Estimation and comparison of nitrogen contents model of rice [Oryza sativa] plant using remote sensing
2007
Ryu, C.(Kyoto Univ. (Japan)) | Suguri, M. | Umeda, M. | .
In this research, estimation of nitrogen content models at panicle initiation (40 plots) and heading stages (22 plots) and validation of models at heading stage (18 plots) were investigated using multi and hyperspectral remote sensing, depending on two test fields of different nitrogen fertilizer application. On the test field 1, nitrogen fertilizer was applied with seven nitrogen treatments at both basal dressing and topdressing. On the test field 2, nitrogen fertilizer was applied with seven nitrogen treatments at basal dressing and by variable rate at topdressing depending on the plant growth. By using multispectral images recorded at each vegetation stage, it is possible to create nitrogen content models using GreenNDVI, as 0.908 and 0.775, respectively. By using hyperspectral image recorded at panicle initiation stage, it also possible to create nitrogen content models using multilinear regression (r=0.958, MLR) and partial least square regression (PLS) methods, such as Leverage method (r=0.912) and Full-Cross method (r=0.895). In the case of heading stage, Leverage method (r=0.999) and Full-Cross method (r=0.909) were better than the MLR method (r=0.805). When these models were compared at each vegetation stage, there were similar results as nitrogen content models of GreenNDVI (RMSEC=0.610), MLR (RMSEC=0.718), PLS analysis (RMSEC=0.430, RMSEC=0.470). At heading stage, however, there were different tendency at nitrogen contents models compared with GreenNDVI (RMSEC=1.808, RMSEP=1.468) and MLR (RMSEC=1.697, RMSEP=2.208) with leverage method (RMSEC=0.083, RMSEP=4.344) and Full-Cross method (RMSEC=0.706, RMSEP=2.494) of PLS, which was better at RMSEC, but worse at RMSEP and maximum error.
Mostrar más [+] Menos [-]Palabras clave de AGROVOC
Información bibliográfica
Este registro bibliográfico ha sido proporcionado por Agriculture, Forestry and Fisheries Research Information Technology Center