for Malawi. Rainwater can be harvested and stored during the rainy season and could be used to irrigate crops during dry season. This may allow for multiple harvests within a year thereby increasing the likelihood of food security for the nation. In addition, Malawi may need to consider introducing or intensifying irrigation farming by introducing user friendly irrigation technologies which could be easily operated by subsistence farmers who are generally illiterate and have little technical know-how. This will not only improve corn production, but also increase the labor force, subsequently boosting economic status of farmers and of the nation as a whole In his article titled “Malawians rethink maize as climate dries”, Sanje (2013) conducted a survey which revealed that due to climate change, some parts of Malawi barely receive rainfall for a period of 120 days, the average time to maturity for local varieties of corn. This leads to accelerated crop growth and hence negatively impacts production. As an adaptation strategy, the 86 study recommends the combined farming of local and hybrid varieties of corn which could increase corn yields. Hybrid varieties mature between 60 to 80 days after planting, on the other hand, local varieties take more time to mature, but have developed resilience to climate changes including dry spells because they have been grown in this area since time immemorial. Therefore by planting different varieties of corn, there is a high probability of realizing a good harvest. Tchale (2009) discovered that on average, the soils in Malawi lose 40, 6.6 and 32.2 kg per hectare per year of nitrogen (N), phosphorus (P) and potassium (K) respectively, mainly due to runoff and soil erosion. Natural soil conservation methods like mulching and zero-tilling are some of the practices that could conserve nutrients like carbon into the soil thereby maintaining its fertility. Natural methods like mulches are said to retain soil moisture and moderate soil temperatures, consequently subduing the spread of diseases and pests (Dea and Scoones, 2003). AGRHYMET (2004) outlines a list of indigenous adaptation strategies that have been adopted in some parts of Africa for preventing nutrient loss, namely: “controlled bush clearing; using tall grasses such as Andropogon gayanus for fixing soil surface nutrients washed away by runoff; erosion-control bunding to reduce significantly the effects of runoff; restoring lands by using green manure; constructing stone dykes; managing low-lying lands and protecting river banks.” These indigenous adaptation strategies may also be considered in Malawi. Crop diversification may also help to deal with food insecurity under climate change. We also discussed earlier on that corn farming is the primary focus food crop in Malawi. However, with climate change, it may be important to consider other food crops like potatoes, rice, and cassava. Food crops like cassava do not require as much water as corn, have better resistance to droughts, and are less fertilizer dependent. Crop diversification may not only ensure food security but also reduce fertilizer costs thereby uplifting the economic welfare of the farmers. 87 Another adaptation strategy under consideration is biotechnology, defined by the online free dictionary as a “technique in which living organisms or parts of organisms are altered to make or modify agricultural products, to improve crops, or develop microbes for specific uses in agricultural processes”. ECA (2002) believes that biotechnology research could also help African nations particularly those in Sub-Saharan Africa to effectively deal with climate change effects on agriculture like pests, insects and droughts. By manipulating the genetic composition of seed crops, new varieties of crops are likely to evolve that will be pest, drought and insect resistant. 6.3 Study Limitations Like any other work, this study also has some limitations which may, to a certain extent, affect the overall results. Human error is one of the crucial areas which may affect the results. Climate change studies involve dealing with a huge amount of data and human error may occur at two stages: data processing and in-situ recordings. However, human error is minimized during observations by taking multiple readings and recording an average value of the climate variables. Second, the study estimates corn production based on monthly average values of a long period of time (20 year in each time period). Long term averages do not account for intra-annual weather variability which may be critical in assessing climate change impacts on crop production. Third, AquaCrop, a crop simulation model used in this study, does not consider the effect of pests and weeds on corn yields. Pests and weeds play a significant role in total biomass production and hence yield. However, the assumption is that proper field management practices are employed which retards the growth of weeds and outbreak of pests. Fourth, average values of some soil properties like hydraulic conductivity were used in the model; in practice, these values may vary immensely from one point to another. Model efficiency assumptions associated with emission scenarios are other important areas that may limit study results. 88 6.4 Future Work 6.4.1 Introduce Supplemental Irrigation Corn Farming The study results have estimated a decrease in rainfed corn production in each of the three time periods. By supplementing rainfed farming with irrigation, the crop model will run on combined rainfed and irrigation scenario and estimated yields compared with the results obtained in this study. 6.4.2 Improving the Water Retaining Capabilities of the Soils We discussed in 6.2, that the soils in Malawi lose substantial amounts of nitrogen (N), phosphorus (P) and potassium (K), mainly due to runoff and soil erosion. Excessive loss of nutrients and water may limit crop growth and hence affect crop yield. Management practices like mulching and erecting soil bunds may to be considered while running the model. The effect of improving the water retaining capabilities through the above mentioned management practices may be quantified by the crop model by comparing the yields of a well-managed field with those of normal managed soils (business as usual). 6.4.3 Intra Annual and Inter Seasonal Variations Future work shall also consider estimating corn yields taking into consideration both intra annual and seasonal variations of climate change. 6.4.4 Using the Up to Date Emission Scenarios (AR5) Estimations of future weather data in this report were based on the assumptions of emission scenarios as outlined in the Fourth IPCC Report (AR4). However, IPCC has just released the fifth assessment report (AR5) which is considered to provide a comprehensive assessment of the physical science basis of climate change. Future work may therefore consider incorporating these up to date assumptions in climate change studies. 89 REFERENCES Abedinpoura M., Sarangib A., Rajputb T.B.S., Singhb M., Pathakc H., and Ahmad T. 2012. Performance evaluation of AquaCrop model for maize crop in a semi-arid environment. Agricultural Water Management 110 (2012) 55– 66 Adams J.E., Arkin G.F., and Ritchie J.T. 1976. Influence of row spacing and straw mulch on first stage drying. Soil Sci. Soc. Am.J. 40:436-442. Adams R.M. 1989. Global climate change and agriculture: An economic perspective. American Journal of Agricultural Economics 71(5):1272-1279. Aerts, J. Droogers, P. (Eds.) 2004. Climate Change in Contrasting River Basins: Adaptation Strategies for Water, Food, and Environment. CABI Books, Londen, p. 288. Aggarwal PK, Kalra N, Chander S. 2006. InfoCrop: A dynamic simulation model for the assessment of crop yields, losses due to pests, and environmental impact of agroecosystems in tropical environments. I. Model description. Agric Syst; 89:1–25 AGRHYMET. 2004. Rapport synthèse de l’enquête générale sur les itinéraires d’adaptation des populations locales à la variabilité et aux changements climatiques conduite sur les projets pilotes/ par AGRHYMET et l’UQAM, par Hubert N’Djafa Ouaga, 13 pp Allen R., Pereira L.S., Raes D., and Smith M. 1998. Crop evapotranspiraron-Guidelines for computing crop water requirements. Irrigation and Drainage Paper No. 56. FAO, Rome. Anderson E.R., Cherrington., E.A., Flores A.I., Perez J.B., Carrillo R., and Sempris E. 2008. Potential Impacts of Climate Change on Biodiversity in Central America, Mexico, and the Dominican Republic, CATHALAC / USAID. Panama City, Panama. 105 pp Assessment: MAGICC and SCENGEN Version Barrow,E.M., 2.4 Workbook, Climate Research Unit, Norwich, U.K. 90 Barron J. 2004. Dry spell mitigation to upgrade semi-arid rainfed agriculture: water harvesting and soil nutrient management for smallholder maize cultivation in Machakos, Kenya. Ph.D. Thesis. Stockholm University, Sweden. Barron J., Rockctröm, J., Gichuki, F., Hatibu, N. 2003. Dry spell analysis and maize yields for two semi-arid locations in East Africa. Agric. For. Meteorol. 117, 23–37. Buishand T.A. 1978. Some remarks on the use of daily rainfall models. Journal of Hydrology 36, 295–308. Cai X, Wang D., Lauren R. 2009. Impact of Climate Change onCrop Yield: A Case Study of Rainfed Corn in Central Illinois, Journal of Applied Meteorology and Climatology. Vol. 48:1868-1881. Carlson R.E., Todey D.P., Taylor S.E. 1996. Midwestern corn yields and weather in relation to extremes of the southern oscillation. J Prod Agric 9:347–352 Carter T. R., Parry M. L., Harasawa., H., and Nishioka S. 1994. IPCC technical guidelines for assessing climate change impacts and adaptations. London,United Kingdom/Tsukuba, Japan : University College/Centre for Global Environmental Research. Challinor A.J., Wheeler T.R. 2008. Crop yield reduction in the tropics under climate change: processes and uncertainties. Agric Forest Meteorol; 148:343–56 Chen J., Zhang X.C., Liu W.Z., and Li Z., 2009. Evaluating and extending CLIGEN precipitation generation for the Loess Plateau of China. Journal of the American Water Resources Association 45 (2), 378–396 Chirwa W.E., Kydd J., and Dorward A. 2006. Future Scenarios for Agriculture in Malawi: Challenges and Dilemmas, Future Agricultures 91 Cline W. 2007. Global Warming and Agriculture (Washington, DC: Peterson Institute for International Economics. Cuculeanu V, Tuinea P, and Balteanu D. 2002. Climate change impacts in Romania: vulnerability and adaptation options. Geol J; 57:203–9. Dea D. and I. Scoones, 2003. Networks of knowledge: how farmers and scientists understand soils and their fertility: a case study from Ethiopia. Oxford Development Studies, 31, 461478. Department of the Environment (DOE). 1996. Review of the potential effects of climate change in the United Kingdom. London: HMSO. Descheemaeker, K. 2006. Pedological and hydrological effects of vegetation restoration in enclosures established on degraded hill slopes in the highlands of Northern Ethiopia. Ph.D. diss., Dissertationes de Agricultura no. 725. K.U. Leuven Univ., Leuven, Belgium. Downscaling Climate Data, http://www.climate-decisions.org Droogers P., van Dam J., and Hoogeveen J. 2004. Adaptation strategies to climate changes to sustain food security. In: Aerts J, Droogers P, editors. Climate change in contrasting river basins: adaptation strategies for water, food and environment. The Netherlands: CABI Publishing; p. 49–73. ECA (Economic Commission for Africa). 2002. Harnessing technologies for sustainable development. Economic Commission for Africa Policy Research Report, Addis Ababa, 178 pp. Eitzinger J., Stastna M., and Zalud Z. 2003. A simulation study of the effect of soil water balance and water stress on winter wheat production under different climate change scenarios. Agric Water Manage; 61:195–217. 92 FAO. 2007b. Adaptation to Climate Change in Agriculture, Forestry, and Fisheries: perspective, framework and priorities.FAO, Rome. FAO. 2008a. The state of world fisheries and aquaculture. FAO, Rome FAO. 2009. ETo Calculator: Land and Water Digital Media Series No. 36.FAO, Rome. Farahani, H. J., Izzi, G. and Oweis, T.y. 2009: Parameterization and Evaluation of the AquaCrop Model for Full and Deficit Irrigated Cotton. Agronomy Journal 101: 469-476. Fereres, E. 2010. Crop Yield Response to water Firmino G. Mucavele. 2009. True Contribution of Agriculture to Economic Growth and Poverty Reduction: Malawi, Mozambique and Zambia Synthesis Report; Food, Agriculture and Natural Resources Policy Network (FANRPAN) Fowler H. J., Blenkinsop S., and Tebaldi C. 2007. Linking climate change modelling to impacts studies: Recent advances in downscaling techniques for hydrological modelling. Int J Climatol, 27, 1547–1578. Geerts S. and Lunacre E. 1998. What are General circulation models? (http://wwwdas.uwyo.edu/~geerts/cwx/notes/chap12/nwp_gcm.html) Geerts S., D. Racs M. Garcia Cardenas O. Condori J. Mamani R. Miranda J. Cusicanqui C. Taboada E., Yucra and J. Vacher. 2008. Could deficit irrigation be a sustainable practice for quinoa (Chenopodium quinoa Willd.) in the Southern Bolivian Altiplano? Agric. Water Manage. 95:909-917. Godwin D.C., Ritchie J.T., Singh U., Hunt, L. 1989. A User’s Guide to CERES WheatV2.10. International Fertilizer Development Center, Muscle Shoals, AL. 93 Gohari A., Eslamian S., Abedi-Koupaei J., Bavani A.M., Wang D. and Madani K. 2013. Climate change impacts on crop production in Iran's Zayandeh-Rud River Basin. Science of the Total Environment, 442: 405–419. GoM (Government of Malawi). 2001. State of Environment Report for Malawi 2001, Lilongwe: Ministry of Natural Resources and Environmental Affairs. GoM (Government of Malawi). 2007b. Economic Report, Ministry of Economic Planning and Development, Lilongwe, Malawi. GoM (Government of Malawi). 2013. Department of Climate Change and Meteorological Services, Ministry of Environment and Climate Management. Accessible at http://www.metmalawi.com/. GoM (Government of Malawi). 2013. Ministry of Agriculture and Food Security, Lilongwe, Malawi GoM (Government of the Republic of Malawi). 2005/2006. Ministry of Agriculture and Food Security, Impact of Green Maize Consumption on Potential Production (Lilongwe and Dowa Case Study) Gommes R.A. 1983. Pocket computers in agrometeorology, Plant production and Protection Paper n. 45. FAO, Rome, Italy, 140 p Harun S. 1999. Forecasting and Simulation of Net Inflows for Reservoir Operation and Management. Universiti Teknologi Malaysia: Ph.D Thesis. Hassan Z.B. 2012. Climate Change Impact on Precipitation and Streamflow in a Humid Tropical Watershed Heng L.k., Hsiao,T.C., Evett, S., Howell, T. and Steduto P. 2009. Validating the FAO AquaCrop Model for Irrigated and Water Deficient Field Maize. Agronomy Journal 101: 488-498. 94 Holton J. R. 1992. An introduction to dynamic meteorology, 3rd Edition. SanDiego, CA: Academic. Hoogenboom G., Jones, J.W., Wilkens, P.W., Batchelor, W.D.,Bowen, W.T., Hunt, L.A., Pickering, N.B., Singh, U., Godwin,D.C., Bear, B., Boote, K.J., Ritchie, J.T., White, J.W., 1994. Crop Models, DSSAT Version 3.0. International Benchmark Sites Network for Agrotechnology Transfer, University ofHawaii, Honolulu, 692 pp Houghton J. T., Ding. 2000. Climate change 2001: the scientific basis:contribution of working group I to the third assessment report of the intergovernmental panel on climate change. Cambridge: Cambridge University Press. Hulme. 2000. Using a Climate Scenario Generator for Vulnerability and Adaptation Hunink J.E., and Droogers P. 2010. Climate Change Impact Assessment on Crop Production in Albania, World Bank Study on Reducing Vulnerability to Climate Change in Europe and Central Asia (ECA) Agricultural Systems Huntingford C., C. H. Lambert, J. H. C. Gash, C. M. Taylor, and Challinor A. J. 2005. Aspects of climate change prediction relevant to crop productivity. Philos. Trans. Roy. Soc. London,360B, 1999–2009, doi:10.1098/rstb.2005.1748. in central-eastern Argentina and El Niño-Southern Oscillation. J Appl Meterol 38:1488– 1498 IPCC (Intergovernmental Panel on Climate Change). 2007. Climate Change: Impacts, Adaptation and Vulnerability, Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (Cambridge Univ Press, Cambridge, UK), in press. 95 IPCC SRES. 2000. Nebojsa Nakicenovic and Rob Swart (Eds.), Cambridge University Press, UK. Pp 570. IPCC. 2001. Impacts, Adaptation and Vulnerability, Contribution of Working Group II to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, UK: Cambridge University Press. Jaehyuk Lee J., Nadolnyak D., and Hartarska V. 2012. Impact of Climate Change on Agricultural Production in Asian Countries: Evidence from Panel Study. Department of Agricultural Economics and Rural Sociology, 202 Comer Hall, Auburn University, Auburn, AL 36849-5406 Japan Association for International Collaboration of Agriculture and Forestry (JAICAF). 2008. The Maize in Zambia and Malawi. Johnson G.L., Hanson C.L., Hardegree S.P., Ballard E.B. 1996. Stochastic weather simulation: overview and analysis of two commonly used models. Journal of Applied Meteorology 35 (10), 1878–1896. Jones C.A. and Kiniry J.R. 1986. CERES-Maize model: A simulation model of maize growth and development. Texas A & M University Press, 194 pp. Jones J.W., G. Hoogenboom C.H. Porter, K.J. Boote, W.D. Batchelor, L.A.Hunt, P.W. Wilkens, U. Singh, A.J. Gijsman, and J.T. Ritchie. 2003. The DSSAT cropping system model. Eur. J. Agron. 18:235–265. Kachule R.N. 1994. Economic analysis of beans production on northern Malawi: Implications for agricultural policy. Ms Theisis. Bunda College, Lilongwe Malawi 96 Kaiser H.M, Riha S.J, Wilks DS, Rossiter D.G, Sampath R. 1993. A farm-level analysis of economic and agronomic impacts of gradual climate warming. AmJ Agric Econ 75(2):387– 398 Kang., Khan S., Ma X. 2009. Climate change impacts on crop yield, crop water productivity and food security-A review. Progress in Natural Science, Pp 1665–1674. Kanyama-Phiri, GY., Kamangira J.B., and Wellard K. 1994. The farmer Participatory Agroforestry Research project. Summary of preliminary findings and proposal for the next phase of research. Lilongwe, Malawi. Kevin M., Ramesh R., John O., Imran A., Bahram and G. 2005. Evaluation of weathergenerator ClimGen for southern Ontario. Canadian Water Resources Journal 30(4), 315–330. Kimball B.A. 1983. Carbon Dioxide and Agricultural Yield: An Assemblage and Analysis of 430 Prior Observations. Agron. J. 75:779-788 Krishnan P., Swain D.K, and Bhaskar B.C. 2007. Impact of elevated CO2 and temperature on rice yield and methods of adaptation as evaluated by crop simulation studies. Agric Ecosyst Environ;122:233–42. Luo Q, Williams M, and Bellotti W, 2003. Quantitative and visual assessments of climate change impacts on South Australian wheat production. Agric Syst; 77:173–86. Mainuddin M., Hoanh C.T., Jirayoot K., Halls A.S., Kirby M., Lacombe G. and Srinetr V. 2010. Adaptation options to reduce the vulnerability of mekong water resources, food security and the environment to impacts of development and climate change. CSIRO: Water for a Healthy Country National Research Flagship. 152 pp. 97 Malawi Government (GoM). 1995. Economic Report, 1995, Ministry of Economic Planning and Development. Zomba Malawi. McMichael A, Woodruff R, Whetton P, Hennessy K, Nicholls N, Hales S, Woodward A, Kjellstrom T. 2003. Human Health and Climate Change in Oceania: Risk Assessment, 2002 (Department of Health and Aging, Canberra, Australia). McMichael A.J., R.E. Woodruff and Hales S. 2006. Climate change and human health: present and future risks. Lancet, 367, 859-869. Mendelsohn R. 2008. The Impact of Climate Change on Agriculture in Developing Countries, Journal of Natural Resources Policy Research, 1:1, 5-19 Mendelson R., Dinar A., and Dalfelt A. 2000. Climate change impacts on African agriculture. http://www.worldbank.org/wbi/sdclimate/pdf. Mohammed Y. 2009. Climate change impact assessment on soil water availability and crop yield in Anjeni Watershed Blue Nile Basin. Arba Minch University: Master Thesis Moriasi D. N., Arnold, J. G., Liew M. W. V., Bingner R. L., Harmel R. D., and Veith T. L. 2007. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Transactions of The ASABE 50, 885–900. Nash J.E., Sutcliffe J.V. 1970. River flow forecasting through conceptual models. I. A discussion of principles. Journal of Hydrology 10, 282–29 Ng'ong'ola D.N., Mtimuni, B.M., Kanyama-Phiri, G., Nothale, D.W. and Wiyo, K.A. 1992. Socio-economic forecasts for land use programmes in Malawi, in Kwapata, M. 1993, Ibid Nordhaus William D. 1991. To slow or not to slow: The economics of the greenhouse effect, The Economic Journal, 101, pp. 920–937. NSO (National Statistics Office). 2008. Population and Housing Census. Final Report, Zomba 98 OECD. 2003. Special Issue on Climate Change: Climate Change Policies: Recent development and long term issues. Paris: Organization for Economic Cooperation and Development (OECD) publications. Owens T., Hoddinot J., and Kinsey B. 2003. Ex-ante actions and ex-post public responses to drought shocks: evidence and simulations from Zimbabwe. World Dev., 31, 1239-1255 Ozdogan M. 2011. Modeling the impacts of climate change on wheat yields in northwestern Turkey. Agr Ecosyst Environ; 141:1-12 Parry M., Rosenzweig C., Iglesias A. 1999. Climate change and world food security: a new assessment. Global Environ Change: S51–67. Parry M.L., Rosenzweig C., Iglesias A., Livermore M., and Fischer G. 2004. Effects of climate change on global food production under SRES emissions and socio-economic scenarios. Global Environ. Change. 14, 53–67 Paul W.G. 2012. Biomass Supply from Corn Residues: Estimates and Critical Review of Procedures, Department of Economics, Iowa State University Phillips JG, Cane M.A, Rosenzweig C. 1998. ENSO seasonal rainfall patterns and simulated maize yield variability in Zimbabwe. Agric For Meteorol 90:39–50 Podestá G.P., Messina C.D.,Grondona M.O, Magrin G.O. 1999. Associations between grain crop yields Popova Z, Kercheva M. 2005. CERES model application for increasing preparedness to climate variability in agricultural planning-risk analyses. Phys Chem Earth; 30:117–24. Racksko P., Szeidl L., and Semenov M. 1991. A serial approach to local stochastic weather models. Ecological Modeling, 57, 27–41. 99 Raes D, Steduto P, Hsiao T. C., and Fereres E. 2012. FAO, Reference Manual for AquaCrop Version 4, Land and Water Division Rome, Italy Raes D. 1982. A summary simulation model of the water budget of a cropped soil: Ph.D. diss. Dissertationes de Agricultura no. 122. K. U. Leuven Univ., Leuven, Belgium. Raes D., Geerts S., Kipkorir E., Wellens J., and Sahli A. 2006. Simulation of yield decline as a result of water stress with a robust soil water balance model. Agric. Water Manage. 81:335-357. Raes, D., H. Lemmens P., Van Aelst M., Vanden Bulcke, and Smith M. 1988. IRSISIrrigationScheduling Information System. Volume 1. Software Reference Manual. K.U. Leuven, Dep. Land Management, Reference Manual 3. Leuven, Belgium. Rallison R.E. 1980. Origin and evolution of the SCS runoff equation. p.912924. In Proc. Symp. on Watershed Management. ASCE, New York. Reddy V.R, Pachepsky Y.A.2000. Predicting crop yields under climate change conditions from monthly GCM weather projections. Environ Modell Software;15:79–86. Riha S.J, Wilks D.S, Simoens P. 1996. Impact of temperature and precipitation variability on crop model predictions. Clim Change 32(3):293–311 Ringler C., Zhu, T., Cai X., Koo J., and Wang,D. 2010. Climate Change Impacts on Food Security in Sub -Saharan Africa. IPFPRI. Rockström J. 2000. Water resources management in smallholder farms in Easternand Southern Africa: an overview. Phys. Chem. Earth (B) 25 (3), 275–283. Rockström J. 2003. Resilience building and water demand management for drought mitigation. Phys. Chem. Earth, 28, 869-877 100 Rosenzweig C, Parry ML. 1994. Potential impact of climate change on world food supply. Nature 367:133–138 Rosenzweig C. 1989. Global climate change: Predictions and observations. American Journal of Agricultural Economics 71(5):1265-1271. Rowlinson, P. 2008. Adapting Livestock Production Systems to Climate Change – Temperate Zones. Livestock andGlobal Change conference proceeding. May 2008, Tunisia. Rubas D.J., Hill H.S.J., Mjelde J.W. 2006. Economics and climate applications: exploring the frontier. ClimRes 33:43–54 Saka A.R, Phiri I.M.G, Kamdonyo D.R, and Kumwenda J.D.T. 2003. The Potential Impact of Climate Change on the Agiculture Sector in Malawi. Ministry of Natural Resources and Environmental Affairs, Agriculture and Livelihood Sanje K. 2013. Malawians rethink maize as climate dries, Thomson Reuters Foundation. http://www.trust.org/item/?map=malawians-rethink-maize-planting-as-climate-dries/ Schmidhuber J. and Tubiello F.N. 2005. Global food security under climate change. vol. 104. pp; 19703–19708. Semenov M. A., and Barrow E. M. 2002. LARS-WG: a stochastic weathergenerator for use in climate impact studies (Version 3.0). User Manual. Semenov M. A., and Brooks R. J. 1999. Spatial interpolation of the LARS-WG stochastic weather generator in Great Britain. Climate Research, 11, 137-148. Seo S., and Mendelsohn R. 2007. An Analysis of Livestock Choice: adapting to climate change in Latin American farms. World Bank Policy Research Working Paper No. 4164. 101 SEPLD. 2006. Social Economic Profile for Lilongwe District (Including District Development Planning Framework Shaka, A. K. 2008. Assessment of climate change impacts on the hydrology of Gilgel Abbay Catchment in Lake Tana Basin, Ethiopia Enschede, Netherlands. The International Institute for Geo-information Science and Earth Observation: Master Thesis. Shimpei M. 1991. A guide to Ecological Agriculture in the tropics, Lessons from Nature Singh J.P., Govindakrishnan P.M., Lal SS and Aggarwal P.K. 2005a. Increasing the efficiency of Agronomy Singh R, Hales S, deWet N, Raj R, Hearnden M, Weinstein P. 2001. Environ Health Perspect 109:55–59. Steduto P.,T., C. Hsiao., D. Raes, and E. Fereres. 2009. AquaCrop-The FAO crop model to simulate yield response to water: I. Concepts and underlying principles. Agron. J. 101:426-437 (this issue). Steenhuis T.S., M. Winchell, J., Rossing J.A., Zollweg, and M. F. Walter. 1995. SCS Runoff equation revisited for variable-source runoff areas. J. Irrig. Drain. Eng. 121:234-238. Stöckle C.O., Donatelli, M., Nelson, R. 2003. CropSyst, a cropping systems simulation model. Eur. J. Agron. 18 (3/4), 289–307 (Second special issue “2nd International Symposium on Modeling Cropping Systems, Florence, Italy”). Stroosnijder L. 2007. Rainfall and land degradation. In: Sivakumar, M.V.K., Ndiang’ui,N. (Eds.), Climate and Land Degradation. Springer, pp. 167–195 102 Tao F, Zhang Z. 2010. Adaptation of maize production to climate change in North China Plain: quantify the relative contributions of adaptation options. Eur J Agron; 33:103– 16 Tao F, Zhang Z. 2010. Adaptation of maize production to climate change in North China Plain: quantify the relative contributions of adaptation options. Eur J Agron, 33:103– 16 Tchale H. 2009. The Efficiency of Smallholder Agriculture in Malawi; World Bank: Lilongwe, Malawi. Thornton P.K., Jones P.G., Owiyo T, Kruska R.L., Herrero M., Kristjanson P., Notenbaert A., Bekele N and Omolo A, with contributions from Orindi V, Otiende B, Ochieng A, Bhadwal S, Anantram K, Nair S, Kumar V and Kulkar U. 2006: Mapping Climate Vulnerability and Poverty in Africa. Report to the Department for International Development, ILRI, Nairobi. p.171. Todorovic M., Albrizio,R., Zivotic, L., Abi Saab, M.T., Stockle, C. and Steduto, P. 2009. Assessment of AquaCrop, CropSyst, and WOFOST Models in the Simulation of Sunflower Growth under Different Water Regimes. Agronomy Journal 101: 509521. Tojo Soler C.M, Sentelhas P.C, Hoogenboom G. 2007. Application of the CSM-CERESMaize model for planting date evaluation and yield forecasting for maize grown offseason in a subtropical environment. Eur J Agron; 27:165–77. Tubiello F.N, Rosenzweig C, Goldberg R. A., Jagtap S., Jones J. W. 2002. Effects of climate change on US crop production: simulation results using two different GCM scenarios. Part I: Wheat, potato, maize, and citrus 103 United States Environmental Protection Agency (EPA), http://www.epa.gov/climatechange/basics/ USDA. 1964. Estimation of direct runoff from storm rainfall. p. 1-24. In National Engineering Handbook. Soil Conservation Service, USDA, Washington, DC. USGCRP. 2009. Global Climate Change Impacts in the United States. Karl, T.R., J.M. Melillo, and T.C. Peterson (eds.). United States Global Change Research Program. Cambridge University Press, New York, NY, USA van Dam, J.C., Huygen, J., Wesseling, J.G., Feddes, R.A., Kabat,P., van Walsum, P.E.V., Groenendijk, P., van Diepen, C.A. 1997. Theory of SWAP Version 2.0, Report #71. DepartmentWater Resources, Wageningen Agricultural University, 167 pp Vasilev V. 2003. Epidemiol Infect 132:51–56 Villalobos F.J., and E. Fereres. 1990. Evaporation measurements beneath corn, cotton, and sunflower canopies. Agron. J. 82:1153-115 Walker N.J., Schulze R.E. 2006. An assessment of sustainable maize production under different management and climate scenarios for smallholder agro-ecosystems in KwaZulu-Natal, South Africa. Phys Chem Earth; 31:995–1002. Willmott, C. J. 1984. On the evaluation of model performance in physical geography. InSpatial Statistics and Models, Gaile GL, Willmott CJ (eds). D. Reidel: Boston. 443–460 World Fish Center. 2007. Fisheries and aquaculture can provide solutions to cope with climate change. Issues brief, 1701. Xie H, Eheart JW. 2004. Assessing vulnerability of water resources to climate change in midwest. Copyright ASCE; 1–10. 104 Xu C. Y. 1999. From GCMs to River Flow: A Review of Downscaling Methods and Hydrologic Modelling Approaches. Progress in Physical Geography, 23(2), 229– 249. Yao F, Xu Y and Lin El. 2007. Assessing the impacts of climate change on rice yields in the main rice areas of China. ClimChange; 80:395–409. You L, Rosegrant MW, Wood S, Sun D. 2009. Impact of growing season temperature on wheat productivity in China. Agric for Meteorol; 149:1009–14. 105