Remote Sensing of Wheat Crop Health in Punjab, Pakistan: Utilizing Sentinel Data and Vegetation Indices

Abstract

The goal of the study is to ascertain the age of wheat crops in Punjab, Pakistan, by utilizing contemporary remote sensors in conjunction with machine learning and estimation methods. A range of vegetation indices, such as NDVI, EVI, SAVI, GDVI, and IVI using Sentinel satellite data, were used to track crop health and crop growth stage. To identify the most relevant features, two distinct selection approaches were applied: Univariate Linear Regression Tests and the Random Forest Feature Importance method. Polynomial regression models of degrees 1 to 3 were then used. The results demonstrated highly precise crop age estimation, reflected by strong R² values, which are 0.68- 0.92, and low Root Mean Square Error (RMSE), 1.14 to 0.5, is observed among polynomial degree 1 to 3, respectively. These findings are very important in making wise judgments on matters that are to be done in respect to water supply and nutrient management. The article draws attention to the power of data-driven by remote sensors and strong feature modeling and selection methods to improve precision agriculture. The approach will help in promoting sustainable agricultural activities because farmers will be able to take timely measures that can enhance productivity and resilience to environmental stresses in the area. The solution is an important development of present-day agricultural monitoring and management in Punjab and similar agro-ecological areas.

Code and data availability

The paper uses Sentinel-2 imagery from USGS Earth Explorer and field measurements, but provides no public dataset, code, model, or supplement. The Data Availability Statement only says the testing data is in the paper itself, with no authors' URL or repository identifier.

No evidence-backed public reproduction asset is currently recorded.