APPLICATION OF VEGETATION AND NDWI INDICES IN REMOTE SENSING OF SOIL MOISTURE

Abstract

Soil moisture is one of the most important factors affecting plant growth and yield (Zhang ir kt., 2018). However, cheaper and more convenient methods that can replace conventional direct moisture measurements are still being sought. This paper examines the accuracy of the application of optical sensors for soil moisture determination based on indices evaluating plant physiology and water content. The research is carried out by applying the "DJI Inspire 2" drone, the "Sentinel-2" satellite, and the gravimetric method of soil moisture determination. The accuracy of the indices is evaluated by constructing linear regression models for the indices most correlated with soil moisture. The reliability of the indices is also determined, which is done by evaluating the significance of the difference between the moisture values ​​obtained by these indices and the in situ moisture and calculating the RMSE.

The results of the study show that soil moisture is best estimated from the NDWI and GCI indices. NDWI accuracy was 35% from drone data and 40% from satellite data. The accuracy of the GCI for determining soil moisture was 30% for drone and 36% for satellite data. No significant difference was found between moisture values from these indices and in situ measurements. The average deviations of indices from direct measurements ranged from 0.22% to 0.83%. In the conditions of lush vegetation, a higher accuracy of the satellite indices was found: 61% (NDWI) and 88% (GCI), suggesting an opportunity to improve the methodology in the conditions of mature vegetation. Average deviations from soil moisture did not exceed 1.10%.