Cropland Suitability Assessment Using Satellite-Based Biophysical Vegetation Properties and Machine Learning

Abstract

The determination of cropland suitability is a major step for adapting to the increased food demands caused by population growth, climate change and environmental contamination. This study presents a novel cropland suitability assessment approach based on machine learning, which overcomes the limitations of the conventional GIS-based multicriteria analysis by increasing computational efficiency, accuracy and objectivity of the prediction. The suitability assessment method was developed and evaluated for soybean cultivation within two 50 × 50 km subsets located in the continental biogeoregion of Croatia, in the four-year period during 2017–2020. Two biophysical vegetation properties, leaf area index (LAI) and a fraction of absorbed photosynthetically active radiation (FAPAR), were utilized to train and test machine learning models. The data derived from a medium-resolution satellite mission PROBA-V were prime indicators of cropland suitability, having a high correlation to crop health, yield and biomass in previous studies. A variety of climate, soil, topography and vegetation covariates were used to establish a relationship with the training samples, with a total of 119 covariates being utilized per yearly suitability assessment. Random forest (RF) produced a superior prediction accuracy compared to support vector machine (SVM), having the mean overall accuracy of 76.6% to 68.1% for Subset A and 80.6% to 79.5% for Subset B. The 6.1% of the highly suitable FAO suitability class for soybean cultivation was determined on the sparsely utilized Subset A, while the intensively cultivated agricultural land produced only 1.5% of the same suitability class in Subset B. The applicability of the proposed method for other crop types adjusted by their respective vegetation periods, as well as the upgrade to high-resolution Sentinel-2 images, will be a subject of future research.

The Role of Remote Sensing Data and Methods in a Modern Approach to Fertilization in Precision Agriculture

“…Four of the most commonly applied machine learning methods in previous studies indexed in the WoSCC were used: random forest (RF), support vector machine (SVM), artificial neural networks (ANN) and decision tree minor-and major-scales [82], climate data and auxiliary soil information are commonly included in the modern approach [73]. These values are generally more suitable for the macro-location studies due to their local homogeneity, as well as the lack of available spatial data at the higher spatial resolution to match those of satellite images and DEMs [2]. Despite the same spatial resolution of the P2O5 and K2O rasters produced by the conventional and modern approaches, modern machine learning methods have resulted in much less smooth areas, retaining specific local information about field conditions, which are a backbone for precision agriculture [7].…”

“…The increasing need for food production due to population growth is the cause of the continuous intensification of agricultural production, which leads to higher yields of agricultural crops on existing agricultural land [1]. Modern intensive agriculture makes fertilization one of the primary ways to achieve this goal; however, the short-term objectives of agricultural land management could seriously affect the long-term sustainability of food production [2]. Several authors have pointed out that farmers should not only consider fertilization with the aim of increasing their economic profits, but also in relation to maintaining the long-term biochemical soil composition and its effects on the ecosystem [3,4].…”

A UAS and Machine Learning Classification Approach to Suitability Prediction of Expanding Natural Habitats for Endangered Flora Species

“…To perform such a procedure, due to the narrow ecological gradients of the present flora species, an accurate evaluation of the habitat suitability according to relevant abiotic factors is necessary [8]. These abiotic factors generally include climate, soil, and topography environmental components, which are continuously interacting with vegetation [9].…”

“…The conventional approach to habitat suitability studies include subjective methods with marginal reproducibility as they are dominantly affected by expert assumptions. Geographic information system (GIS)-based multicriteria anaylsis in combination with individual weight determination methods, such as analytic hierarchy procedure, enables flexible suitability calculations but is deficient with regard to computational efficiency and reliability [9]. To overcome these limitations while maintaining flexibility and straightforwardness in spatial prediction, machine learning methods have been increasingly adopted in habitat suitability analyses [17].…”

“…To provide an efficient multispectral imaging solution with high spatial resolution in restricted locations, unmanned aerial systems (UASs) have been successfully implemented in various nature conservation studies, specifically ensuring non-invasive data collection in sensitive study areas [22]. Various machine learning classification algorithms have recently been established as state-of-the-art methods for suitability prediction in various disciplines, such as agriculture [9], forestry [23], nature and environment conservation [24], including land and marine contamination studies [25,26]. While these methods have been success-fully utilized in previous studies [27,28], habitat suitability prediction methods according to environmental criteria have been relatively unexplored, especially for the purpose of extending the habitat of endangered flora species [29].…”