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

a single mission, the phrase “near-real-time” within thiscontext refers to an effective latency of approximately 5 to 15days. For the last few years of the mission, the GRACEscience data system has been producing near real-timeestimates of the Earth gravity field on a best-effort basis,using automated processes. The Level-1B tracking data,produced for monitoring the health of the flight system, isbeing opportunistically used for producing these gravityfield estimates. The short latency, and process automation,implies that the ancillary data and models used in thisprocessing cannot be put to the same scrutiny as theoperational Level-2 gravity field data products. Theseproducts have so far been used for correlative interpretationwith other remote-sensing and in situ data during floods inthe Amazon (spring 2009), Pakistan (fall 2009), and inQueensland, Australia (winter 2010). This paper, after a briefpresentation of the processing approach, will focus on thechallenges imposed on the interpretation of these lowlatencydata products due to processing methods and due tothe choice of background models.Billah, Mirza M.Impacts of Different Evapotranspiration Estimateson Quantify Regional Scale Terrestrial WaterStorageBillah, Mirza M. 1 ; Goodall, Jonathan L. 1 ; Narayan, Ujjwal 2 ;Lakshmi, Venkat 11. Civil and Environmental Engg., University of SouthCarolina, Columbia, SC, USA2. Department of GIS, Richland County, Columbia, SC,USAEvapotranspiration is difficult flux to quantify atregional spatial scales. It is a crucial component of terrestrialwater balance, which is important to estimate wateravailability and sustainable water resources management. Wetest three different approaches for estimatingevapotranspiration and evaluate how well each approachperforms at closing the water budget for sub-watersheds thatrange in size from 1.2 km2 to 3350 km2 in South Carolina.The Variable Infiltration Capacity (VIC) model, NorthAmerican Regional Reanalysis (NARR) program and remotesensing derived estimates are used for evapotranspiration.Results from the analysis show that all the three methods forestimating evapotranspiration produce similar variation inseasonal water storage (positive in fall and winter, negative inspring and summer), but differences exist in the magnitudeand spatial patterns of the estimates. In the spring andsummer months, relatively low evapotranspiration rates wereestimated by remote sensing as compared to VIC and NARRmodels. The remotely sensing evapotranspiration in fall andwinter months fell between the higher VICevapotranspiration and the lower corrected NARRevaporation estimates. We compared our estimates of changein terrestrial water storage using the threeevapotranspiration estimates with drought indices providedby the Drought Monitor (DM) program and observedgroundwater levels as independent means for validating theestimates. On an annual and seasonal basis, the change interrestrial water storage estimated using remote sensingevapotranspiration was consistent with annual and seasonaldrought variation recorded by the DM program.Comparison with groundwater levels showed that remotesensing evapotranspiration approach resulted in the highestcorrelation among the three estimates of evapotranspiration.We conclude from this study that remote sensing is morereliable and consistent at estimating regional scaleevapotranspiration as compared to the two model-basedestimates in our study area.Bitew, Menberu M.Can One Use Streamflow Observations as a Way ofEvaluating Satellite Rainfall Estimates?Bitew, Menberu M. 1 ; Gebremichael, Mekonnen 11. Civil & Environmental Engineering, University ofConnecticut, Storrs, CT, USAObserved streamflow data are increasing used as a wayof evaluating the accuracy of satellite rainfall estimates,particularly in gauged watersheds where there are no reliableground-based rainfall measuring sensors. The procedureconsists of (1) calibrating hydrologic models with satelliterainfall inputs, (2) using satellite rainfall estimates as inputsinto hydrologic model, and (3) comparison of simulated andobserved streamflow. The authors investigated the feasibilityof this approach for two watersheds within the Blue NileRiver Basin in Ethiopia: Gilgel Abay Watershed (Area of1,656 km2, rainfall accounting for 71% of streamflow), andBlue Nile River Basin (Area of 176,000 km2, rainfallaccounting for 20.4% of streamflow). The approach was putto test to evaluate PERSIANN rainfall estimates, which hadlarge underestimation biases as found out throughcomparison with rain gauge values. The approachsuccessfully detects the underestimation bias in the GilgelAbay watershed, but fails to detect it in the Blue Nile Riverbasin. Apparently, in the Gilgel Abay Watershed,precipitation is a significant portion of the streamflow, andany errors committed in the model parameter estimatespertaining to evapotranspiration and groundwater flow(during calibration phase due to lack of these datasets)cannot hide the substantial bias in the rainfall estimates.However, in the Blue Nile River Basin, precipitation is asmall portion of the streamflow, and any errors committedin model parameter estimates can easily hide the substantialbias in the rainfall estimates. The authors recommend thatstrong caution be exercised in using observed streamflow toevaluate the accuracy of satellite rainfall estimates inwatersheds where precipitation is only a small fraction of thestreamflow. In the absence of reliable model parameterestimates pertaining to evapotranspiration and groundwater,the authors recommend against the use of streamflowobservations a way of evaluating satellite rainfall estimates inregions where precipitation is not a large portion of thestreamflow.40