Monitoring Water Status Based On Hyperspectra In Wheat

Real-time and non-destructive monitoring of crop water status based on hyperspectra is of significant importance for improving crop water management level and use efficiency. As one of the limiting nutrition factors of crop, nitrogen impacts on plant morphogenesis and growth and development directly, thereby affects agronomic parameters and spectral reflectance, furthermore, interferes crop water monitoring. This study was based on experiments on wheat plants under different water and nitrogen treatments in two consecutive growing seasons, obtained information about leaf and canopy hyperspectral reflectance and the corresponding agronomic parameters in critical growing stage, and then analyzed impact of different nitrogen nutrition on water parameters and spectral reflectance, determined the core band and optimal spectral indices at different scale, established stable and reliable water monitoring models in wheat.Firstly, choosing leaf equivalent water thickness (EWT) as the indicator of crop water status at leaf scale. Systematically constructed normalized difference spectral indices (NDSI), ratio spectral indices (RSI), deferent spectral index (DSI) and three-band spectral indices with the original spectrum in the range of350-2500nm, and the optimal spectral indices were selected for models of EWT monitoring in leaf of wheat. The results showed that NDSI (1204,1314) could be used for EWT monitoring during the entire growing season of wheat, but the model performed differently before and after anthesis. Therefore, two spectral indices (NDSIb at1445nm,487nm and NDSIa at1714nm,1395nm) were constructed for the two different growth periods. The attempt to construct three-band spectral indices based on NDSI (1429,416) found in the common sensitive areas before and after anthesis confirmed that the EWT monitoring models based on NDSIb and NDSIa displayed high precisions and low errors.Secondly, obtained information about canopy water content (CWC) and canopy spectral reflectance in critical growing stage at canopy scale. Analyzed changes of CWC and canopy spectral reflectance and their correlation, constructed two-band spectral indices systematically with all possible two-band combinations by original spectrum in the range of350-2500nm. The results showed that NDSI(762,1458), RSI(1458,1156) and DSI(1458,2301) displayed high modeling precision (R2>0.84) but ordinary testing performance, which were the optimal band combinations, and the CWC model based on DSI(1458,2301) was less stable between years. Furthermore, constructed new three-band spectral indices based on NDSI(762,1458) and RSI(1458,1156) by adding a new third band. It found that (R762-R1458)/(R762+R1458-R2301) improved prediction accuracy while reduced systematic errors caused by different nitrogen rates between two years than two-band spectral indices (with R2of0.8661and SE of2.9512for calibration, and R2of0.6421, RRMSE of0.0437, Slope of1.2886for validation), which new third band2301nm was sensitive to protein and nitrogen.Finally, did research at plant scale. Analyzed PWC and canopy spectral reflectance and their correlation. Then constructed spectral indices systematically with all possible two-band combinations in the range of350-2500nm, selected the optimal spectral indices constructed in three different ways and established PWC monitoring models. The results indicated that NDSI(1175,1305) selected in common sensitive area before and after anthesis, and NDSI(1302,1190) selected in common sensitive area in different PNC levels, performed better than the whole growth stage spectral index NDSI(1727,1539), especially NDSI(1302,1190), which had the best linear relationship on calibration and validation with PWC and insensitive to PNC. The new ways constructed spectral index insensitive to nitrogen and the results in our study would help to provide technical support for water status monitoring and management in wheat.