Remote sensing of maize intercropping: A multi- sensor approach

Cropping patterns, particularly intercropping - cultivating two or more crops within the same field, are prevalent in smallholder farming systems, enabling farmers to optimize land use, improve soil fertility, and increase resilience to environmental pressures such as pests, diseases, and climate extremes. When implemented effectively, intercropping promotes biodiversity and stabilizes yields. However, the diversity of crops and their interactions in intercropped fields make monitoring crop growth and estimating crop yields challenging. The dynamic and diverse nature of these practices often goes unreported in agricultural statistics, making them even more challenging to identify and document accurately.
Traditional methods, such as field surveys and manual crop assessments, often lack the precision needed to capture detailed variations in crop types, growth and health within these cropping patterns. Further, the capturing methods can be time-consuming, labor-intensive, and prone to human-error and prone to missing important aspects of crop growth. While remote sensing offer a more efficient way to map and monitor crops, the mapping of cropping patterns is often generalized. The fields are often classified as broad cropland categories, or identified only by the most dominant crop in the field. This oversimplification result in the loss of critical information regarding the location of various types of cropping patterns, as a result, key data needed for improving yields and managing resources effectively are often overlooked. Therefore, using both remote sensing data and field data to explore and understand the dynamics of these cropping patterns is crucial for eventually optimizing cropping practices and boosting crop productivity.
The small field sizes and variability in cropping practices of diverse agricultural landscapes requires high spatial and high temporal resolution data for discrimination and monitoring of cropping patterns. Remote sensing plays a key role in monitoring the spatial and temporal variability of growth phases in cropping patterns. This thesis demonstrates the spatiotemporal and spectral variability of cropping patterns using imagery data from Sentinel-2, PlanetScope, and DESIS hyperspectral sensors. The aim was to enhance our understanding of how to monitor and distinguish the complex cropping patterns, particularly those involving intercropping.
This thesis addressed three key aspects of monitoring and discriminating cropping patterns using remote sensing data: the temporal, spectral, and spatial variability of cropping patterns, with a focus on intercropping. First, we focused on understanding annual cropping patterns using remote sensing, which highlighted the gaps and challenges in mapping intercropping. We then identified critical crop growth phases for discriminating cropping patterns using Sentinel-2, emphasizing the importance of temporal resolution. Using Sentinel-2 and DESIS hyperspectral data, we also examined key spectral regions for better discrimination of cropping patterns, exploring the role of spectral resolution. Lastly, we assessed the spatial variability of cropping patterns within different intercropping patterns using PlanetScope. Both primary and secondary field data were used in this study. Key challenges included the lack of detailed field data, diverse cropping practices and the small field sizes, which played a role in the classification accuracy. Additional challenges arose from the complexity of intercropped fields and their dynamic interactions during the crop growing phases. The study highlighted the importance of resolution in identifying critical crop growth phases and improving pattern discrimination through the analysis of temporal, spectral, and spatial variability using Sentinel-2, DESIS, and PlanetScope data, further, despite these challenges, the findings contribute valuable insights into the potential of remote sensing for better cropping pattern monitoring and mapping of smallholder farms.