NASA IDS project meeting:
Precipitation and LULC change datasets
Hongjie Xie
The University of Texas at San Antonio
March 19, 2009 at Texas A&M, Corpus Christi
Year 1• Downscaled the NEXRDA MPE 4 km to 1
km provided for Liang’s group for modeling
2004-2007, totally 34,842 hourly data
Guan, Xie, WilsonSubmitted to J of Hydrology
Year 2
• Processed NEXRAD Stage IV (2004-2007)• Examined methods for improving NEXRDA
accuracy• Started to build the SWAT model for Guadalupe
River Bain• Started working on the USGS LULC change
dataset
2.1 NEXRAD Stage IV (2004-2007)
• Stage IV data of 2004-2007 has been downloaded, processed, aggregated into 3 hourly files, and provided for Liang’s group.– All 3 hourly files were provided in the NetCDF
format
22º N to 37º N, and 100º W to 75º W
Missing files only found in 2004
2.2 Improving NEXRDA accuracy
• Explored four methods to improve the NEXRDA precipitation accuracy, by incorporating rain gauge measurements– Bias Adjustment (BA)– Simple Kriging with varying Local Means (SKlm)– Kriging with External Drift (KED)– Regression Kriging (RK)
• To evaluate which method is better, four evaluation parameters are used:– Percentage Bias, – Mean Absolute Error, – Coefficient of Determination, – Nash-Sutcliffe efficiency
NoCorrection BA SKlm KED RK
Mean 7.16 9.32 9.15 9.12 9.08Min 0 0 0 0 0Max 32 41 42 46 44CV 1.33 1.33 1.36 1.38 1.37
PBIAS -22.41 0.04 0.5 0.79 0.43MAE 3.36 2.64 1.95 1.94 1.93
R2 0.87 0.86 0.92 0.92 0.92NSE 0.81 0.86 0.92 0.92 0.92
Hour 9, June 22nd, 2004
Hour 8, April 24nd, 2004
Hour 8, June 22nd, 2004
NoCorrection BA SKlm KED RK
PBIAS 2.93 0.29 2.52 1.68 2.65MAE 0.32 0.02 0.04 0.07 0.04
R2 0.9 1 1 0.99 1NSE 0.9 1 1 0.99 1
PBIAS 4.33 -3.92 0.21 5.55 0.25MAE 0.45 0.4 0.35 0.41 0.35
R2 0.77 0.76 0.76 0.77 0.76NSE 0.79 0.84 0.88 0.83 0.88
PBIAS 6.54 -0.15 1.13 0.52 0.57MAE 0.41 0.4 0.36 0.37 0.36
R2 0.65 0.65 0.65 0.64 0.65NSE 0.65 0.74 0.8 0.73 0.79
PBIAS 36.41 50.12 31.98 28.51 32.24MAE 0.98 0.98 0.84 0.91 0.85
R2 0.8 0.81 0.8 0.81 0.8NSE 0.69 0.73 0.79 0.71 0.79gp1
Evaluation Parameters
Areal mean
kr4
kr18
Evaluation parameters of predicted areal mean precipitation and precipitation at three rain gauges for the year of 2004
No CorrectionBA SKlm KED RK0
10
20
30
40
50
Abs
olut
e P
BIA
S
No CorrectionBA SKlm KED RK
0.4
0.6
0.8
1
MA
E
No CorrectionBA SKlm KED RK0.55
0.6
0.65
0.7
0.75
0.8
R2
No CorrectionBA SKlm KED RK
0.65
0.7
0.75
0.8
0.85
0.9
NS
E
Box-and-whisker plot of the PBIAS, MAE, R2 and NSE values of the 50 rain gauge for four different methods
Results show that the average performance of SKlm is similar to or better than the other three methods. -- A paper is in submission
2.3 SWAT model for Guadalupe River Bain
• Soil and Water Assessment Tool (SWAT)• DEM, precipitation, temperature, land cover/land u
se and soil type as input • Use measured water quality data to tune the para
meters. – nitrate (NO3), ammonium (NH4), dissolved oxygen (D
O), total dissolved phosphorus (TDP), phosphate (PO4), total organic carbon (TOC), total nitrogen (TN) and total phosphorus (TP)
– Data are getting from the Jim’s group (not yet).• Once well-tuned and calibrated, it can be used to
estimate the water quality, which can be used to compare with water quality estimates, from Jim’s empirical modeling approach.
90 m DEM were used to delineate the basin and subbasins: 21 subbasins and 226 HRUs (hydrological response units)
78 stations
Key Procedures
• Create SWAT project• Delineate watershed (DEM, basin shapefile)• Define land use/soil/slope data grids (land cover/
land use, soil type)• Determine the distribution of HRUs • Define rainfall, temperature and other weather d
ata• Write the SWAT input files• Setup and run SWAT
Water quality calibration Using NCDC gauge precipitation as input
• Data Source: – rainfall and temperature data 1990-2007 – Water quality data: 1993 – 2007 (daily, monthly or yearly)
• Model spin-up: 1990-1992• Calibration period: 1993-2007• Calibration parameters:
– nitrate (NO3), ammonium (NH4), dissolved oxygen (DO), total dissolved phosphorus (TDP), phosphate (PO4), total organic carbon (TOC), total nitrogen (TN) and total phosphorus (TP)
2.4 LULC change dataset
• We examined the USGS LULCC dataset– Landsat 1992– Landsat 2001
• A preliminary analysis was done in Bexar county
• For new changes since 2001, we need MODIS yearly 1 km (we can process them if any group need them)
LULC and Change 1992-20011. Open water2. Urban3. Barren4. Forest5. Grassland/Shrub6. Agriculture7. WetlandsNN. Changes
Example of San Antonio/Bexar County
Bexar County
Edwards Aquifer
Thank you
NEXRAD precipitation products
• NCDC Level II, III • Level II (base) data
– Reflectivity, mean radial velocity, and spectrum width
– 1 km x 1 degree– 5 or 6 minutes in rain model
and 10 minutes in clear sky mode
• Level III products (total 41)– DPA (4.7625 km HRAP grid,
hourly, but every 5 or 6 minutes)
• RFC Stage I, II, III• Stage I Stage I - Hourly digital
precipitation (HDP), 4 km• Stage II Stage II - HDP merge with
gauges• Stage III (or MPE)Stage III (or MPE) -
Mosaicked Stage II cover a RFC area. MPE since 2004.
• NCEP Stage IV • Stage IVStage IV – Mosaicked Stage
III or MPE for the USA.