Abstract:
The decline in the availability of farm products has had a significant negative impact on society, driven by reduced crop production and an increasing gap between global food demand and agricultural output. Crop production faces numerous challenges, such as water scarcity, poor soil quality, unsuitable temperatures, and attacks by pests, diseases, and weeds. Conventional ground-based methods for detecting and controlling pest invasions are time-consuming, labor-intensive, and not well-suited for large-scale farmlands. In contrast, remote sensing offers a rapid, non-intrusive, and viable option for collecting and analyzing spectral properties of Earth’s surfaces from various distances, including satellites and ground-based platforms.
This study aims to assess the effectiveness of remote sensing instrumentation and techniques in evaluating and estimating crop health, with the goal of circumventing the lengthy, costly, and labor-intensive biological laboratory tests typically used by agricultural scientists for this purpose. Sentinel-2A and Landsat 8 images covering an area of 101 hectares were acquired for the study and preprocessed using ArcGIS software to remove atmospheric effects on image reflectance properties. These images were then processed to derive various representations of vegetation indices, soil indices, and tasseled cap indices.
Statistical tools such as correlation, regression, and analysis of variance (ANOVA) were employed to evaluate the agreement between vegetation and soil indices and their correlation with tasseled cap indices. Concurrently, laboratory tests were conducted on sampled crops to assess their health status. A Principal Component Analysis (PCA) model was developed to convert the laboratory test results into an equivalence of the remote sensing NDVI (Normalized Difference Vegetation Index), a widely used index for assessing crop health.
The statistical analysis revealed a very weak correlation (0.16) between the outputs from the remote sensing approach and the conventional agricultural approach to crop health estimation. The overall likelihood that the remote sensing technique would yield results equivalent to laboratory tests in agriculture was approximately negligible (1.6%). However, individual crop species exhibited varying levels of similarity. For instance, cassava showed a 48.8% similarity, groundnut had 50.2%, maize had 63.8%, and rice had 23.9% similarity with laboratory results. The study indicated that the remote sensing technique is unreliable for estimating maize health. Furthermore, when employing satellite images for estimating soya beans health, the output showed a negative correlation with laboratory results. In the case of yam species, there was no significant correlation (0.021) between the laboratory and remote sensing results, suggesting that cultivation efforts should be focused on the northern region of the study area.
To establish a more valid PCA model for the study variables, the results obtained from this study should be further validated.
Utilizing Satellite-Based Remote Sensing for Crop Health Estimation and Monitoring. GET MORE, ACTUARIAL SCIENCE PROJECT TOPICS AND MATERIALS
