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

Tripoli, Gregory J.Trends in Evapo-transpiration in the Great LakesStatesTripoli, Gregory J. 1 ; Kung, Sam 21. Atmospheric and Oceanic Scienc, University of Wisconsin- Madi, Madison, WI, USA2. Soil Science, University of Wisconsin - Madison,Madison, WI, USARecently, an alarming drop in lake levels and water tablelevels in the Central Sands of Wisconsin over the past 12years has raised concerns with regard to the relationshipbetween domestic, agricultural and natural uses of water.Angry residents have argued that high capacity wells beingused by agriculture are draining water supplies, vital to therecreational industry. Others suggest that forests,particularly conifer forests, use as much or more waterannually as agriculture where fields are dormant for much ofthe year. Recent hydrological studies, carried out inWisconsin, used measurement of stream flow andprecipitation over the past 12 years and in a more limitedstudy, over the past century demonstrated that suggestedthat (1) the precipitation over the State has remainedrelatively constant, and (2) stream flow is decreasing statewide, suggesting that evapo-transipration has beenincreasing, particularly in the last 12 years. This is due insome part for an increase in the length of the growingseason and the ice free period over historic periods. Inaddition, these studies have found that the climatologicalseasonal profile of precipitation has been changing from amore flat or constant rate across the warm season, to onedominated by drought with interludes of heavyprecipitation, particularly in the early months and latemonths of the growing season when plant life cannot benefitas much. These studies seem to suggest that the recentchanges in the evapo-transpiration rate over Wisconsin maybe indicative of a general trend affecting the entire GreatLakes Basin, rather than being focused on the Central Sandsof Wisconsin, and are not a result of local excessive wateruse. In fact, the documented changes in Wisconsin, showshorter term variability consistent with the climaticfluctuations of the ENSO cycle, and perhaps consistent alsowith long term climate change. To learn more about theregional footprint of how evapo-transpiration has beenchanging, the authors have proposed a study to NASA to useModis to document the evolution of the growing season overthe past 10-12 years. The results of the completed water cyclebudget studies in Wisconsin and the progress on basin widestudies using satellite will be shown at the oral presentation.http://cup.aos.wisc.edu:/NEWSTsend-Ayush, JavzandulamGenerating a long-term vegetation index datarecord from AVHRR and MODISTsend-Ayush, Javzandulam 1 ; Miura, Tomoaki 11. Natural Resources and Environmental Management,University of Hawaii, Honolulu, HI, USAA long-term global data record of spectral vegetationindex (VI) is considered one of the key datasets for water andenergy, and terrestrial carbon cycle studies. Such a long-termdata record needs to be constructed from multi-sensor data.However, extending a VI data record from one sensor toanother requires ensuring cross-sensor continuity andcompatibility because datasets from different sensors are notexactly the same due to differences in sensor/platformcharacteristics and product generation algorithms. In thisstudy, we evaluated cross-sensor Normalized DifferenceVegetation Index (NDVI) compatibility for translating theLong-term Data Record (LTDR) AVHRR NDVI dataset to aTerra MODIS-compatible dataset by “top-down”, directimage comparisons. There was about one year of anoverlapping observation period for NOAA-14 AVHRR andTerra MODIS in years 2000-2001; however, the LTDRAVHRR product for that period was not generated due tosignificant orbital drift. Thus, we designed our compatibilityanalysis approach to use SPOT-4 VEGETATION (VGT) forbridging the AVHRR to MODIS datasets. Three global daily5-km datasets, LTDR NOAA-14 AVHRR NDVI, TerraMODIS Climate Modeling Grid (CMG) NDVI, and VGT S1NDVI, were obtained for their overlapping periods ofobservations (1998-1999 for AVHRR, 2001-2002 for MODIS,and 1998-1999 and 2001-2002 for VGT). A set of twocomparisons were made for (1) AVHRR vs. VGT and (2) VGTvs. MODIS. Cross-sensor NDVI compatibility was measuredand analyzed using agreement coefficients (ACs), whichcomputes a set statistics for evaluating the degree ofagreement between two datasets, including R2 values,geometric mean functional regression (GMFR), andsystematic and unsystematic differences. Our analysisshowed good linear agreements for both sensor pairs (R2 =0.53 for AVHRR and VGT and R2 = 0.95 for VGT andMODIS). GMFRs indicated that MODIS NDVI was slightlyhigher than that of VGT and that VGT NDVI was largerthan that of AVHRR. Likewise, these systematic differenceswere larger for larger NDVI values for both pairs.Unsystematic differences, which are measured as scatteringabout the mean trend (GMFR line) in AC, were larger for theAVHRR-VGT pair than for the VGT-MODIS pair. Furtherevaluation of systematic and unsystematic differencesindicated that the viewing angle constraint could improvethe predictive power of the regression equations as itincreased consistency (i.e., similar observation geometry) ofpaired observations. This study suggests that compatibilityof the NDVI from AVHRR to MODIS is achievable, but theconsiderations of sensor geometry, aerosols, residual cloudcontaminations, and surface conditions are needed to reduceunsystematic differences or uncertainties in cross-sensortranslation.144