Satellite-Based Estimation of Evapotranspiration in Florida David M. Sumner 1 ,Jennifer M. Jacobs 2 , John R. Mecikalski 3 , and Michael Holmes 4 1 U. S. Geological Survey, Florida Integrated Science Center, Orlando, Florida 2 University of New Hampshire, Department of Civil Engineering, Durham, New Hampshire 3 University of Alabama in Huntsville, Atmospheric Sciences Department, Huntsville, Alabama 4 U. S. Geological Survey, Florida Integrated Science Center, Tampa, Florida
57
Embed
Satellite-Based Estimation of Evapotranspiration in Florida David M. Sumner 1,Jennifer M. Jacobs 2, John R. Mecikalski 3, and Michael Holmes 4 1 U. S.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Satellite-Based Estimation of Evapotranspiration in Florida
David M. Sumner1,Jennifer M. Jacobs2, John R. Mecikalski3, and Michael Holmes4
1U. S. Geological Survey, Florida Integrated Science Center, Orlando, Florida
2University of New Hampshire, Department of Civil Engineering, Durham, New Hampshire3University of Alabama in Huntsville, Atmospheric Sciences Department, Huntsville, Alabama
4U. S. Geological Survey, Florida Integrated Science Center, Tampa, Florida
Varieties of ET
• Actual ET
• Reference ET- hypothetical surface ET
• Potential ET - ET when moisture is not limiting
- surface-dependent
Problem & Need
• Potential ET is common input for hydrologic models (surface dependent)
• Reference ET is needed for allocation of water (for real or hypothetical “reference” surface)
• PET and RET are inconsistently determined among the five Florida Water Management Districts
• Areally-continuous coverage of both PET and RET is lacking
• Potential ET = ET without water limitation
Actual ET = parameterized function of:PET, water level, soil moisture, and/or LAI
MODFLOW, MikeShe, HSPF, VS2D, etc.
MODFLOW ET conceptualization
Reference ET w/crop coefficients
• Reference ET computed using:- - weather station data- - selected RET equation
(varieties of Penman-Monteith, Blaney-Criddle, Hargreaves, etc.)
and crop coefficent specific to crop type and phase is applied as multiplier
AET = kcRET
Crop CoefficientsCrop Coefficients
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Original BC Shih Modified BC
Problem & Need
• Potential ET is common input for hydrologic models (surface dependent)
• Reference ET is needed for allocation of water (for real or hypothetical “reference” surface)
• PET and RET are inconsistently determined among five Florida Water Management Districts
• Areally-continuous coverage of both PET and RET is lacking
Objectives
Estimate reference and potential ET - - throughout State of Florida - - 1995 to 2004 … and beyond - - at 2 km spatial resolution - - at daily temporal resolution - - with spatial grid consistent with NEXRAD grid
RET computations
Meteorological data
Inverse distance weighting interpolation of numerous NOAA, UF, and WMD weather station data to 2-km grid.
UF and WMDNOAA
• Simple method
• Priestley-Taylor
• Penman-Monteith
PET models considered
slow RKET *
GRET now
as
apanow rr
rDcGRET
1
Comparison of PET models with AET measured during low Bowen ratio conditions
….. Supports choice of Priestley-Taylor PET
SFWMD / USGS ET station at WRWX in Polk County
Bowen ratio ET station in Everglades
Calculation of PET was performed with the Priestley-Taylor method
E = (Rn – G)
PET computations (daily)
Solar radiation measured via satellite …
… other variables estimated using spatial interpolation of land-based station data.
Required input = net radiation (Rn) = 1.26 = f(air temperature)G is assumed zero over a day
Solar and terrerestrial radiation
Net Radiation
Incoming solar (Rs)
Reflected solar = Rs
Longwave down (Ld)
Longwave up (Lu)
Net radiation = Rs – Rs +Ld - Lu
4-component radiation sensors (11) used to define means to estimate reflected solar and longwave terms
Longwave radiation simulation
Stefan-Boltzmann equationRadiation = T4
surface ~ 0.97 for soil/grass/snow
atmosphere =
f (vapor, temperature, cloudiness)
Clear sky clear (e) – Sellers (1965)
Cloudy sky (clear ,Rs/Ro) - Crawford and Duchon (1999)
Longwave radiation
Longwave radiation simulation
Stefan-Boltzmann equationRadiation = T4
surface ~ 0.97 for soil/grass/snow
atmosphere =
f (vapor, temperature, cloudiness)
Clear sky clear (e) – Sellers (1965)
Cloudy sky (clear ,Rs/Ro) - Crawford and Duchon (1999)
Satellite-based estimation of incoming solar radiation
Inco
min
g so
lar
radi
atio
n (
MJ/
m2/d
)R
efe
renc
e or
pot
entia
l ET
(m
m/d
)
Incoming solar radiation has strong explanatory value (> 80%) fortemporal variability of PET and RET in Florida
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
Longwave terms are large … but solar terms exhibit most variability
0
100
200
300
400
500
600
0 50 100 150 200 250
LW up
LW down
solar down
solar up
Day of year 2008
Rad
iatio
n, in
W/m
2
Rad
iatio
n, in
W/m
2Solar radiation terms explain most (~ 84% at central Florida station) of temporal variability in net radiation
Day of year 2008
-200
-100
0
100
200
300
400
0 50 100 150 200 250
net solar
net longwave
net radiation
Frequency (%) of clear sky conditionsFrom: Climatic Atlas of Clouds over Land and Ocean by Warren and Hahn (2007)
Solar radiation is strongly affected by cloud cover … and Florida is relatively cloudy.
Large spatial variability in cloud cover --->
Large spatial variability in incoming solar radiation and ET
GOES East 8 and 12 geostationary satellites provide spatially (1-km in Florida) and temporally (30 minute) data,
capturing spatial and diurnal changes in cloud cover and solar radiation
Polar-orbiting satellites (MODIS, AVHRR, Landsat) provide less frequent monitoring.
Gautier-Diak-Masse model – simple radiative transfer model incorporating:
1. Clouds
2. water vapor absorption
3. Raleigh/Mie scattering
4. ozone absorption
Gautier et al. (1980)
Diak and Gautier (1983)
Diak et al. (1996)
1. 2-week minimim noon albedo
2. Is pixel cloudy?
3. If so, solve for cloud albedo.
4. Solve for incident solar radiation (full SW bandwidth)
Approach
GOES albedo =
solar albedo
GOES “visible” bandpass
Albe
do
AMS Agriculture and Forest Meteorology Conf.
Orlando, Florida 30 April 2008
GDM model has shown regional utilityOtkin et al. (2005) :
Method has error on theorder of ~7-8% during clear-skyconditions, and ~17% duringcloudy-sky conditions.
Calibration of GDM incoming solar radiation product for Florida
1) Clear-sky conditions
2) Cloudiness bias correction
3) Temporal bias correction
Uncorrected
Clear-Day Comparison Initial Model Calibration
simon
* John, this example is for 3 stations from one WMD (South West Florida, SWF)* Image hanging off bottom displays ok
CorrectedCorrected +4%
Clear-Day Comparison Initial Model Calibration
Calibration of GOES daily solar product under cloudy conditions
19 pyranometer stations~ 36,000 station-days over 1995-2004
GOES daily solar bias related to cloudiness
Mean solar radiation ~ 190 w/m^2
… winter solstice ~ 100 w/m^2
… summer solstice ~ 270 w/m^2
Temporal bias in initial solar product
Measured vs. GOES insolation
Orange = southBlue = north
simon
John, this a comparison with one South Florida station (ENR308) of the DAILY product. I guess if we compared the half-hourly data, things may not look as good, but apparently it's the daily product that the WMDs are interested in anyway.
Error statistics of 9 “validation” stations during calibration path
Initial Clear-sky Cloudiness Temporal trend
Examples of incoming solar radiation daily product
Summer Winter
Mean annual solar cross-sections across Florida
Data availability
10 year dataset (1995-2004) and summary report available on webFISC Hydrologic Data Portal ….. http://hdwp.er.usgs.gov/
Daily 2-km values throughout Florida:
Potential ETReference ETIncoming solar radiationAir temperature (min and max)Relative humidity (min and max)Wind speed (mean)
Work to extend dataset thru 2007 will be completed in September 2008
Applications:Applications:
1.1. PET and RET for hydrological modeling (surface & ground water) PET and RET for hydrological modeling (surface & ground water)
2.2. RET for water allocation RET for water allocation
3.3. Solar radiation distributions for studies of:Solar radiation distributions for studies of:ecosystem (e.g., implications for Primary Production)ecosystem (e.g., implications for Primary Production)oceans (e.g., light availability in subsurface)oceans (e.g., light availability in subsurface)
4.4. Monitoring for climate change (cloud cover and solar radiation)Monitoring for climate change (cloud cover and solar radiation)
Note: Florida is flat …. expansion to more topographically rugged areas Note: Florida is flat …. expansion to more topographically rugged areas will require consideration of slope and aspect in computation of will require consideration of slope and aspect in computation of incident solar radiation ....incident solar radiation ....……. including differentiation of diffuse and direct solar radiation. including differentiation of diffuse and direct solar radiation
Extend approach Nationally?
1. Puts hydrologic modelers on common PET framework.