Premium
In‐Season Canopy Reflectance‐Based Estimation of Rice Yield Response to Nitrogen
Author(s) -
Tubaña B. S.,
Harrell D. L.,
Walker T.,
Teboh J.,
Lofton J.,
Kanke Y.
Publication year - 2012
Publication title -
agronomy journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.752
H-Index - 131
eISSN - 1435-0645
pISSN - 0002-1962
DOI - 10.2134/agronj2012.0214
Subject(s) - normalized difference vegetation index , mathematics , yield (engineering) , oryza sativa , panicle , growing season , canopy , agronomy , coefficient of variation , zoology , coefficient of determination , leaf area index , statistics , botany , biology , biochemistry , materials science , gene , metallurgy
A sensor‐based approach to derive N recommendation based on crop needs and seasonal variation in yield potential and plant‐available soil N requires an estimate of in‐season rice ( Oryza sativa L.) response to applied N. The objectives of this study were to: (i) determine if normalized difference vegetation index (NDVI) readings can be used to predict rice grain yield response to applied N and (ii) establish the relationship between NDVI‐estimated response index (RI‐NDVI) and measured response index at harvest (RI‐YIELD). Sensor and yield data were collected from multiple N‐rate trials from different locations in Louisiana and Mississippi from 2008 to 2010. The NDVI readings were collected using a ground‐based active sensor at panicle initiation (PI), one (PI + 1 wk) and 2 wk (PI + 2 wk) after PI. Two sets of RI values were computed: (i) the ratio of the highest yielding N‐fertilized plot to the check plot and (ii) ratio of N fertilized plots to the check plot. Across sampling dates, significant linear relationships were obtained between RI‐NDVI and RI‐YIELD ( P < 0.01). The highest coefficient of determination value ( r 2 ) obtained between estimated and measured RIs was obtained at PI + 1 wk ( r 2 = 0.63). The quality of the models was improved when RI values were computed by comparing all variable rate N‐fertilized plots to the check plot. Our results showed that RI‐YIELD can be predicted using RI‐NDVI measured within a 3‐wk period starting at PI which is in synchrony with the schedule of topdress N application in mid‐southern United States rice production systems.