Location: Hydrology and Remote Sensing LaboratoryTitle: Assessing correlations of satellite-derived evapotranspiration, precipitation and leaf area index anomalies with yields of major Brazilian crops
|ZOLIN, C. - Collaborator|
|SENTELHAS, P. - Universidad De Sao Paulo|
|HAIN, C. - Collaborator|
|SEMMENS, K. - Collaborator|
|YILMAZ, M.T. - Middle East Technical University|
|OTKIN, J. - University Of Wisconsin|
|TETRAULT, R. - Foreign Agricultural Service (FAS, USDA)|
Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/24/2015
Publication Date: 3/1/2016
Citation: Anderson, M.C., Zolin, C., Sentelhas, P., Hain, C., Semmens, K., Yilmaz, M., Gao, F.N., Otkin, J., Tetrault, R. 2016. Assessing correlations of satellite-derived evapotranspiration, precipitation and leaf area index anomalies with yields of major Brazilian crops. Remote Sensing of Environment. 174:82-99.
Interpretive Summary: Global yield monitoring efforts are increasingly utilizing geospatial information about surface moisture and crop conditions provided by satellites. This paper explores the utility of remote sensing maps of evapotranspiration (ET, a measure of crop water use), rainfall, and green biomass (leaf area index, or LAI) as predictors of at-harvest yield for major crops grown in Brazil. The crops investigated include soybeans, corn (first and second growing seasons), and cotton. The analyses shows that anomalies in ET, rainfall and LAI are most strongly correlated with yield estimates over the period 2003-2013 in regions where yields were highly variable - particularly in southern and northeastern Brazil. These areas were impacted by several severe drought events over the past decade, and signals of these droughts were clearly evident in all satellite indicators investigated. Of these, the Evaporative Stress Index, based on anomalies in ET, provided best overall performance in terms of strength and timing of peak correlations with yields. This suggests that remotely sensed reductions in crop water use relative to normal rates at critical points during the crop growing cycle may be an effective signal of impending crop failure.
Technical Abstract: To effectively meet growing global food demands will require a better understanding of factors that are currently limiting agricultural yields and where production can be viably expanded with minimal environmental consequences. Remote sensing can help to inform these analyses, providing valuable spatiotemporal information about yield-limiting moisture conditions and crop response under current climate conditions. In this paper we study correlations for the period 2002-2013 between yield estimates for major crops grown in Brazil and satellite indicators of crop water use or evapotranspiration (as conveyed by the Evaporative Stress Index; ESI), water supply (rainfall from the Tropical Rainfall Mapping Mission; TRMM) and biomass accumulation (leaf area index; LAI from the Moderate Resolution Imaging Spectroradiometer - MODIS). Correlations with yield data reported at both the state and municipality levels were computed as a function of satellite index composite date through the growing season to evaluate strength and timing of peak correlations for each index class.Correlation patterns were in general similar between all indices, both spatially and temporally. Spatial variability in correlation strength was largely driven by variability in yield over the period of record, with strongest correlations found in the south and northeast where severe flash droughts have occurred over the past decade. Peak correlations tended to occur during sensitive crop growth stages. At regional scales using state-level yield, the ESI provided higher correlations for most crops and regions in comparison with TRMM and LAI anomalies. Using finer scale yield data at the municipality level, ESI provided higher peak correlations with soybean yields, while TRMM precipitation had marginally better performance with corn in most states examined. For all crops, ESI peak correlations occurred earlier by 10 days on average over the country in comparison with TRMM precipitation anomalies and 25 days in comparison with MODIS LAI anomalies. These results provide insight into when and where different remote sensing indicators are likely to add significant value to yield forecasts, setting the stage for future work involving assimilation into crop modeling systems.