Prediction of dry direct-seeded rice yields using chlorophyll meter, leaf color chart and GreenSeeker optical sensor in northwestern India

Prediction of potential yields during crop growth season is important for successful agricultural decision-making. The objective of this study was to predict grain yield of dry direct-seeded rice (DDSR) using leaf greenness as measured by chlorophyll meter (SPAD) and leaf color chart (LCC) and using normalized difference vegetation index (NDVI) worked from GreenSeeker optical sensor measurements. Regression analysis was performed to predict rice grain yield at maturity from the LCC, SPAD and NDVI readings recorded from two multi-rate nitrogen level experiments conducted in two consecutive rice seasons. The measurements made at early growth stage could not explain satisfactorily the variations in rice grain yield. Predictions made by the LCC were not reliable. The SPAD meter was superior to the LCC at booting growth stage. The NDVI readings at panicle initiation growth stage exhibited the highest coefficient of determination and explained 63% of the variation in rice grain yield. Yield predictability with SPAD measurements at 70 and 84 DAS, and NDVI readings at 70, 84 and 98 DAS did not improve by introducing the element of cumulative growing degree days (CGDD). However, CGDD based SPAD meter readings at 98 DAS, and LCC readings at 70, 84 and 98 DAS improved the crop yield predictability. The regression models were validated on an independent data set obtained from experiment conducted in the same area. The root mean square error (RMSE) for NDVI and SPAD readings was lower than the LCC readings. On the contrary, adjusting the LCC score with CGDD reduced the RMSE. The study reveals that yield of DDSR can be satisfactorily predicted with in-season measurements of NDVI, SPAD meter and the LCC scores adjusted with CGDD.