info:eu-repo/semantics/article
Early Maize Yield Forecasting from Remotely Sensed Temperature/Vegetation Index Measurements
Fecha
2016-01Registro en:
Holzman, Mauro Ezequiel; Rivas, Raúl Eduardo; Early Maize Yield Forecasting from Remotely Sensed Temperature/Vegetation Index Measurements; Institute of Electrical and Electronics Engineers; IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing; 9; 1; 1-2016; 507-519
1939-1404
CONICET Digital
CONICET
Autor
Holzman, Mauro Ezequiel
Rivas, Raúl Eduardo
Resumen
High and low soil moisture availability is one of the main limiting factors-Affecting crops productivity. Thus, determination of the relationship between them is crucial for food security and support importing-exporting strategies. The aim of this work was to analyze the aptitude of temperature vegetation dryness index (TVDI) to forecast maize yield. MODIS/AQUA enhanced vegetation index and land surface temperature (LST) at 1 km were used to calculate TVDI and maize yield over a large agricultural area of Argentine Pampas. The comparison between TVDI and official yield statistics was carried out to derive regression models in two agro-climatic zones, obtaining linear and quadratic adjustments. The models account for between 73% and 83% of yield variability, with the best prediction in the humid zone. The RMSE values ranged from 14% to 19% of average yield. The bias showed a slightly higher difference between predicted and observed yield data in semi-Arid zone. The models showed aptitude to estimate yield with reasonable accuracy 8-12 weeks before harvest. In addition, the TVDI-maize yield relationship and the impact of submonthly water stress were evaluated at field scale using yield measurements to ensure the analysis on maize. The highest ext{R}^{2} (0.61) was obtained using monthly values suggesting that the entire critical stage should be taken into account for yield forecasting. Although these results would not be directly extrapolated to other agricultural regions in the world, the proposed model is promising for forecasting spatial yield in other regions with poor data coverage several weeks before harvest.