Exploration of high-resolution multispectral UAV imagery for non-destructive estimation of chlorophyll content in lemon lea

Abstract

Unmanned Aerial Vehicle (UAV) technology also known as drones together with Remote Sensing (RS) have become evident as great potential for the agriculture industry in making informed decisions. The overall aim of using this technology is to improve productivity and support sustainable agricultural practices. South African farmers are faced with many challenges, including climate change, load shedding, and water shortages. Despite these challenges, they must sustain the environment and enhance food security. Chlorophyll plays a great role as a crop predictor of plant health status and can reveal nutrient deficiency. The estimation of chlorophyll requires selecting vegetation indices (VI) that are sensitive to chlorophyll or physiological properties of crops. Five indices were selected for this study: the Normalized Difference Vegetation Index (NDVI), Normalized Difference Red Edge (NDRE), Modified Soil Adjusted Vegetation Index (MSAVI), Leaf Chlorophyll Index (LCI), and Difference Vegetation Index (DVI). The study focused on estimating chlorophyll content in lemon leaves using VIs and validating these estimates with ground-measured chlorophyll data. The Partial Least Squares Regression (PLSR) method was employed to create a predictive model for chlorophyll content in the study area. Through Pearson Correlation analysis, it was found that VIs that used the combination of the NIR and red band were more influential and sensitive indices to chlorophyll than the red edge indices. The DVI and MSAVI yielded the highest correlation, with values of 0.85 and 0.47 respectively. These were followed by the LCI, NDVI, and NDRE which had r values of 0.26, 0.23, and 0.22 respectively. Among the spectral bands, the NIR band exhibited the strongest correlation with a coefficient of r = 0.89, followed by the red edge, green, red, and blue bands with their respective r values of 0.57, 0.18, and 0.11. The blue band showed the lowest correlation with measured chlorophyll at r = 0.04. The PLSR estimation models were evaluated using the following metrics: coefficient of determination (R2), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). The model based on UAV spectral bands achieved a slightly higher estimation accuracy with an R2 = 0.85, and the lowest values of RMSE = 2.11 and MAE = 1.79. This model was subsequently used to create a chlorophyll concentration map for the study area. These findings demonstrate that high-resolution UAV data can be effectively utilized for field management in many ways including crop monitoring and field mapping. This data can significantly help in making more informed decisions.