Effect of Water Background and Depth on Red and Red-Edge Reflectance Based Yield Prediction Models in Rice.

Nitrogen (N) management based on remote sensing technology has been studied in cereal crops. In rice flooded growth condition, water background is a unique feature and may require additional investigation prior to implementing remote sensor technology. Collection of normalized difference vegetation index (NDVI) readings for rice grain yield prediction in Louisiana is done in mid-season. At this stage, sensor can capture water background in its field of view. Therefore, the objective of this research was to investigate the effect of water background on rice yield prediction models using red and red-edge spectral reflectance in two varieties. Varieties x N trials were established at the LSU AgCenter Rice Research Station located in Crowley, Louisiana in 2011 and 2012. Canopy spectral reflectance under clear and turbid water, biomass yield, N content, plant canopy coverage, and water depth were collected each week for three consecutive weeks beginning two weeks before panicle differentiation;  at harvest, plot grain yield was determined. There was no significant effect of water background on the spectral reflectance readings at red and red-edge wavebands.  Water depth slightly influenced the reflectance at red waveband however, this effect was not carried over when reflectance readings were transformed to vegetation indices. The red-edge based vegetation indices had a better relationship with grain yield (r²=0.89). The r² and components of yield prediction models changed for each variety depending on the transformation process done on reflectance readings within the red-edge and near infrared bands. However, no significant effect of variety was observed for models which used derivative based red-edge vegetation indices. Our findings showed that:1) water background poses miniimal concern when using remote sensing technology for mid-season N application in rice, and 2) red-edge based vegetation indices showed potential as good predictors of rice grain yield.