Presentation on theme: "Adapting SEBAL for Ogallala region: Limitations and Improvements in ET mapping George Paul 1, Prasanna H. Gowda 3, P.V. Vara Prasad 1, Terry A. Howell."— Presentation transcript:
Adapting SEBAL for Ogallala region: Limitations and Improvements in ET mapping George Paul 1, Prasanna H. Gowda 3, P.V. Vara Prasad 1, Terry A. Howell 3, Paul D. Colaizzi 3, Stacy L. Hutchinson 2, David Steward 2 1 Agronomy, 2004 Throckmorton Hall, Kansas State University, Manhattan, KS 66506, USA 2 Biological and Agricultural Engineering, 43B Seaton Hall,, Kansas State University, Manhattan, KS 66506, USA 3 USDA-ARS Conservation and Production Research Laboratory, P.O. Drawer 10, Bushland, TX 79012, USA
Penman, H.L. Natural evaporation from open water, bare soil and grass. Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences Vol. 193, No. 1032, Apr, 22, 1948, pg. 120-145 Background
Motivation Different ET measurement method in field (typical errors) 1.Eddy Covariance (15% ̶ 30%) 2.Bowen Ratio (10% ̶ 20%) 3.Soil Water Balance (10 % ̶ 30%) 4.Lysimeter (5 % ̶ 15%) 5.Remote Sensing EB (10 % ̶ 20%) Why do we need Remote Sensing-ET (ET a ) ? 1.Directly gives crop water demand (ET a ) 2.Spatial coverage 3.Large utility (apart from agriculture) 4.Based on physical parameterization 5.Accuracy as good as any other measurement 6.Water-Energy-Carbon Nexus Field Lysimeter (mmh -1 ) ET cadj =(K s K cb +K e ) Et o (mmh -1 ) RS_ET (mmh -1 ) NE 0.760.720.77 SE 0.850.720.79 NW 0.520.480.42 SW 0.470.480.46 Do we need RS-ET for crop water requirement assessment (ET a ) ? Kc values unreliable, too much idealized, cannot cope with the newer hybrids.
Motivation AGRONOMY JOURNAL, VOL. 66, MAY-JUNE 1974 Landsat Data Continuity Mission/ Landsat 8, 11 Feb 2013
Research Gaps SEBAL SEBS TSM METRIC MOD16 Where are we heading to ? ReSET SEBTA SSEB REEM AHAS SEBI SW Model
Goals and Objectives The primary goal of our research is to comprehensively evaluate and improve Remote Sensing based ET mapping techniques for developing an operational water resource management and planning tool. In this study we are looking specifically on SEBAL Model
Materials and Methods Bushland Evapotranspiration and Agricultural Remote Sensing Experiment (BEAREX) 1.Conducted at the USDA-ARS Conservation Production Research Laboratory, Bushland, TX, during the 2007 and 2008 summer cropping season. 2.Multi-institutional (Four USDA-ARS Labs, Univ. Texas, Utah State Univ., and Kansas State Univ.) research effort initiated in 2007 and continued through 2008. 3.Consisted of 3 large aperture scintillometers, 4 weighing lysimeters, 2 eddy covariance stations and 2 Bowen Ratio stations 4.Included MODIS, LANDSAT, and ASTER satellite data, and airborne multispectral digital data
BEAREX :Uniqueness Materials and Methods 1.Simultaneous evaluation of dryland and irrigated conditions 2.Use of high resolution (0.5-3 m) airborne images 3.Multiple images acquired from emergence to vegetative growth period from two years for tall and short crop. 4.Evaluating instantaneous ET (mm h -1 ) values 5.Evaluating against large precision Lysimeter
No.Year Acquisition Date (DOY) Time Lysimeter Field NE (Irrigated) SE (Irrigated) NW (Dryland) SW (Dryland) 1 2007 June,24 (175)10:20 Forage Sorghum May 30 Forage Corn May 17 Grain Sorghum June 6 Clumped Grain Sorghum June 6 2June,25 (176)11:33 3July,02 (183)03:27 4July,10 (191)09:53 5July,10 (191)11:15 6July,10 (191)02:50 7July,11 (192)12:40 8July,26 (207)11:37 9July,27 (208)09:55 10July,27 (208)11:16 11July,27 (208)01:33 12 2008 June,26 (178)10:52 Cotton May 21 Cotton May 21 Cotton June 05 Cotton June 05 13July,12 (194)11:20 14July,20 (202)11:06 15 July,28 (210)11:24 16 Aug,05 (218)11:43 17 Aug, 13 (226)11:25 Materials and Methods Image acquisition
Materials and Methods Canopy Cover and Water Availability A1–6 June, 07’ Dryland field Grain Sorghum A2–6 June, 07’ Dryland field C. G. Sorghum B1– 30 May, 07’ Irrigated field Forage Sorghum B2– 17 May, 07’ Irrigated field Forage Corn
Materials and Methods Canopy Cover and Water Availability A1–26 June, 08’ irrigated field A2–26 June, 08’ dryland field B1– 5 August, 08’ irrigated field B2– 5 August, 08’ dryland field.
Materials and Methods Statistical variable DescriptionUse and desired value n Number of observations - R2R2 Coefficient of determination Degree of collinearity +1or -1 m Slope of the best fit regression line Relative relationship ~1 y-intercept y-intercept of the best fit regression line Lag or lead indicator ~0 MBE Mean bias error underestimation/ overestimation indication ~0 RMSE Root mean square error Indicates error in the constituents unit ~0 MAE Mean Absolute error individual absolute differences are weighted equally ~0 NSE Nash-Sutcliffe efficiency Indicative of the strength of model to predict the observed ~0-1 Performance statistics
Theories and Concepts Sensible Heat: The bulk formulation kB -1
The Principle: HOT AND COLD PIXEL CONCEPT Theories and Concepts The dT formulation in SEBAL and METRIC for a July 28, 2008 image
ParameterN Obs. Mean Est. Mean MBE%MBERMSE%RMSEMAEMAPDNSER2R2 T s ( o C)6833.533.60.030.071.123.340.792.360.97 R n (W m -2 )625775715.61.028.85.023.04.00.860.87 G o (W m -2 )6838.731.96.717.416.943.813.033.50.210.34 Results Performance statistics for T s, R n, and G o
Results SEBAL (kB -1 =2.3) Observation points NMeanMBE%MBERMSE%RMSEMAEMAPDNSE All fields 480.44-0.09-16.50.1732.30.1426.50.51 Table: Performance statistics for instantaneous ET (mm h -1 ) for the complete data set (Obs. mean 0.52) Figure: Modeled versus observed ET
Results Observation pointsMeanMBE%MBERMSE%RMSEMAEMAPDNSE All fields 0.44-0.09-16.50.1732.30.1426.50.51 Irrigated fields0.57-0.06-10.10.13188.8.131.52.69 Dryland fields0.30-0.11-26.40.1948.70.1741.6-0.18 Irrigated Field Dryland Field Table: Performance statistics for instantaneous ET (mm h -1 ) for the irrigated field (Obs. mean 0.64) and dryland field (Obs. mean 0.41)
Results SEBAL (z 1 = 0.1) Observation pointsnMeanMBE%MBERMSE%RMSEMAEMAPDNSE SEBAL (kB -1 =2.3) 480.44-0.09-16.50.1732.30.1426.50.51 SEBAL (z 1 =0.1)480.48-0.04-7.30.1427.10.1222.60.66 Table: Performance statistics for instantaneous ET (mm h -1 ) for the complete data set (Obs. mean 0.52) Figure: Modeled versus observed ET SEBAL (kB-1=2.3)
SEBS-SEBAL Hybrid Algorithm Results 1. Linear temperature gradient approach from SEBAL. dT versus T s linearity assumption could not address the spatial variation of z oh (kB -1 ) or in other words address the differences between T o and T s 2. Excess resistance to heat transport from SEBS. Under both sparse and full vegetation conditions, an appropriate value of kB -1 is required for accurate estimation of H using T s Observation pointsMBE%MBERMSE%RMSEMAEMAPDNSE All fields 0.000.330.1019.20.0916.40.73 Irrigated fields0.000.640.0913.60.0711.010.80 Dryland fields0.000.160.1227.40.1125.50.31
Selection of hot and wet pixel and the variability in 'a' and 'b' coefficient (subset image ).
Summary 1.Uncertainty exists in the selection of ‘Hot’ and ‘Cold’ pixel. The expected variation in the final ET value due to different selection criteria may ranges from 10% to 30%. 2.SEBAL performance for irrigated fields (greater ET rates, limited soil water deficits, and complete ground cover) and dryland fields (lower ET rates, greater soil water deficits, and sparse ground cover) were markedly different. 3.SEBAL is sensitive to the value of z oh or kB -1. The approach of evading z oh by adopting constant z 1 appeared to be a good option under the greater ET rates, limited soil water deficits, and greater ground cover; however, under sparse ground covers this would completely fail. 4.A kB -1 model incorporated into SEBAL performed better for both dryland and irrigated fields
Acknowledgements Colleagues at Crop Physiology Lab, Agronomy, KSU. Scientists at USDA-ARS-CPRL, Bushland, Texas. We are grateful to Dr. Christopher M.U. Neale, Professor, Civil and Environmental Engineering, Utah State University for conducting the airborne missions for acquiring the remote sensed imagery.