* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download How might Australian rainforest cloud interception
Climate change feedback wikipedia , lookup
Climate change adaptation wikipedia , lookup
Media coverage of global warming wikipedia , lookup
Scientific opinion on climate change wikipedia , lookup
Climate change and agriculture wikipedia , lookup
Public opinion on global warming wikipedia , lookup
Effects of global warming on human health wikipedia , lookup
Attribution of recent climate change wikipedia , lookup
Climate change in Tuvalu wikipedia , lookup
Surveys of scientists' views on climate change wikipedia , lookup
Solar radiation management wikipedia , lookup
IPCC Fourth Assessment Report wikipedia , lookup
Climate change in Australia wikipedia , lookup
Climate change and poverty wikipedia , lookup
Climate change, industry and society wikipedia , lookup
Journal of Hydrology 481 (2013) 85–95 Contents lists available at SciVerse ScienceDirect Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol How might Australian rainforest cloud interception respond to climate change? Jim Wallace a,b,⇑, Dave McJannet c a Centre for Tropical Water & Aquatic Ecosystem Research (TropWATER), James Cook University, Townsville, QLD 4811, Australia CSIRO Land and Water, Australian Tropical Science and Innovation Precinct, Townsville, QLD 4814, Australia c CSIRO Land and Water, EcoSciences Precinct, Dutton Park, QLD 4120, Australia b a r t i c l e i n f o Article history: Available online 5 January 2013 This manuscript was handled by K. Georgakakos, Editor-in-Chief, with the assistance of Matthew McCabe, Associate Editor. Keywords: Rainforest Climate change Cloud interception Canopy wetness s u m m a r y The lower and upper montane rainforests in northern Queensland receive significant amounts of cloud interception that affect both in situ canopy wetness and downstream runoff. Cloud interception contributes 5–30% of the annual water input to the canopy and this increases to 40–70% of the monthly water input during the dry season. This occult water is therefore an important input to the canopy, sustaining the epiphytes, mosses and other species that depend on wet canopy conditions. The potential effect of climate change on cloud interception was examined using the relationship between cloud interception and cloud frequency derived from measurements made at four different rainforest locations. Any given change in cloud frequency produces a greater change in cloud interception and this ‘amplification’ increases from 1.1 to 1.7 as cloud frequency increases from 5% to 70%. This means that any changes in cloud frequency will have the greatest relative effects at the higher altitude sites where cloud interception is greatest. As cloud frequency is also a major factor affecting canopy wetness, any given change in cloud frequency will therefore have a greater impact on canopy wetness at the higher altitude sites. These changes in wetness duration will augment those due to changes in rainfall and may have important implications for the fauna and flora that depend on wet canopy conditions. We also found that the Australian rainforests may be more efficient (by 50% on average) in intercepting cloud water than American coniferous forests, which may be due to differences in canopy structure and exposure at the different sites. Ó 2013 Published by Elsevier B.V. 1. Introduction Australia contains some unique tropical rainforests in the Wet Tropics region of northern Queensland. These forests vary in structure and composition as the soils and climate change with increasing distance from the coast (Tracey, 1982; McJannet et al., 2007d,b). There is also a marked effect of altitude as the terrain rises rapidly from the coastal plain (50–100 m ASL) to the ‘Tablelands’ plateau (1000 m ASL) and further to the mountains of the Great Dividing Range (up to 1600 m ASL). On the seaward side of these mountains, rainfall increases with altitude and this has a major effect on the forest water balance and the duration of wet conditions in the canopy (McJannet et al., 2007a,b; Wallace and McJannet, 2012). However, the rainforests above 1000 m also receive significant inputs of water as cloud interception (McJannet et al., 2007c), which constitutes between 4% and 29% of the annual water input. More importantly, the above studies have also shown that the contribution of cloud interception during dry season ⇑ Corresponding author at: Centre for Tropical Water & Aquatic Ecosystem Research (TropWATER), James Cook University, Townsville, QLD 4811, Australia. Tel.: +44 1635 770 634. E-mail address: [email protected] (J. Wallace). 0022-1694/$ - see front matter Ó 2013 Published by Elsevier B.V. http://dx.doi.org/10.1016/j.jhydrol.2012.12.028 months can be as high as 65%, when it becomes the major water input sustaining in situ canopy wetness and downstream runoff. If climate change alters the frequency and amount of cloud interception there could therefore be important hydrological and ecological consequences. The potential impacts of climate change on the water balance of these forests have been reported by Wallace and McJannet (2012). A range of climate simulations for north Queensland predict future (2050) rainfall and temperature changes of ±20% and 1–3 K respectively (see Section 2.5), and using these figures they found that the water balance was primarily affected by rainfall changes rather than temperature. Runoff downstream of the forests changed by a greater amount than rainfall, especially in the dry season, where there were important effects on the onset and duration of the period when there is no runoff. They also found potential in situ impacts of climate change that affect how long the rainforest canopy is wet, which may have important implications for the epiphytes, mosses and other species that depend on these wet canopy conditions (Hölscher et al., 2004; Köhler et al., 2006). However, the climate impacts on canopy wetness analysis presented by Wallace and McJannet (2012) only dealt with possible changes in rainfall and no account was taken of how climate change might influence cloud interception. As this is an important input to the higher alti- 86 J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 tude rainforests in Australia, this paper presents an analysis of how cloud interception might change under future climate scenarios and in turn how this might affect the seasonality of canopy wetness and downstream runoff. 2. Methods 2.1. Locations and weather The cloud interception and climate impact analyses were performed for four rainforest locations where significant cloud interception has been recorded by McJannet et al. (2007a,b), Fig. 1. Detailed descriptions of the forest sites and methods used to measure rainfall and cloud interception are given by McJannet et al. (2007a,b), so only brief site descriptions are given below. The first site, Upper Barron (UB), is a lower montane rainforest on the Atherton Tablelands (1050 m ASL). The two sites at Mt Lewis are located in the Mt Lewis Forest Reserve and are also classified as lower montane rainforest. The lower site (ML1; 1100 m ASL) is in a shallow valley while the second site (ML2; 1160 m ASL) is located on a more exposed hill slope. The forest at ML2 is much more stunted because of this exposure, with a canopy height of 12 m, compared to 32 m at ML1. The fourth site is located in upper montane cloud forest near the summit of Bellenden Ker (BK) (1560 m ASL), which is the second highest mountain in Queensland. The rainforests types vary from complex notophyll vine forest on the Tablelands to simple microphyll vine fern thicket at the top of the mountains (Tracey, 1982), Table 1. The rainforests generally form closed canopies and are up to 30 m tall. The upper montane cloud forest at the top of Bellenden Ker is much more stunted (8 m) and has the lowest leaf area index (3.3). Further details the forest types are given by McJannet et al. (2007b,c). The Australian Wet Tropics region has high annual rainfall, but with distinct wet and dry seasons. The dry season (six driest months) runs from June to November with an average of 20% of the annual rainfall, while the wet season (six wettest months) generally occurs from December to May with 80% of annual rainfall. The rainforest furthest from the coast (UB) has the highest radia- tion and warmest air and all sites have very high humidity (Table 1). The upper montane forest (BK) grows in a much cloudier and cooler climate and is exposed to the highest wind speeds. Long term weather records for the four sites were obtained from the Australian SILO meteorological database (Jeffrey et al., 2001). Locations within a few kilometres of the rainforest sites were chosen and daily rainfall, temperature, solar radiation and vapour pressure extracted for the period 1950–2008. For the periods where weather data (618–1113 days) were also measured at the rainforest sites by McJannet et al. (2007a) correlations were established between the actual site and the SILO weather data (see Wallace and McJannet, 2012). Site rainfalls, solar radiation and temperature were significantly different from those in SILO, so in subsequent calculations of canopy wetness duration and downstream runoff using SILO long term data we therefore applied the regressions given by Wallace and McJannet (2012) for rainfall, temperature, solar radiation and vapour pressure. 2.2. Rainforest water balance The water balance of a rainforest has been described by McJannet et al. (2007c) as: ðPg þ Pc Þ ¼ I þ Et þ Es þ R þ D þ dh; ð1Þ where (Pg + Pc) represents total precipitation input to the forest as rainfall, Pg and cloud interception, Pc. Water losses from the canopy are given by the sum of the canopy interception I, and transpiration, Et. Direct evaporation from the forest floor, Es, is usually small and R and D represent the processes of runoff (i.e. lateral flows whether overland or subsurface) and drainage, respectively. Over short time periods (days to months) changes in soil moisture storage (dh) can be a significant part of the water balance. The difference between the rainfall, Pg and the total evaporation (I + Et + Es) is referred to as ‘excess water’ by Wallace and McJannet (2012), and is given by (R + D + dh). When the change in dh is small compared to the other water balance components (e.g. high rainfall months and/or annual periods) the excess water gives a good measure of (R + D). Fig. 1. Location map showing Wet Tropics region and location of the four rainforest sites that have significant cloud interception (D). Altitudes of field sites are given in brackets. Reproduced from McJannet et al. (2007c). 87 J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 Table 1 Details of the location, forest type and structure and weather characteristics for the four rainforest sites that receive significant cloud interception. Further details of the size classes used to determine stem density and basal area are given by McJannet et al. (2007a). Lat. (S) Long. (E) Altitude (m) Forest type LAI (m2 m2) Basal area (m2 ha1) Stem density (ha1) Canopy height (m) Average precipitation (mm year1)a Average potential evaporation (mm year1) Average daily solar radiation (MJ m2) Average daily temperature (°C) Average daily relative humidity (%) Average daily wind speed (m s1) a Upper Barron Mount Lewis 1 Mount Lewis 2 Bellenden Ker 17°27.1 145°29.7 1050 Complex Notophyll Vine Forest 4.1 69 925 25 2087 1237 16°31.7 145°16.7 1100 Simple Notophyll Vine Forest 4.5 62 650 32 3471 982 16°31.8 145°16.3 1160 Simple Notophyll Vine Forest 4.1 75 2150 12 2673 N/A 17°16.0 145°51.0 1560 Simple Microphyll Vine Fern Thicket 3.3 74 2019 8 7386 1073 17.2 18.3 94 2.9 14.4 16.8 90 1.1 N/A 16.3 93 N/A 12.6 14.2 88 3.2 Corrected rainfall from McJannet et al. (2007b). 2.3. Cloud interception and frequency Estimates of cloud interception were made at all four field sites using a combination of cloud interception gauges and the canopy water balance during periods when clouds interception was detected (McJannet et al., 2007a). This method overcomes problems associated with the difference in the cloud droplet interception efficiency between the gauge and the forest canopy. Cloud frequency (the % of time when clouds are present within the rainforest canopy) can be derived from the times when the cloud interception gauges recorded events. However, as cloud could also be passing through the forest canopy before the first recorded event, in any gaps between events and after the final event, the true cloud frequency may be significantly longer than simply the sum of when gauged events occurred. Several estimates of cloud frequency were therefore made (see Section 3) using a combination of cloud interception and high relative humidity data, measured using a Vaisala capacitance sensor located in clearing near the forest (for ML1, ML2 and UB) and above the forest canopy at BK. 2.4. Canopy wetness duration Rainforest canopies can be wetted by rain and/or cloud interception which may or may not occur simultaneously. The total duration (% of time) of canopy wetness is therefore given by; W total ¼ fcnr þ fcr þ T nc r Wr; ð2Þ where the total cloud frequency duration (fc) is split into fractions that occurred when it was (fcr ) and was not (fcnr ) raining. Total monthly durations of canopy wetness due to cloud interception were estimated from mean monthly cloud frequency (as described above) and the split into fcr and fcnr was made using 30 min recordings of rainfall from each site. Monthly wetness durations due to rain in the absence of canopy cloud were derived from the product of the total wetness duration due to rainfall (W r ) and the fraction of time rain occurred in the absence of cloud interception (T nc r ). W r was calculated using a modification of Gash et al. (1995) model described by Wallace and McJannet (2012). Wetness durations during rainfall are given by the amount of precipitation (Pg) that fell divided by the mean rainfall rate (given by Wallace and McJannet (2008)). Additional wetness durations after rainfall were derived by dividing the canopy storage by the mean potential evaporation rate after rain, Ep, derived using 30 min weather data recorded at each site (see Wallace and McJannet, 2012). The mean monthly fraction of time rain occurred in the absence of cloud interception (T nc r ) was calculated using 30 min recordings of rainfall and cloud interception from each site. 2.5. Climate scenarios Wallace and McJannet (2012) have pointed out that general circulation model (GCM) simulations of climate change for the north coast of Queensland are highly uncertain (Cai et al., 2003; CSIRO, 2007; Li et al., 2009; Suppiah et al., 2010; QCCCE, 2010; McGregor and Nguyen, 2010) with different GCM’s predicting either decreases (up to 20%), increases (up to 20%) or no change in rainfall by 2050 on a medium (SRES A1B) CO2 emissions scenario (see Suppiah et al., 2010). To examine the potential effects of climate change on lower and upper montane rainforest canopy wetness we have therefore perturbed rainfall by ±20% (as in Wallace and McJannet (2012)). GCM’s also include predictions of how the cloud condensation level will change; however, these are not very accurate in individual grid cells, especially near steep topography (Still et al., 1999). Furthermore, information on the change in cloud base alone is not sufficient to provide an estimate of change in cloud interception as this is not well correlated with altitude alone. In the absence of any better information we have therefore estimated a range of potential effects of climate change on rainforest cloud interception (Pc) by assuming that cloud frequency (fc) changes by a similar percentage as rainfall. We chose to perturb fc by ±20% rather than Pc, since much of the cloud interception occurs when it is not raining and, as we shall see later, there is a good relationship between Pc and fc. The above GCM climate change studies include temperature changes, predicted to be between 1 and 3 K by 2050 (Suppiah et al., 2010), so our perturbation fc by ±20% should also include the effects of temperature change. However, we also discuss how climate induced temperature change alone might affect the cloud condensation level. 3. Results and discussion 3.1. Canopy water inputs The contribution of cloud interception (Pc) to the total water input to the forest canopy (Pg + Pc) is shown in Fig. 2 for the four rainforest locations. The number of months averaged at each site (from 1 to 3.5) is determined by the length of each data set (Table 2) and is shown on the x-axis. Most of the rainfall and much of the cloud 88 J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 Fig. 2. Seasonal variation in monthly rainfall ( ) and cloud interception ( ) for four rainforest locations. Also shown is the percentage of total water input occurring as cloud interception (h). The number of months averaged is shown in brackets on the x-axis for each site. Note the y-axis scale for Bellenden Ker is different from the other three sites. interception occurs during the wet season (December to May). However, significant cloud interception continues during the dry season and this becomes an increasing fraction of the total water input to the canopy at this time. The least cloud interception occurred at Mount Lewis 1, reaching a monthly maximum of 15% of the canopy water input and averaging 5% over the entire year 89 J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 Table 2 Annual average rainfall (Pg), cloud interception (Pc) and total water input (Pg + Pc) to the forest canopy at four rainforest locations. Also shown are the start and end dates of the different periods of data availability at each site. Observation period Mount Lewis 1 Mount Lewis 2 Upper Barron Bellenden Ker 01-June-2001 01-June-2001 06-July-2001 03-October-2002 17-June-2004 17-June-2004 11-June-2004 11-June-2004 Duration (days) Pg (mm) Pc (mm) Pg + Pc (mm) Pc (%) 1113 1113 1072 618 3172 2633 1675 6880 181 390 715 2090 3353 3023 2390 8970 5 13 30 23 from 69% at BK to 19% at ML1. However, as cloud may not be present for the entire day these values will be overestimates, or at best an upper limit. Several studies of cloud interception have estimated cloud frequency by assuming cloud presence when the relative humidity is >95% (e.g. Dasch, 1988; Saxena and Lin, 1990). We estimated fc from our 30 min relative humidity data for periods of 618, 1072 and 1113 days at BK, UB and ML1 and 2 respectively, Table 3. It is clear that this method can grossly overestimate fc, especially at the lower montane sites UB, ML1 and ML2, since the values are well in excess of the upper limit derived in the first method. The third estimate of fc was made using 30 min recordings of cloud interception and this yielded much lower values of cloud frequency. These values can be regarded as minimum values of fc for each site, since they are based on the sum of the 30 min occasions when the cloud gauge recorded a tip (or more) on its tipping bucket. Cloud interception will have been occurring before the first tip, in gaps between tips as the gauge collected enough water for the next tip and again for a period after the last tip. A fourth estimate of fc was therefore made that included an allowance for these periods by including 30 min periods where the relative humidity was >95% for periods up to 6 h after the last tip. The period after the last tip was chosen as 6 h as this was the maximum time taken for the cloud interception tipping bucket to fill (derived by dividing the bucket size (mm) by the average cloud interception minus the wet canopy evaporation rate). Sensitivity analysis revealed that fc was not very sensitive to the threshold value of relative humidity (95%), i.e. a 4% point change in RH gives a 5–10% change in fc. Similarly, fc is not too sensitive to the post-last tip duration value, where a ± 50% change only leads to a 5–20% change in fc. The values of fc derived using the fourth method above are therefore considered the best estimates, noting they are approximately half way between the alternative method minimum and maximum values (Table 3). Comparison of the two similar altitude Mount Lewis forests illustrates the effect of site exposure on cloud frequency, with the forest on the exposed (windward) hill slope (ML2) experiencing twice the cloud frequency as the nearby forest in a shallow valley on the leeward side of the hill (ML1). Clouds were present for over 50% of the time at the highest and most exposed site (BK); 3–6 times as long as the other forest sites. The seasonal pattern of cloud frequency derived using the fourth method above at the four rainforest sites is shown in Fig. 3. Also shown are the monthly mean values of cloud interception (Pc) and wind speed recorded during the interception periods. The number of months averaged at each site is the same as in Fig. 2. At all four sites cloud frequency and interception are highly seasonal with maximum values between March and July and minimum rates in October. Cloud frequency varied greatly between (Table 2). This site is in a shallow valley and is less exposed to the prevailing winds than the nearby Mount Lewis 2 site on an exposed ridge, where much more cloud interception occurred. For example, the monthly Pc at ML2 reached 30% in June and July and the annual average Pc was 13%. The relative contribution of Pc was even greater at Upper Barron, peaking at close to 70% of the water input in June and averaging 30% of the annual water input (Fig. 2;Table 2). The largest amounts of cloud interception occurred at Bellenden Ker at just over 2000 mm per annum; again dry season cloud water inputs at this site were between 40% and 50% of the total. The cloud interception rates at ML1 (0.5 mm day1), ML2 (1.1 mm day1) and UB (1.4 mm day1) are very similar to the typical rates of interception of passing cloud reported by Bruijnzeel (2005) and Bruijnzeel et al. (2012) in their reviews of tropical montane cloud forests (range 0.2–4 mm day1). The Pc rate for BK (5.7 mm day1) is greater than the range given by Bruijnzeel (2005), but there have been other studies of cloud interception in rainforests that found high rates between 5 and 6 mm day1, e.g. by Cavelier et al. (1996), Brown et al. (1996) and Holder (1998). It is also possible that the Pc values at BK were enhanced by wind blown rain, although corrections for this have been made in the rainfall estimate (see McJannet et al., 2007b). 3.2. Cloud frequency and interception The amount of water intercepted by a forest is determined by a number of factors including wind speed, cloud liquid water content and the geometry of the forest leaves and branches (Bruijnzeel, 2005). A number of process models have been developed to describe cloud interception to forest canopies (e.g. Shuttleworth, 1977; Lovett, 1984; Slinn, 1982), but their practical application is limited by lack of information on key parameters such as droplet size distribution and liquid water content. Although cloud interception is a complex process, Vong et al. (1991) have presented data from a number of studies in the Appalachian Mountains that indicate a positive relationship between annual cloud interception and cloud frequency (fc). This kind of relationship allows estimates of climate impacts on cloud interception to be made if we know how fc might vary with climate. The frequencies with which the Australian rainforest locations, UB, ML1, Ml2 and BK are immersed in cloud were derived from measurements of the occurrence of cloud interception and high relative humidity. The four methods used to estimate fc are summarised in Table 3. The first estimate of fc is made from daily data, where any day when cloud interception is detected by the cloud gauge is designated as a cloud day. This gives values for fc ranging Table 3 A summary of four different estimates of cloud frequency at the Australian rainforest sites, Mount Lewis 1 and 2, Upper Barron and Bellenden Ker. 1. 2. 3. 4. % % % % of days with cloud interception time when relative humidity (RH) > 95% of time when cloud interception recorded of time when cloud interception recorded and afterwards if RH > 95% Mount Lewis 1 Mount Lewis 2 Upper Barron Bellenden Ker 19 68 1 8 30 68 3 16 35 63 6 18 69 62 35 51 90 J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 Fig. 3. Seasonal variations in monthly average cloud frequency (h), cloud interception () and wind speed (+) at four rainforest locations. Note the vertical axes for Bellenden Ker have different scales to those for the other three sites. the different sites; with the lowest values at ML1 and the highest at BK. Within individual sites there was little or no correlation between cloud interception (Pc) and wind speed during the interception periods (r2 values ranged from 0.00 to 0.34). However, the highest Pc rates were observed at the windiest site (BK; mean wind speed 5.1 m s1) and the lowest Pc rates at the calmest site (ML1; mean wind speed 1.1 m s1). Cloud interception was highly correlated with cloud frequency (fc) at all four sites, where the individual linear regressions (through the origin) for each of the sites (not shown in Fig. 4 for clarity) had slopes that increased from 1.9 (r2 = 0.99) at ML1, 2.1 at ML2 (r2 = 0.81), 3.5 at UB (r2 = 0.78) and 3.6 at BK (r2 = 0.64). The lower correlation coefficient at BK may, in part, be due to the smaller number of months averaged for the shorter data set at this site (see Table 2 and Fig. 2). The best fit line to all the Australian data points in Fig. 4a is curvilinear, as this allows for the combined effects of cloud frequency and wind speed on cloud interception. The physical explanation for this may be twofold; J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 91 Fig. 4. (a) The relationship between monthly cloud interception (Pc) and cloud frequency (fc) for four Australian rainforest sites (Mount Lewis 1 (h), ML1; Mount Lewis 2 (s), ML2; Upper Barron (N), UB and Bellenden Ker (), BK). The fitted line has the form Pc = 0.039 fc2 + 1.30 fc; r2 = 0.87. For comparison the linear regression through the origin Pc = 3.44 fc; r2 = 0.79 is also shown. (b) equivalent data from a number of forested sites in the Appalachian Mountains (Vong et al., 1991;Bruijnzeel, 2005) along with the curvilinear regression from Fig. 4a. (1) for a given amount of time more cloud interception occurs at the higher altitude sites where the higher wind speeds force more cloud droplets per unit time past the condensation structures in the forest and (2) theory suggests that the liquid water content within clouds increases with altitude (e.g. Korolev et al., 2007). For comparison, cloud interception and frequency data from a number of studies of cloud interception in the Appalachian Mountains in the eastern United States (US) of America by Vong et al. (1991) (with additional data reported by Bruijnzeel (2005)) are shown in Fig. 4b. It should be noted that the US data uses different methods and periods to estimate cloud interception and there are considerable uncertainties in the estimation of cloud frequency. For example, many of the Appalachian Mountain studies were quite short (1–2 months) and fc was derived from relative humidity (e.g. Dasch, 1988; Saxena and Lin, 1990), shown here (Table 3) to be an overestimate, or inferred from nearby studies at different times (Reiners and Lang, 1979). However, despite these uncertainties it appears that the Australian rainforests may be more efficient (by 50% on average) in intercepting cloud water than the Ameri- can coniferous forests, which may be due to differences in canopy structure and exposure at the different sites. The fitted curve in Fig. 4 can be used to derive predictions of how cloud interception may change if the change in cloud frequency can be predicted. In the absence of any better information (discussed further below), cloud frequency is assumed to change by the same amount as future rainfall change, which is ±20% for the Australian rainforest locations. A 20% change in fc produces a larger relative change in cloud interception because of the increasing slope of the best fit curve in Fig. 4a. Furthermore, as fc increases, the same relative change in fc produces a larger relative change in cloud interception. This amplification (ratio of the relative change in Pc to the relative change in fc) is around 1.1 when fc is 5% and grows to 1.7 when fc is 70%. This implies that changes in cloud frequency will be amplified in cloud interception at all rainforest sites, with the greatest amplification at the higher altitude sites such as Upper Barron and Bellenden Ker. Again this is consistent with what would be expected as both wind speed and/or cloud liquid water content increase with altitude. 92 J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 A number of studies have examined the potential impacts of climate change on tropical forests that are exposed to cloud interception (e.g. Pounds et al., 1999; Still et al., 1999; Foster, 2001). Foster (2001) reports global and regional climate modelling studies that indicate that there would be an increase in dry days due to a rise in the cloud level. For a 2 CO2 scenario she reported that absolute humidity surfaces would move upslope by 200–300 m, similar to values reported by Still et al. (1999). Other studies have reported higher estimates of the shift in cloud base height, i.e. up to 500 m for a 3 K rise in temperature (Peters and Darling, 1985; Halpin, 1994), which is very close to the moist adiabatic lapse rate (Held and Soden, 2000). However, as the actual amount of cloud interception is not well correlated with altitude alone (Vong et al., 1991; McJannet et al., 2007a), it is difficult to use this approach to estimate the quantitative change in cloud interception as the climate temperature changes. Climate change predictions are made using General Circulation Models (GCM’s) that represent the key climate processes in a multi-layered atmosphere. Their most common outputs are predictions of rainfall and temperature change that occur as the concentration of atmospheric CO2 is increased. However, these models (or their nested higher resolution meso-scale models) also include predictions of how the cloud condensation level will change, although Still et al. (1999) point out that these models are not expected to produce highly accurate predictions of the amount of water vapour in individual grid cells, especially near steep topography. This location specific uncertainty is also recognised in the recent GCM study by Foster (2010) who report small and/or insignificant changes in the cloud condensation level in northern Queensland up to 2050. Their work highlights the importance of knowing how relative humidity will change as the atmosphere warms, since where this is small (as suggested by Held and Soden (2000)) then changes in the cloud condensation level will also be small. Even if it were possible to more accurately predict climate induced changes in cloud condensation level, without any commensurate indication of how cloud frequency might change, this information is of limited use in predicting quantitative changes in cloud interception as climate temperature changes. Much of the biodiversity of tropical montane rainforests depends on the availability of water in the forest canopy, which is highly regulated by cloud interception. For example, in their study of birds, lizards and frogs in the highland rainforests at Monteverde in Costa Rica, Pounds et al. (1999) reported that altitudinal shifts in these faunal populations were most highly correlated with variations in cloud frequency, as indicted by total dry days. They found an increase in the frequency of dry days, which they ascribe to a raising of the cloud level due to atmospheric warming. They also found that changes in bird populations and a decline in lizard and frog numbers were all associated with changes in cloud frequency. 3.3. Canopy wetness duration The seasonal variation in canopy wetness duration due to rainfall and cloud interception is shown in Fig. 5. Wetness due to rain (in the absence of cloud interception), cloud (in the absence of rainfall) and the simultaneous occurrence of cloud and rain are shown separately. At the lower montane sites (ML1, ML2 and UB) wetness durations follow the pattern of long term rainfall and are highest around February/March (40–50%) and lowest in September/October (10–20%). Wetness due to cloud interception follows the different seasonal pattern of cloud frequency (see Fig. 3) and makes the smallest contribution in the summer months, but becomes the dominant wetting mechanism in winter. For example, throughout the months from May to September canopy wetness at ML2 and UB is more influenced by cloud interception than by rainfall alone. The pattern of canopy wetness at the highest altitude site (BK) is quite different. Here wetness durations are much higher, 60–80% for over 9 months and are dominated by cloud interception throughout the year. In the winter months (June, July and August) canopy wetness is almost entirely due to cloud passing through the canopy, even though it only constitutes less than half of the water input to the canopy. This is because for a given amount of water input to the canopy, wetness durations due to cloud interception are much longer than those due to rainfall as the rate of cloud interception is so much slower (one fifth) than typical rainfall rates (Wallace and McJannet, 2008). Potential climate impacts on canopy wetness duration at Mount Lewis 1 and Bellenden Ker is shown in Fig. 6. Climate change has less effect on wetness duration at ML1 as wetness is more dependent on rainfall at this site. This is because changes in rainfall (of ±20%) can reduce the duration of wetness during rain, but not after rain; also discussed in Wallace and McJannet (2012), who found that depending on the relative contributions of these two periods, wetness duration due to rainfall changed by between 10% and 15% for a 20% change in rainfall. On the other hand, since wetness durations at Bellenden Ker are much more dominated by cloud interception, a 20% change in this produces a similar change in canopy wetness duration. Any given change in rainfall and cloud interception will therefore have a greater impact on canopy wetness at the higher altitude sites where cloud interception is a larger fraction of total canopy water input. However, the ecological implications of wetness duration changes at higher and lower altitude sites will depend on how close the species are to their desiccation limits. 3.4. Downstream impacts The seasonal variation in excess water (R + D + dh) generated by both rainfall and cloud interception is shown in Fig. 7. Excess water was calculated on a monthly basis as the difference between total water input (Pg + Pc) and total evaporation (I + Et + Es) using the SILO long term rainfall data for ML1 and BK. Interception (I) and transpiration (Et) were calculated from monthly (Pg + Pc) using the regression equations derived for these sites by Wallace and McJannet (2012). Es is a small component and was assumed to be fixed at 0.2 mm day1. Cloud interception (Pc) was derived for each month using the seasonally varying cloud frequencies (fc) shown in Fig. 3 in the Pc versus fc relationship shown in Fig. 4. These calculations were repeated for normal, +20% and 20% rainfall and cloud frequency. At the lower montane Mount Lewis 1 site excess water is dominated by rainfall during the summer and as this declines in the winter months there is a period when total evaporation exceeds rainfall and so (R + D + dh) becomes negative, implying that the forest is dependent on soil moisture and/or groundwater at this time. Under normal rainfall conditions the duration of this period is 96 days and this changes by 44 and + 23 days respectively as rainfall and cloud frequency are altered by +20% and 20%. As reported by Wallace and McJannet (2012) for other rainforest sites, the climate induced shift in the start of the zero excess water period is greater than the shift in the end date. Excess water is always positive at Bellenden Ker (Fig. 7b), but there are large changes associated with the ±20% rainfall and cloud frequency climate change scenarios. As cloud interception is very large at BK (Fig. 3), the contribution that the 20% change in cloud frequency makes to the change in excess water becomes quite significant in the winter months, accounting for up to 30% of the total change in (R + D + dh). A given change in rainfall and cloud frequency produces a larger change in (R + D + dh), similar to that reported by Wallace and McJannet (2012) for rainfall alone. The greatest amplification (ratio 93 J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 Canopy wetness duration (%) 100 (a) Mount Lewis 1 90 80 70 60 50 40 30 20 10 0 Jan Feb Mar Apr Canopy wetness duration (%) 100 May Jun Jul Aug Sep Oct Nov Dec Sep Oct Nov Dec (b) Mount Lewis 2 90 80 70 60 50 40 30 20 10 0 Jan Feb Mar Apr Canopy wetness duration (%) 100 May Jun Jul Aug (c) Upper Barron 90 80 70 60 50 40 30 20 10 0 Jan Feb Mar Apr Canopy wetness duration (%) 100 May Jun Jul Aug Sep Oct Nov Dec (d) Bellenden Ker 90 80 70 60 50 40 30 20 10 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Fig. 5. Seasonal variation in canopy wetness duration at four rainforest locations. Wetness due to rain with no canopy cloud (black), rain and cloud occurring together (mid grey) and cloud occurring in the absence of rain (light grey) are shown separately for each site. of the % change in (R + D + dh) over the % change in Pg and fc) occurs at ML1, where the ratio is 3.4 in June and 1.7 for the entire year. The amplification at BK is smaller, around 1.6 in winter and 1.3 as an annual average. The amount of amplification is greater when rainfall is lower because of the highly non-linear relationship between (R + D + dh) and rainfall (see Wallace and McJannet, 2012). Similarly, non-linearity in the relationship between cloud intercep- tion and cloud frequency (Fig. 4) also contributes to amplification of changes in (R + D + dh). 4. Conclusions This paper confirms that cloud interception is a very significant water input to rainforests in north Queensland, especially during 94 J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 100 Canopy wetness duration (%) 90 Bellenden Ker 80 70 60 50 40 30 Mount Lewis 1 20 10 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Fig. 6. Seasonal variation in monthly mean canopy wetness duration at Mount Lewis 1 and Bellenden Ker under normal rainfall and cloud frequency (h, j), 20% more rainfall and cloud (D,N) and 20% less rainfall and cloud (e, ). 600 (a) Mount Lewis 1 500 400 300 200 100 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec -100 1800 60% (b) Bellenden Ker 1600 1200 40% 1000 30% 800 600 20% 400 10% Cloud frequency change contribution 50% 1400 200 0 0% Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Fig. 7. The seasonal variation in excess water (R + D + dh) at (a) Mount Lewis 1 and (b) Bellenden Ker under normal rainfall and cloud frequency (), 20% more rainfall and cloud frequency (j) and 20% less rainfall and cloud frequency (N). The contribution that cloud frequency change makes to the overall change in (R + D + dh) is also shown for Bellenden Ker. the dry season when it can exceed the rainfall input. Cloud interception also makes a major contribution to the length of time the rainforest canopy is wet, again becoming the dominant wetting mechanism during the dry season at all of the locations studied. At the upper montane rainforest (BK), cloud interception is the dominant wetting mechanism throughout the year. The much slower rate of cloud interception (0.8 mm h1; Wallace and McJannet, 2008) compared with rainfall (3.4 mm h1) means that a given amount of cloud interception will produce a wetness duration that is over four times that produced by the same amount of rainfall. This is the main reason why cloud interception has a greater impact on canopy wetness duration. There may be important impacts on cloud interception (Pc) at these higher altitude rainforests due to climate change, especially as any given change in cloud frequency (fc) produces a greater change in cloud interception. This ‘amplification’ (ratio of the % change in Pc to the % change in fc) increases with fc growing from 1.1 when fc is 5% to 1.7% when fc is 70%. This means that any changes in cloud frequency will have an amplified impact on cloud interception at all rainforest sites, with the greatest amplification at the highest altitude sites such as Bellenden Ker where cloud J. Wallace, D. McJannet / Journal of Hydrology 481 (2013) 85–95 interception is a large fraction of total canopy water input. As cloud frequency is also a major factor affecting canopy wetness, any given change in cloud frequency will therefore have a greater impact on canopy wetness at the higher altitude sites. These changes in wetness duration will augment those due to changes in rainfall and may have important implications for the epiphytes, mosses and other species that depend on these wet canopy conditions (Hölscher et al., 2004; Köhler et al., 2006). More combined hydro-ecological studies of canopy wetness and rainforest flora and fauna are needed to address this question. Downstream freshwater species may also be impacted by climate induced changes in rainfall and cloud frequency where again any given climate change will produce a greater change in runoff by up to a factor of 3.4 (seasonally) and 1.7 (annually). Large shifts may also occur in the timing and duration of the winter period when there is no excess water for downstream runoff and any freshwater species that respond to flows at this time may be affected. More certain estimation of these flow regime changes on aquatic biota require better estimates of how climate change will affect not only rainfall, but also condensation levels and the frequency of cloud formation at rainforest canopy levels. Acknowledgements The analysis in this paper are based on extensive field data collected with the help of Mark Disher, Andrew Ford, Peter Richardson, Trevor Parker, Adam McKeown, Trudi Prideaux, Jenny Holmes and Pepper Brown. We are also grateful to the internal CSIRO reviewers and the external referees for their helpful comments and advice on the draft manuscript. Funding for this research was provided by the Cooperative Research Centre for Tropical Rainforest Ecology and Management. References Brown, M.B., de la Roca, I., Vallejo A., Ford, G., Casey, J., Aguilar, B., Haacker, R., 1996. A Valuation Analysis of the Role of Cloud Forests in Watershed Protection. Sierra de las Minas Biosphere Reserve, Guatemala and Cusuco N.P., Honduras. RARE Center for Tropical Conservation: Philadelphia, USA; 133. Bruijnzeel, L.A., 2005. Tropical montane cloud forest: a unique hydrological case. In: Bonell, M., Bruijnzeel, L.A. (Eds.), Forests, Water and People in the Humid Tropics. Cambridge University Press, Cambridge, pp. 462–483. Bruijnzeel, L.A., Scatena, F., Hamilton, L.S. (Eds.), 2012. Tropical Montane Cloud Forests: Science for Conservation and Management. Cambridge University Press, Cambridge UK, 740 pp. Cai, W., Crimp, S., McInnes, K., Hunt, B., Suppiah, R., Collier, M., Elliot, T., Henessy, K., Jones, P., Page, C., Whetton, P., 2003. Climate change in Queensland under enhanced greenhouse conditions. Annual Report 2003 No C/0708, CSIRO Atmospheric Research. Aspendale, Victoria, 72pp. Cavelier, J., Solis, D., Jaramillo, M.A., 1996. Fog interception in montane forests across the Central Cordillera of Panama. J. Trop. Ecol. 12, 357–369. CSIRO, 2007. Climate change in Australia – Technical Report 2007. CSIRO Marine and Atmospheric Research. Aspendale, Victoria. <http:// www.climatechangeinaustralia.gov.au/index.php>. Dasch, J.M., 1988. Hydrological and chemical inputs to fir trees from rain and clouds during a 1-month study at Clingmans Peak, NC. Atmos. Environ. 22, 2255–2262. Foster, P., 2001. The potential negative impacts of global climate change on tropical montane cloud forest. Earth Sci. Rev. 55, 73–106. Foster, P., 2010. Changes in mist immersion. In: Bruijnzeel, L.A., Scatena, F.N., Hamilton, L.S. (Eds.), Tropical Montane Cloud Forests: Science for Conservation and Management. Cambridge University Press, 57–66. Gash, J.H.C., Lloyd, C.R., Lachaud, G., 1995. Estimating sparse forest rainfall interception with an analytical model. J. Hydrol. 170, 79–86. Halpin, P.N., 1994. Latitudinal variation in montane ecosystem response to potential climate change. In: Beniston, M. (Ed.), Mountain Ecosystems in Changing Climates. Routledge Publishing, London, pp. 180–203. 95 Held, I.M., Soden, B.J., 2000. Water vapour feedback and global warming. Ann. Rev. Energy Environ. 25, 441–475. Holder, C.D., 1998. Fog precipitation in the Sierra de las Minas Biosphere Reserve, Guatemala. In: Schemenauer, R.S., Bridgman, H.A. (Eds.), First International Conference on Fog and Fog Collection. ICRC, Ottawa. Hölscher, D., Köhler, L., van Dijk, A.I.J.M., Bruijnzeel, L.A., 2004. The importance of epiphytes to total rainfall interception by a tropical montane rain forest in Costa Rica. J. Hydrol. 292, 308–322. Jeffrey, S.J., Carter, J.O., Moodie, K.B., Beswick, A.R., 2001. Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environ. Modell. Softw. 16 (4), 309–330. Köhler, L., Hölscher, D., Tobón Marin, C., Frumau, A., Bruijnzeel, L.A., 2006. Biomass and water storage of epiphytes in old-growth and secondary montane rain forest in Costa Rica. Plant Ecol. 193, 171–284. Korolev, A.V., Issac, G.A., Strapp, W., Cober, S.G., Barker, H.W., 2007. In situ measurements of liquid water content profiles in midlatitude stratiform clouds. Quart. J. R. Meteorol. Soc. 133, 1639–1699. Li, L.T., Donohue, R.J., McVicar, T.R., Van Niel, T.G., Ten, J., Potter, N.J., Smith, I.N., Kirono, D.G.C., Bathols, J.M., Cai, W., Marvanek, S.P., Gallant, S.N., Chiew, F.H.S., Frost, A.J., 2009. Climate data and their characterisation for hydrological scenario modelling across northern Australia. In: A report to the Australian Government from the CSIRO Northern Australia Sustainable Yields Project. CSIRO Water for a Healthy Country Flagship, Australia, 63 pp. Lovett, G.M., 1984. Rates and mechanisms of cloud water deposition to a subalpine balsam fir forest. Atmos. Environ. 18, 361–371. McGregor, J.L., Nguyen, K.C., 2010. An ensemble of high-resolution climate simulations for the MTSRF region. In: Marine and Tropical Sciences Research Facility (MTSRF) Annual Conference 2010, Cairns, Australia. Reef & Rainforest Research Centre Ltd., Cairns. <http://www.rrrc.org.au/news/downloads/ john_mcgregor_day2_climatechangeonland.pdf>. McJannet, D.L., Wallace, J.S., Reddell, P., 2007a. Precipitation interception in Australian tropical rainforests: I. Measurement of stemflow, throughfall and cloud interception. Hydrol. Process. 21, 1692–1702. McJannet, D.L., Wallace, J.S., Reddell, P., 2007b. Precipitation interception in Australian tropical rainforests: II. Altitudinal gradients of cloud interception, stemflow, throughfall and interception. Hydrol. Process. 21, 1703–1718. McJannet, D.L., Wallace, J.S., Fitch, P., Disher, M., Reddell, P., 2007c. Water balance of tropical rainforest canopies in north Queensland, Australia. Hydrol. Process. 21, 3473–3484. McJannet, D.L., Fitch, P., Disher, M., Wallace, J.S., 2007d. Measurements of transpiration in four tropical rainforest types of north Queensland, Australia. Hydrol. Process. 21, 3549–3564. Peters, R.L., Darling, J.D.S., 1985. The greenhouse effect and nature reserves: global warming would diminish biological diversity by causing extinctions among reserve species. Bio-science 35, 707–717. Pounds, J.A., Fogden, M.P.L., Campbell, J.H., 1999. Biological response to climate change on a tropical mountain. Nature 398, 611–615. QCCCE, 2010. Climate change in Queensland: what the science is telling us. Queensland Climate Change Centre of Excellence. Department of Environment and Resource Management, Brisbane, Queensland, 95 pp. Reiners, W.A., Lang, G.E., 1979. Vegetational patterns and processes in the balsam fir zone, White Mountains, New Hampshire. Ecology 60, 403–417. Saxena, V.K., Lin, N.-H., 1990. Cloud chemistry measurements and estimates of acidic deposition on an above cloud base coniferous forest. Atmos. Environ. 24A, 329–352. Shuttleworth, W.J., 1977. The exchange of wind-driven fog and mist between vegetation and the atmosphere. Bound.-Layer Meteorol. 12, 463–489. Slinn, W.G.N., 1982. Predictions for particle deposition to vegetative canopies. Atmos. Environ. 16, 1785–1794. Still, C.J., Foster, P.N., Schneider, S.H., 1999. Simulating the effects of climate change on tropical Montane cloud forests. Nature 398 (6728), 608–610. Suppiah, R., Watterson, I.G., Macadam, I., Collier, M.A., Bathols, J. 2010. Climate Change Projections for the Tropical Rainforest Region of North Queensland. Final Report on MTSRF Activities, 2006–2010. Report prepared for the Marine and Tropical Sciences Research Facility (MTSRF) Research Report Series. Published by the Reef & Rainforest Research Centre Ltd., Cairns. 81pp. Tracey, J.G., 1982. The Vegetation of the Humid Tropical Region of North Queensland. CSIRO, Melbourne, FL, 124. Vong, R.J., Sigmon, J.T., Mueller, S.F., 1991. Cloud water deposition to Appalachian forests. Environ. Sci. Technol. 25, 1014–1021. Wallace, J.S., McJannet, D.L., 2008. Modelling interception in coastal and montane rainforests in northern Queensland, Australia. J. Hydrol. 348, 480–495. Wallace, J.S., McJannet, D.L., 2012. Climate change impacts on the water balance of coastal and montane rainforests in north Queensland. Australia. J. Hydrol. 475, 84–96.