Advancing agricultural monitoring for improved yield estimations using SPOT-VGT and PROBA-V type remote sensing data

Accurate and timely crop condition monitoring is crucial for food management and the economic development of any nation. However, accurately estimating crop yield from the field to global scales is a challenge. According to the global strategy of the World Bank, in order to improve national agricultural statistics, crop area, crop production, and crop yield are key variables that all countries should be able to provide. Crop yield assessment requires that both an estimation of the quantity of a product and the area provided for that product should be available. The definition seems simple; however, these measurements are time consuming and subject to error in many circumstances. Remote sensing is one of several methods used for crop yield estimation. The yield results from a combination of environmental factors, such as soil, weather, and farm management, which are responsible for the unique spectral signature of a crop captured by satellite images. Additionally, yield is an expression of the state, structure, and composition of the plant. Various indices, crop masks, and land observation sensors have been developed to remotely observe and control crops in different regions. This thesis focuses on how much low spatial resolution satellites, such as Project for On Board Autonomy Vegetation (PROBA V), can contribute to global crop monitoring by aiding the search for improved methods and datasets for better crop yield estimation. This thesis contains three chapters. The first chapter explores how an existing product, Dry Matter Productivity (DMP), that has been developed for Satellites Pour l’Observation de la Terre or Earth observing Satellites VeGeTation (SPOT VGT), and transferred to PROBA V, can be improved to more closely relate to yield anomalies across selected regions. This chapter also covers the testing of the contribution of stress factors to improve wheat and maize yield estimations. According to Monteith’s theory, crop biomass linearly correlates with the amount of Absorbed Photosynthetically Active Radiation (APAR) and constant Radiation Use Efficiency (RUE) downregulated by stress factors such as CO2, fertilization, temperature, and water stress. The objective of this chapter is to investigate the relative importance of these stress factors in relation to the regional biomass production and yield. The production efficiency model Copernicus Global Land Service Dry Matter Productivity (CGLS DMP), which follows Monteith’s theory, is modified and evaluated for common wheat and silage maize in France, Belgium, and Morocco using SPOT VGT for the 1999–2012 period. The correlations between the crop yield data and the cumulative modified DMP, CGLS DMP, Fraction of APAR (fAPAR), and Normalized Difference Vegetation Index (NDVI) values are analyzed for different crop growth stages. The best results are obtained when combinations of the most appropriate stress factors are included for each selected region, and the modified DMP during the reproductive stage is accumulated. Though no single solution can demonstrate an improvement of the global product, the findings support an extension of the methodology to other regions of the world. The second chapter demonstrates how PROBA V can be used effectively for crop identification mapping by utilizing spectral matching techniques and phenological characteristics of different crop types. The study sites are agricultural areas spread across the globe, located in Flanders (Belgium), Sria (Russia), Kyiv (Ukraine), and Sao Paulo (Brazil). The data are collected for the 2014–2015 season. For each pure pixel within a field, the NDVI profile of the crop type for its growing season is matched with the reference NDVI profile. Three temporal windows are tested within the growing season: green up to senescence, green up to dormancy, and minimum NDVI at the beginning of the growing season to minimum NDVI at the end of the growing season. In order of importance, the crop phenological development period, parcel size, shorter time window, number of ground truth parcels, and crop calendar similarity are the main reasons behind the differences between the results. The methodology described in this chapter demonstrates the potentials and limitations of using 100 m PROBA V with revisiting frequency every 5 days in crop identification across different regions of the world. The final chapter explores the trade off between the different spatial resolutions provided by PROBA V products versus the temporal frequency and, additionally, explores the use of thermal time to improve statistical yield estimations. The ground data are winter wheat yields at the field level for 39 fields across Northern France during one growing season 2014–2015. An asymmetric double sigmoid function is fitted, and the NDVI values are integrated over thermal time and over calendar time for the central pixel of the field, exploring different thresholds to mark the start and end of the cropping season. The integrated NDVI values with different NDVI thresholds are used as a proxy for yield. In addition, a pixel purity analysis is performed for different purity thresholds at the 100 m, 300 m, and 1 km resolutions. The findings demonstrate that while estimating winter wheat yields at the field level with pure pixels from PROBA V products, the best correlation is obtained with a 100 m resolution product. However, several fields must be omitted due to the lack of observations throughout the growing season with the 100 m resolution dataset, as this product has a lower temporal resolution compared to 300 m and 1 km. This thesis is a modest contribution to the remote sensing and data analysis field with its own merits, in particular with respect to PROBA V. The experiments provide interesting insight into the PROBA V dataset at 1 km, 300 m, and 100 m resolutions. Specifically, the results show that 100 m spatial resolution imagery could be used effectively and advantageously in agricultural crop monitoring and crop identification at local – field level – and regional – the administrative regions defined by the national governments – levels. Furthermore, this thesis discusses the limitations of using a low resolution satellite, such as the PROBA V 100 m dataset, in crop monitoring and identification. Also, several recommendations are made for space agencies that can be used when designing the new generation of satellites.