Retrieval of cotton plant water content by UAV-based vegetation supply water index (VSWI)

Knowing plant water content (PWC) is of great significance for precision irrigation of field crop. The aim of this study is to monitor the PWC of cotton non-destructively in situ. A six-band multispectral camera embedded on an unmanned aerial vehicle (UAV) was used to collect images at flowering and boll-forming stages of the cotton. Thirteen vegetation indices (VI) were extracted from the camera. Consequently, all the VIs were fused with canopy temperature mathematically into vegetation supply water indices (VSWI). Unary and multivariate models were used to establish the relationship between VSWIs and the water content of leaf, petiole, stalk as well as bud & boll, respectively. Results indicated significant correlations (P < 0.01) between the VSWI from green index (VSWI_GI) and leaf water content (LWC), and between the VSWI from MERIS terrestrial chlorophyll index by the second infrared band (VSWI_MTCI2) and bud & boll water content (BWC). The correlation coefficients between the stalk water content (SWC) and VSWI_MTCI2 as well as VSWI_DATT2 were both −0.895. The best retrieval model of LWC, SWC, and BWC were the multivariate linear models for the much higher estimation ability. Coefficients of determination for modelling and validation were close to or greater than 0.8, and the root-mean-square errors (RMSE) for validation were less than 0.17, and the relative errors (RE) were less than 18%. The results showed all these models have relatively high accuracy and can provide a new method to efficiently monitor water content in cotton plants.