2012 AGU Chapman Conference on Remote Sensing of the ...

Gebregiorgis, Abebe S.Characterizing Satellite Rainfall Errors based onLand Use and Land Cover and Tracing Error Sourcein Hydrologic Model SimulationGebregiorgis, Abebe S. 1 ; Peters-Lidard, Christa D. 2 ; Tian,Yudong 2, 3 ; Hossain, Faisal 11. Civil and Environmental Eng’g, Tennessee TechnologicalUniversity, Cookeville, TN, USA2. Hydrological Science Branch, NASA Goddard SpaceFlight Center, Greenbelt, MD, USA3. Goddard Earth Sciences and Technology Center,University of Maryland Baltimore County, Greenbelt,MD, USAHydrologic modeling has benefited from operationalproduction of high resolution satellite rainfall products. Theglobal coverage, near-real time availability, spatial andtemporal sampling resolutions have advanced theapplication of physically based semi-distributed anddistributed hydrologic models for wide range ofenvironmental decision making processes. Despite thissuccess, the existence of uncertainties inherent in theindirect way of satellite rainfall estimation and hydrologicmodels pose a challenge in making meaningful and practicalpredictions. This study comprises breaking down of totalsatellite rainfall error into three independent components(hit bias, missed precipitation and false alarm),characterizing them as function of land use and land cover(LULC), and tracing back the source of simulated soilmoisture and runoff error in physically based distributedhydrologic model. Here, we asked “on what way the threeindependent total bias components, hit bias, missed, andfalse precipitation, affect the estimation of soil moisture andrunoff in a physically based hydrologic model?” Weimplemented a systematic approach by characterizing anddecomposing the total satellite rainfall error as a function ofland use and land cover of the Mississippi basin. Thisfacilitated the understanding of the major source of soilmoisture and runoff errors in hydrologic model simulationand tracing back of the information to algorithmdevelopment and sensor type. Consequently, we believe sucha forensic approach stands to improve algorithmdevelopment, application and data assimilation scheme forGlobal Precipitation Measurement (GPM) mission. Keywords: total bias, hit bias, missed precipitation, false alarm,soil moisture error, and runoff error, land use and land cover66Gebremichael, MekonnenEstimation of Daily Evapotranspiration over Africausing MODIS/Terra and SEVIRI/MSG dataGebremichael, Mekonnen 1 ; Sun, Zhigang 21. Civil and Env. Engineering, University of Connecticut,Storrs, CT, USA2. National Institute of Environmental Studies, Tsukuba,JapanMost existing remote sensing-based evapotranspiration(ET) algorithms rely exclusively on polar-orbiting satelliteswith thermal infrared sensors, and therefore the resulting ETvalues represent only “instantaneous or snapshot” values.However, daily ET is more meaningful and useful inapplications. In this study, daily ET estimates are obtainedby combining data from the MODIS sensor aboard thepolar-orbiting Terra satellite and the SEVIRI sensor aboardthe geostationary-orbiting MSG satellite. The procedureconsists of estimating the instantaneous evaporative fraction(EF) based on the MODIS/Terra land data products, andestimating the daily net radiation and daily available energybased on the 30-min SEVIRI/MSG data products. Assumingconstant EF during the daytime, daily ET is estimated as theproduct of the SEVIRI/MSG-based daily available energy andMODIS/Terra-based instantaneous EF. The daily ETestimates are evaluated against flux tower measurements atfour validation sites in Africa. Results indicate that thesynergistic use of SEVIRI/MSG and MODIS/Terra has thepotential to provide reliable estimates of daily ET during wetperiods when daily ET exceeds 1 mm/day. The satellite-baseddaily ET estimates however tend to underestimate ET by 13%to 35%. The daily ET estimation algorithm can further beimproved by incorporating a temporal data-fillinginterpolation technique to estimate the unavailable netradiation information during cloudy sky conditions, and byimproving the accuracy of the instantaneous EF. Theassumption of constant evaporative fraction during the dayis reasonable, and does not result in substantial errors in thedaily ET estimates.Gebremichael, MekonnenError Model and Tuning Extremes for SatelliteRainfall EstimatesGebremichael, Mekonnen 11. Civil and Env. Engineering, University of Connecticut,Storrs, CT, USAA new model is developed that generates thedistribution of actual rainfall values for any given satelliterainfall estimate. The model handles the conditionaldistribution as the mixture of a positive continuousdistribution and a point mass at zero. A method is alsopresented to probabilistically transform a set of satelliterainfall estimates into the more accurate rainfall estimates.As our concern lies with extreme precipitation, a peaks overthreshold extreme value approach is adopted that fits aPareto distribution to the large precipitation estimates. Asimple distributional transformation result is then used to