Estimation of maize plant height in North China by means of backscattering coefficient and depolarization parameters using Sentinel-1 dual-pol SAR data

Abstract

ABSTRACT SAR parameters have been used many times to characterize crops. However, it was found that the scattering mechanisms derived from quad-pol SAR were different from that of the dual-pol SAR data. The objective of this study is to better understand the scattering mechanism of the maize field to dual-pol (VV, VH) SAR data and propose a new radar vegetation index to improve the inversion accuracy of the crop height. We first analyzed the temporal variation of SAR parameters before the maize height stopped increasing. It showed that the sensitivity of SAR parameters to the plant height was strong in the early stage, but the correlation became weak later. Then, according to the scattering mechanism of the maize field to the dual pol (VV, VH) SAR data, we combined use of the backscattering coefficient of σ0VH and the depolarization parameter proposed the mPRVI index. In this paper, the simple linear regression model was used to fit the six parameters separately to estimation the plant height, including σ0VH (R = 0.84), VIRatio (R = 0.82), RVI (R = 0.83), DOD+RVI (R = 0.85), PRVI (R = 0.86) and mPRVI (R = 0.88), which showed a relatively high Pearson correlation coefficient (R) with the plant height. According to adjusted coefficient of determination R2 (0.70, 0.70, 0.71, 0.74, 0.76, 0.80) and RMSE (62.8 cm, 62.7 cm, 61.8 cm, 57.5 cm, 56.3 cm, 51.2 cm) for σ0VH, VIRatio, RVI, DOD+RVI, PRVI, mPRVI, respectively, the mPRVI index outperforms other parameters throughout the maize growing season. And this outperformance is not only manifested in the early growth stage of the maize, but also in the later stage before the plant height stopped increasing.