Download Improving the Depiction of Moisture Transport in Short

Improving the Depiction of Moisture Transport in Short-Range
Forecasts of the Pre-Convective Environment
1,2
1
William E. Line , Ralph Petersen and Robert Aune
3
1
CIMSS/SSEC/University of Wisconsin – Madison, Madison, Wisconsin
Current: CIMMS/University of Oklahoma and NOAA/NWS/Storm Prediction Center, Norman, Oklahoma
3
NOAA/NESDIS/STAR, Advanced Satellite Products Branch, Madison, Wisconsin
2
Abstract
The CIMSS NearCast model is a Lagrangian model that dynamically projects GOES moisture
and temperature observations forward in time to provide detailed, hourly updated information about the
vertical moisture and stability structure of the pre-convective environment 1-9 hours in advance. Since
these observations are made from the clear sky where flow is mostly adiabatic, an isentropic version of
the model has been developed to more accurately project the observations through space in threedimensions. In addition to providing more accurate stability information, this new version of the model
depicts adiabatic vertical motion as well as total isentropic layer mass and moisture content, as shown in
an April 09, 2011, case study.
INTRODUCTION
Predicting the location and timing of excessive precipitation and other hazardous weather events
remains a challenge to forecasters today, especially in the warm season (Fig. 1). This is partially due to
the inability of current numerical weather prediction (NWP) products to capture and predict mesoscale
moisture features accurately. Although satellite retrieval observations reveal details about the vertical
moisture and stability structure of the atmosphere, they remain a highly under-utilized dataset.
Furthermore, GOES soundings of temperature and moisture over land are not being assimilated into any
current NWP models.
Figure 1: 2010-2012 monthly averages of National Centers for Environmental Prediction (NCEP) Global Forecast System
(GFS) and North American Mesoscale Model (NAM) 3-hourly precipitation forecast threat scores. Equitable threat scores
are a measure of the model forecast accuracy when compared to what was observed. Data obtained from the NCEP/EMC
webpage.
The CIMSS NearCast model is a data-driven, quick updating tool that takes advantage of underutilized satellite retrieval data to provide forecasters with analyses and 1-9 hour forecasts of the preconvective environment. The simple model uses a Lagrangian trajectory approach to dynamically project
the temperature and moisture observations forward in space and time at multiple atmospheric levels.
This thermodynamic information is used to help forecasters make short-term predictions about where and
when convection is most (and least) likely to occur, as well as whether existing convection will persist or
dissipate in the near future. The technique preserves fine details (moisture minima, maxima, and
boundaries) present in the observations and retains up to 10 hours of previous observations in its
analysis and forecast products to fill in data gaps caused by cloud contamination.
The original configuration of the NearCast model was in isobaric coordinates, meaning
trajectories are made along constant pressure surfaces. However, since the retrievals are made from the
clear sky where diabatic processes are at a minimum, the flow is assumed to be adiabatic. In an adiabatic
atmosphere, flow is isentropic with three-dimensional moisture transport through constant pressure
surfaces instead of along them (Oliver and Oliver, 1951). Given the success of the isobaric version of the
NearCast model in improving forecasts of the timing and location of convection, and the assumption that
the flow is adiabatic, an isentropic version of the NearCast model has been developed which projects the
observations along constant potential temperature surfaces. Not only does the isentropic NearCast model
more accurately move the observations through space, but it allows for the depiction of adiabatic vertical
motion in the atmosphere, further narrowing down where convective development is most and least likely
to occur. Additionally, the isentropic model retains information about total isentropic layer moisture
content, which can be used to help predict the occurrence of high-precipitation events.
DATA
The temperature and moisture sounding observations (retrievals) that are used in the NearCast
model over the United States and discussed in this paper come from the GOES sounding device. Details
into the GOES sounding retrieval process can be found in Li et al. (2009). The information observed by
the GOES sounder has been shown to correct errors in the GFS model first guess especially in the warm
season, when NWP short-term precipitation forecast skill is poorest (Petersen et al., 2012).
NWP wind and geopotential height field forecasts are used as initial conditions and help compute
the parcel trajectories in isobaric coordinates. For the examples shown here, these data were obtained
from the GFS 0.5 degree model output at 50 hPa intervals in the vertical. For the isentropic NearCast
model, a linear interpolation is applied to move the data to isentropic coordinates at 2 K intervals using
the General Meteorological Package (GEMPAK). The isentropic fields derived from the GFS data include
winds, Montgomery stream function, and pressure all on constant potential temperature surfaces.
METHODS
The NearCast model uses an explicit method for computing parcel trajectories based off of
Petersen and Uccellini (1979). In short, the wind field is used at the first timestep to initiate the parcel
trajectories, and mass field gradients (geopotential height in isobaric coordinates, Montgomery stream
function in isentropic coordinates) are used to accelerate the parcels forward in space and time at all
subsequent timesteps. The trajectories are made on constant pressure surfaces in isobaric coordinates
and constant potential temperature surfaces in isentropic coordinates. The levels used in each coordinate
system are chosen based on their respective close proximities to the lower and upper pressure levels
where the GOES weighting functions indicate independent moisture information.
The model is run out to 10 hours, provides half-hourly forecast output to nine hours, and uses a
10-minute timestep which leads to relatively quick run-times of 1-2 minutes. The half hourly parcel data
are saved from 10 successive cycles so that they can be used to enhance output of future model runs by
filling in data void areas where retrievals were not available in areas of cloud cover. These trajectories are
computed at an upper and lower level in both the isobaric and isentropic versions of the model. The
NearCast model updates every hour as new GOES observations become available.
In both models, equivalent potential temperature ( ) is computed from the GOES sounder
temperature and moisture observations for each level on which the NearCast model is run. Layer stability
is determined by computing the difference between
at the upper and lower level. The vertical
difference, or lapse rate, reveals whether the atmosphere is convectively stable or unstable, as well as
the degree of stability. When the difference is negative (positive), is decreasing (increasing) with
height through the layer, so the atmosphere is convectively unstable (stable). The deep-layer difference
parameter provides an objective way to show when and where dry, cool (warm, moist) air at the upper
levels is advancing over relatively warm, moist (dry, cool) air at the lower levels, leading to convective
instability (stability). Given the proper forcing, regions of relatively high convective instability are likely to
experience convective development in the near future. Furthermore, tendencies in the convective
instability parameter reveal areas that are destabilizing the fastest, further narrowing down when and
where convection is most likely to occur.
Additional information is made available by running the NearCast model in an isentropic
framework. Vertical motion ( ) in isentropic coordinates can be derived from the expansion of the total
derivative of pressure, resulting in three terms (Moore, 1987):
⃗
,
(1)
A
B
C
where A is the local pressure tendency term, B is the advection of pressure on an isentropic surface, and
C is the diabatic heating/cooling term. Since the observations used in the NearCast model are made from
clear-sky atmosphere, diabatic processes are at a minimum, so term C is neglected. The sum of terms A
and B reveals the adiabatic vertical motion on an isentropic surface, where negative (positive) values of
indicate upward (downward) vertical motion.
Another unique source of information in the observations that is retained by running the NearCast
model in an isentropic framework is the total moisture in each isentropic layer computed as
.
represents the total mass in a layer centered at an isentropic level and averaged over a depth of 2 K,
where is inverse static stability. The total amount of moisture being transported adiabatically (the Total
Isentropic Layer Moisture) can be determined by multiplying the layer mass term by the average mixing
ratio ( ) in the layer, with higher (lower) values indicating more (less) moisture in the isentropic layer.
APRIL 09, 2011, CASE STUDY
On the evening of April 09, 2011, a ¾ mile wide EF3 tornado damaged or destroyed nearly twothirds of the town of Mapleton, Iowa (Gallagher and Dreeszen, 2011). Deep convection associated with
this tornadic thunderstorm began its initial development in far east-central Nebraska around 2200 UTC
moving east-northeast (Fig. 2). The tornadic cell, which also produced large hail and damaging wind
along its path, was not associated with widespread, heavy rainfall amounts (Fig. 2). The area of
convection traveled northeast into north-central Iowa by 0300 UTC before weakening considerably. By
this time, a new area of less severe but more widespread and longer-lasting convection had developed in
southeast Minnesota, ahead of the weakening earlier convection. These new storms produced much
heavier and widespread rainfall as they tracked across southeast Minnesota into northeast Wisconsin by
th
the early morning hours of April 10 . The performance of the isobaric NearCast model leading up to this
event will be discussed first, followed by that of the isentropic model.
Figure 2: On left, 3-hourly sequence of 10.7
imagery measured by the GOES-East imager on April 09-10 2011. On right,
Storm Prediction Center (SPC) severe storm reports for the 24 hour period ending 1200 UTC on April 10, 2011. Inset,
observed precipitation for the same period, obtained from the National Weather Service (NWS) Advanced Hydrologic
Prediction Service.
ISOBARIC NEARCAST RESULTS
The 1500 UTC isobaric NearCast cycle on April 09, 2011, initialized approximately 7.5 hours
before the onset of convection in eastern Nebraska, showed the predicted evolution of and convective
instability for the hours leading up to the convective event (Fig 3). At the lower level (780 hPa), there was
fairly strong southerly to southwesterly flow forecast from Texas north into southern Minnesota throughout
the period drawing up warm, moist (high ) air from the south. The model predicted a local maximum
in , originating in central Kansas at 1500 UTC, to move into western Iowa by 0000 UTC behind
relatively cool and dry (low ) air. This meant that a rapid moistening and warming was expected at the
lower levels along a path from central Kansas to western Iowa.
Figure 3: 1500 UTC April 09, 2011, isobaric NearCast model cycle analysis and forecasts out to nine hours, or 0000 UTC.
Upper level is 500 hPa, lower level is 780 hPa, and
difference is upper-level minus lower-level . Higher (lower) values
of
indicate warm and moist (dry and cool) air. Negative (positive) values of
difference indicate the layer is
convectively unstable (stable). Winds are in knots, computed from the trajectories. White areas are where no current or
past (up to 10 hours ago) retrievals were made (or projected into) due to the presence of cloud cover. Black contours are
difference two hour time tendencies, starting at -6K/2 hr, decreasing by 3 K increments. More negative values signify
more rapid destabilization.
In the mid troposphere (500 hPa), values were considerably lower throughout the region,
indicating the air was drier and cooler than below. Flow at this level had a more westerly component to it,
with further drying predicted to occur in eastern Nebraska between 2100 and 0000 UTC. With westerly
winds at the upper levels and weaker southerly flow at the lower levels, the wind shear profile as depicted
by the NearCast model was looking favorable for the support of severe convective development.
The evolution of convective instability during the forecast period was found by taking the
difference between at the upper and lower levels. Initially, there was a local maximum of strong
convective instability in central Kansas with stable air lying just ahead of it to the northeast. The
convective instability was a result of the westerly dry air overlaying the northeast-bound low-level
moisture. By 2100 to 0000 UTC, the instability maximum was predicted to have advanced north-northeast
into eastern Nebraska and northwestern Iowa, an area where destabilization tendencies were also
peaking from previously stable conditions. This is the precise location and timing of the convection that
spawned the EF3 tornado.
Subsequent NearCast model forecasts continued to show the same pattern leading up to the
development and evolution of the tornado-producing convection. Consistency between model runs and
validation of the forecasts by later model analyses provided confidence in the model output. Forecasts
from 2100 UTC onward revealed the instability shifting northeastward from southeast Minnesota through
central Wisconsin by 0600 UTC. Destabilization tendencies and wind shear were both predicted to
become considerably weaker than they had been earlier, perhaps contributing to less-severe convection.
Reasons for the increase in widespread heavy rainfall, however, were unclear from the isobaric version of
the NearCast model
ISENTROPIC NEARCAST RESULTS
The April 09, 2011, case is examined using output from the isentropic version of the NearCast
model, starting with 1500 UTC model run (Fig. 4). Along the lower isentropic surface (312 K), similar to
what was seen in the isobaric model, a maximum was present in central Kansas at the analysis time.
The feature originated just below 750 hPa, close to the 780 hPa level that was used as the lower level in
the isobaric model. The maximum was not forecast to move along at a constant pressure (as in the
isobaric model configuration), but was instead projected to move upward in the atmosphere as it
advanced to the north. By 0000 UTC, the low-level maximum was predicted to have moved to around
650 hPa over northwest Iowa, an ascent of close to 100 hPa over the nine hour time period.
Figure 4: 1500 UTC April 09, 2011, isentropic NearCast model cycle. Similar to Fig. 3, but with trajectories along constant
isentropic surfaces instead of isobaric surfaces. Upper level is 318 K, lower level is 312 K, and red contours are pressure
in hPa.
At the upper, 318 K isentropic surface, adiabatic flow continued to have an upward component
through much of the domain over the forecast period. The deep layer of adiabatic ascent further
supported the potential for convection. The movement of a more significant dry air boundary was also
forecast from the west into eastern Nebraska between 2100 UTC and 0000 UTC. The gradients were
considerably larger on the sloping isentropic surfaces than in isobaric coordinates.
Similar to what is done in the isobaric model, stability in the isentropic model is found by
computing the difference between the upper and lower isentropic surfaces (this is closely related to ,
where is mixing ratio). Comparing Fig. 4 with Fig. 3, both versions of the 1500 UTC model run moved
the maximum in instability roughly along the same horizontal path throughout the cycle. The feature,
however, became more convectively unstable throughout the isentropic cycle due to its more accurate
depiction of a stronger upper-level dry air boundary and movement of low-level moisture. Consequently,
more rapid destabilization tendencies were predicted from eastern Nebraska into northwest Iowa where
convection did eventually develop. The isentropic model also predicted a more pronounced stabilization
in the area of tornadic development before the destabilization, indicating a stronger capping prior to
convective initiation.
The isentropic model is also able to depict wind shear more accurately since it takes into account
changes due to the changing distance between the upper and lower isentropic level at any given location.
Also, as the low-level parcels move upward in a baroclinic atmosphere along the isentropic surface, they
experience additional acceleration associated with the increasing pressure gradient force due to thermal
wind relationships, producing the veering winds and increased vertical speed shear.
Figure 5: 1500 UTC April 09, 2011, isentropic NearCast model cycle adiabatic vertical motion and its two components along
the lower (312 K) isentropic surface. Higher values indicate more rapid upward vertical motion (UVM) or downward vertical
motion (DVM). Units are in
.
The components of the adiabatic vertical motion predicted in the 1500 UTC NearCast cycle at the
312 K isentropic surface are shown quantitatively in Fig 5. The pressure tendency term was greatly
impacted by the movement of a warm (high pressure) thermal ridge across the central United States. The
isentropic surface was lower in the thermal ridge and higher in the colder air behind and ahead of it. As
the thermal ridge advanced eastward through eastern Nebraska, the isentropic surface initially
descended, followed by an ascending pattern by the end of the forecast cycle. The other component
contributing to the adiabatic vertical motion was pressure advection along the isentropic surface. A broad
area of strong negative pressure advection (flow perpendicular to the isobars towards lower pressure)
was present along the northern and eastern edge of the advancing thermal ridge due to substantial warm
air advection. The collocation of significant adiabatic lift in the vicinity of an instability maximum and
strong destabilization tendencies increased confidence that convective development would occur in
eastern Nebraska shortly after 2100 UTC.
Figure 6: 1500 UTC April 09, 2011, isentropic NearCast model cycle average mixing ratio (
layer moisture (
) within the lower (312 K) isentropic layer.
), layer mass (
), and
As mentioned, the isentropic NearCast model retains total isentropic layer moisture information
from the GOES sounding observations. Figure 6 shows the components of this parameter from the 1500
UTC NearCast cycle in the 312 K layer. Over the nine hour period, a 5 g/kg maximum in mixing ratio was
forecast to move along the isentropic surface from central Kansas northeastward into northwest Iowa. A
strip of relatively high layer mass (weak static stability) was oriented just ahead of the moisture maximum
throughout the cycle. Combining the two terms reveals the total isentropic moisture within the layer. Layer
moisture content was greatest where the pocket of highest mass intersected the leading edge of the
highest measured moisture. By 2100 UTC, the plume of highest layer moisture was predicted to have
already moved through northwest Iowa, advancing into north-central Iowa and southern Minnesota by
0000 UTC, well ahead of the strongest convective instability. The lack of heavy/widespread rainfall from
the initial tornado-producing convection was consistent with the smaller amounts of total layer moisture
predicted by the NearCast model. Furthermore, there was enough moisture in the lower layers to support
initial convective growth, but not enough to sustain it over a longer period of time.
The isentropic NearCast cycles initialized between 1500 and 2100 UTC were consistent in
predicting the rapid destabilization, strong low-level adiabatic lift, and relatively low amounts of 312 K
layer moisture coming together in eastern Nebraska between 2100 UTC and 0000 UTC. The 2100 UTC
NearCast model cycle provided information to help make the prediction that future convection would be
much longer-lived and could support higher precipitation amounts from southeast Minnesota into central
Wisconsin (Fig. 7). Between 0000 and 0600 UTC, the low-level maximum was predicted to take on a
more westerly track as it shifted into western Wisconsin. The dry air boundary at the upper level
continued to override the low-level moisture from the west, leading to destabilization from southeast
Minnesota into northeast Wisconsin by the end of the period. The destabilization tendencies were
forecast to be weaker than they were earlier in the cycle because areas ahead of the instability maximum
had already experienced gradual destabilization for several hours due to the changes in flow direction at
both levels to a more uniform westerly track. The shifting winds also led to a less favorable wind shear
environment, further reducing the likelihood of severe convection in this region. There was, however, still
ample low-level adiabatic ascent in the area of convective instability, as the cross-isobar flow was
forecast to remain quite strong, contributing to the eastward shift of the strongest lift.
Figure 7: 2100 UTC April 09, 2011, isentropic NearCast model cycle. Similar information to that found in Fig.’s 4, 5 and 6.
Between the six and nine hour forecasts of the 2100 UTC cycle (valid between 0300 and 0600
UTC), the highest amounts of total moisture in the lower isentropic layer were forecast to be collocated
with areas of convective instability and adiabatic lift from southeast Minnesota through much of central
Wisconsin. Widespread, longer-lasting convection did indeed develop across this region producing
significantly higher precipitation amounts than was seen earlier in eastern Nebraska and northwest Iowa.
SUMMARY AND CONCLUSIONS
The original version of the NearCast model computes trajectories along constant pressure
surfaces and has proven that the Lagrangian trajectory approach using satellite observations can be
successfully applied to make predictions of atmospheric stability. Highlighting some of the key points
regarding the interpretation of output from the isobaric NearCast model: 1) the model is useful in
identifying areas where the atmosphere is or will become convectively unstable (where cool, dry air
overlays warm, moist air), 2) convection tends to develop within maxima of negative difference, or
convective instability, and along
and convective instability boundaries, 3) convection is most likely to
occur and be strongest in areas that experience the most rapid destabilization tendencies and 4) the
prediction of wind shear may be useful in determining the type and strength of the convection. These
points are consistent with evaluations from GOES-R Proving Ground activities.
However, since the sounding observations are made from the clear sky where flow is mostly
adiabatic, it was hypothesized atmospheric motion could be predicted more accurately in an isentropic
framework with trajectories along constant potential temperature surfaces. A three- dimensional,
isentropic version of the model, therefore, has been developed to more accurately predict the movement
of the observations and to retain more information from them.
The April, 09, 2011, Mapleton, Iowa, tornado followed by heavy precipitation across parts of
Minnesota and Wisconsin showcased the additional information gained from running the model in an
isentropic framework. By more appropriately predicting the movement of these observations in threedimensions along isentropic surfaces, the model can 1) depict adiabatic lift, 2) give more details about
vertical wind shear, 3) more accurately predict the three-dimensional movement of and 4) provide more
accurate predictions of convective instability and destabilization. Additionally, the isentropic model
provides more information from the temperature and moisture observations, including total layer moisture.
This information can be used to identify 1) whether there will be enough moisture to support convective
growth, 2) the longevity of potential convection and convection that has already formed and 3) the
potential for a convective event to produce heavy rainfall.
REFERENCES
Gallagher, T. & Dreeszen, D. (2011) ‘GRACE OF GOD’: Mapleton devastated, but thankful lives spared.
Sioxcityjournal.com.
Li, Z., Li, J., Menzel, W. P., Nelson, J. P., Schmit, T. J., Weisz, E., & Ackerman, S. A. (2009) Forecasting
and nowcasting improvement in cloudy regions with high temporal GOES sounder infrared
radiance measurements. J. Geophys. Res., 114, D09216.
Moore, J. T., (1986) Isentropic analysis and interpretation: operational applications to
synoptic and mesoscale forecast problems. Saint Louis University, Department of Earth and
Atmospheric Sciences.
Oliver, V. J., & Oliver, M. B. (1951) Compendium of meteorology: meteorological analysis in the middle
latitudes. American Meteorological Society, pp 715-727.
Petersen, R. A., Aune, R., Dworak, R., & Line, W. (2012) Using analysis of the information content of
GOES/SEVIRI moisture products to improve very-short-range forecasts of the pre-convective
environment. 2012 EUMETSAT meteorological satellite conference.
Petersen, R. A., & Uccellini, L. (1979) The computation of isentropic atmospheric trajectories using a
discrete model formulation. Mon. Wea. Rev., 107, pp 566-574.