Evaluation of Landsat 9 and Sentinel 2 satellites in estimation of cotton leaf area index using M5 tree model

Abstract

The availability of satellite data has caused a continuous increase in the accuracy of spatial information and provides significant conditions for monitoring cultivation in large areas. LAI is used as an important parameter using remote sensing techniques for simple crop growth modeling. In this article, estimation of LAI, as a common vegetation cover index, through the collection and field measurements of LAI, and comparison with the multispectral data of Landsat 9 operational terrain imager (OLI) and Sentinel 2 multispectral instrument were done with the minimum percentage of clouds to extract LAI. The relationship between plant indices as an independent variable and field LAI as a dependent variable was searched using linear multivariate regression and M5 tree regression methods. LAI calculated with plant indices is not very accurate and needs to be modeled and recalculated using spectral indices. Considering the non-linearity of the relationship between LAI and spectral reflectance, linear multivariate regression showed almost satisfactory results, and in the best conditions, this relationship has a correlation coefficient of 75.46 and 72.91, with an error of 0.229 and 0.308, respectively, for the Landsat 9 and Sentinel 2. LAI estimation using machine learning techniques is suitable and very capable for observing LAI developments and increases the accuracy of calculations.