ABSTRACT
A spatiotemporal machine learning framework for automated prediction and analysis of long-term Land Use/Land Cover dynamics is presented. The framework includes: (1) harmonization and preprocessing of spatial and spatiotemporal input datasets (GLAD Landsat, NPP/VIIRS) including five million harmonized LUCAS and CORINE Land Cover-derived training samples, (2) model building based on spatial k-fold cross-validation and hyper-parameter optimization, (3) prediction of the most probable class, class probabilities and model variance of predicted probabilities per pixel, (4) LULC change analysis on time-series of produced maps. The spatiotemporal ensemble model consists of a random forest, gradient boosted tree classifier, and an artificial neural network, with a logistic regressor as meta-learner. The results show that the most important variables for mapping LULC in Europe are: seasonal aggregates of Landsat green and near-infrared bands, multiple Landsat-derived spectral indices, long-term surface water probability, and elevation. Spatial cross-validation of the model indicates consistent performance across multiple years with overall accuracy (a weighted F1-score) of 0.49, 0.63, and 0.83 when predicting 43 (level-3), 14 (level-2), and five classes (level-1). Additional experiments show that spatiotemporal models generalize better to unknown years, outperforming single-year models on known-year classification by 2.7% and unknown-year classification by 3.5%. Results of the accuracy assessment using 48,365 independent test samples shows 87% match with the validation points. Results of time-series analysis (time-series of LULC probabilities and NDVI images) suggest forest loss in large parts of Sweden, the Alps, and Scotland. Positive and negative trends in NDVI in general match the land degradation and land restoration classes, with "urbanization" showing the most negative NDVI trend. An advantage of using spatiotemporal ML is that the fitted model can be used to predict LULC in years that were not included in its training dataset, allowing generalization to past and future periods, e.g. to predict LULC for years prior to 2000 and beyond 2020. The generated LULC time-series data stack (ODSE-LULC), including the training points, is publicly available via the ODSE Viewer. Functions used to prepare data and run modeling are available via the eumap library for Python.
Subject(s)
Environmental Monitoring , Urbanization , Probability , Europe , Time FactorsABSTRACT
This paper will assess the sensitivity of soil moisture anomaly (SMA) obtained from the Soil water index (SWI) product Metop ASCAT, to identify drought in Romania. The SWI data were converted from relative values (%) to absolute values (m3 m-3) using the soil porosity method. The conversion results (SM) were validated using soil moisture in situ measurements from ISMN at 5 cm depths (2015-2020). The SMA was computed based on a 10 day SWI product, between 2007 and 2020. The analysis was performed for the depths of 5 cm (near surface), 40 cm (sub surface), and 100 cm (root zone). The standardized precipitation index (SPI), land surface temperature anomaly (LST anomaly), and normalized difference vegetation index anomaly (NDVI anomaly) were computed in order to compare the extent and intensity of drought events. The best correlations between SM and in situ measurements are for the stations located in the Getic Plateau (Bacles (r = 0.797) and Slatina (r = 0.672)), in the Western Plain (Oradea (r = 0.693)), and in the Moldavian Plateau (Iasi (r = 0.608)). The RMSE were between 0.05 and 0.184. Furthermore, the correlations between the SMA and SPI, the LST anomaly, and the NDVI anomaly were significantly registered in the second half of the warm season (July-September). Due to the predominantly agricultural use of the land, the results can be useful for the management of water resources and irrigation in regions frequently affected by drought.
