Comparing Machine and Deep Learning Methods for the Phenology-Based Classification of Land Cover Types in the Amazon Biome Using Sentinel-1 Time Series

Abstract

The state of Amapá within the Amazon biome has a high complexity of ecosystems formed by forests, savannas, seasonally flooded vegetation, mangroves, and different land uses. The present research aimed to map the vegetation from the phenological behavior of the Sentinel-1 time series, which has the advantage of not having atmospheric interference and cloud cover. Furthermore, the study compared three different sets of images (vertical–vertical co-polarization (VV) only, vertical–horizontal cross-polarization (VH) only, and both VV and VH) and different classifiers based on deep learning (long short-term memory (LSTM), Bidirectional LSTM (Bi-LSTM), Gated Recurrent Units (GRU), Bidirectional GRU (Bi-GRU)) and machine learning (Random Forest, Extreme Gradient Boosting (XGBoost), k-Nearest Neighbors, Support Vector Machines (SVMs), and Multilayer Perceptron). The time series englobed four years (2017–2020) with a 12-day revisit, totaling 122 images for each VV and VH polarization. The methodology presented the following steps: image pre-processing, temporal filtering using the Savitsky–Golay smoothing method, collection of samples considering 17 classes, classification using different methods and polarization datasets, and accuracy analysis. The combinations of the VV and VH pooled dataset with the Bidirectional Recurrent Neuron Networks methods led to the greatest F1 scores, Bi-GRU (93.53) and Bi-LSTM (93.29), followed by the other deep learning methods, GRU (93.30) and LSTM (93.15). Among machine learning, the two methods with the highest F1-score values were SVM (92.18) and XGBoost (91.98). Therefore, phenological variations based on long Synthetic Aperture Radar (SAR) time series allow the detailed representation of land cover/land use and water dynamics.

Big Data and Machine Learning to Improve European Grapevine Moth (Lobesia botrana) Predictions

“…The F1 score performance metric was used to assess the accuracy of both the Touzeau and ML models. Although it is slightly more difficult to calculate for the Touzeau model results than other metrics such as R 2 , it has been used by several authors to compare the results of classification models [ 53 , 84 , 85 ]. F1 is a widely-used metric for assessing the performance of binary classification problems and for comparing the results obtained by different models.…”

Feasibility analysis of using Sentinel-1 images to phenologically differentiate the areas of soybean seed and sub-irrigated bean planting in the period of sanitary void in the tropical floodplains of the Formoso River basin, Tocantins, Brazil

“…Therefore, the LSTM method and its variants have been widely used in optical image time series [19]- [21]. The advent of Sentinel-1 images with high temporal resolution provided an increase in the use of these algorithms in radar images for mapping crops [22]- [24] and land use and land cover [25], [26].…”

Automated Mapping of Land Cover Type within International Heterogenous Landscapes Using Sentinel-2 Imagery with Ancillary Geospatial Data

“…It is important to note that such intercomparison of studies can be misleading, as accuracy metrics are typically impacted by the total number of classes, definition of classes, image resolution, local geography, and climate. Studies have evaluated deep-learning- and shallow-machine-learning-based classifiers and found the latter to have similar or slightly lower accuracy, typically with less than a 5% difference [ 52 , 67 , 68 , 69 , 70 , 71 , 72 , 73 ] but occasionally higher than 10% [ 74 , 75 ].…”