Search

The current study assesses the performance of several machine learning (ML) and deep learning (DL) models for detecting and mapping floods using Sentinel-1 SAR imagery. Three distinct approaches were used: pixel classification, object-based image analysis and object instance segmentation. The ML models are Random Forest, and Support Vector Machine and the DL models are U-NET, DeepLabV3 and Mask RCNN. Five different case studies were selected to test the models’ scalability. These areas are in Romania (Prut River, at the border between Romania, the Republic of Moldova and Ukraine, Timiș River, and Râul Negru River), the United States of America (Missouri River) and Australia (Broughton Creek). For each flood, five Sentinel-1 images were used, four collected before the flood and one collected after the flood. The intensity images were stacked and scaled in the range of the intensity thresholds associated with water and non-water so that all the case studies have the same margins for intensity. Samples of water, vegetation, agricultural fields, and bare soil were collected only from the Prut River case study and used in the training process. Out of all models, the U-Net model returned the highest accuracy with a value for Intersect over Union of 0.763 for a tile size of 128x128 pixels.

Abstract The largest megalake in the geological record formed in Eurasia during the late Miocene, when the epicontinental Paratethys Sea became tectonically-trapped and disconnected from the global ocean. The megalake was characterized by several episodes of hydrological instability and partial desiccation, but the chronology, magnitude and impacts of these paleoenvironmental crises are poorly known. Our integrated stratigraphic study shows that the main desiccation episodes occurred between 9.75 and 7.65 million years ago. We identify four major regressions that correlate with aridification events, vegetation changes and faunal turnovers in large parts of Europe. Our paleogeographic reconstructions reveal that the Paratethys was profoundly transformed during regression episodes, losing ~ 1/3 of the water volume and ~ 70% of its surface during the most extreme events. The remaining water was stored in a central salt-lake and peripheral desalinated basins while vast regions (up to 1.75 million km2) became emergent land, suitable for development of forest-steppe landscapes. The partial megalake desiccations match with climate, food-web and landscape changes throughout Eurasia, although the exact triggers and mechanisms remain to be resolved.

Earth Observation (EO) makes it possible to obtain information on key parameters characterizing interactions among Earth’s system components, such as evaporative fraction (EF) and surface soil moisture (SSM). Notably, techniques utilizing EO data of land surface temperature (Ts) and vegetation index (VI) have shown promise in this regard. The present study investigates, for the first time, the accuracy of one such technique, known as the “simplified triangle”, using Sentinel-3 EO data, acquired for 44 days in 2018 at three savannah FLUXNET sites in Spain. The technique was found to be able to predict both EF and SSM with reasonable accuracy when compared to collocated ground measurements. Comparisons performed for all days together showed relatively low Root Mean square Difference (RMSD) for both EF (0.191) and SSM (0.012 cm3 cm−3) and good correlation coefficients (R) of 0.721 and 0.577, respectively. Both EF and SSM were also largely in agreement with land cover and seasonal variability. The present study comprises the first detailed assessment of the “simplified triangle”, in this case, using Sentinel-3 data and in a Mediterranean setting. Findings, albeit preliminary, are of significant value regarding the use of the investigated technique as a tool of environmental management, and towards ongoing, worldwide efforts aiming at developing operationally relevant products based on the Ts/VI feature space and EO data based on new satellites such as Sentinel-3.

Traditional methods for individual tree-crown (ITC) detection (image classification, segmentation, template matching, etc.) applied to very high-resolution remote sensing imagery have been shown to struggle in disparate landscape types or image resolutions due to scale problems and information complexity. Deep learning promised to overcome these shortcomings due to its superior performance and versatility, proven with reported detection rates of ~90%. However, such models still find their limits in transferability across study areas, because of different tree conditions (e.g., isolated trees vs. compact forests) and/or resolutions of the input data. This study introduces a highly replicable deep learning ensemble design for ITC detection and species classification based on the established single shot detector (SSD) model. The ensemble model design is based on varying the input data for the SSD models, coupled with a voting strategy for the output predictions. Very high-resolution unmanned aerial vehicles (UAV), aerial remote sensing imagery and elevation data are used in different combinations to test the performance of the ensemble models in three study sites with highly contrasting spatial patterns. The results show that ensemble models perform better than any single SSD model, regardless of the local tree conditions or image resolution. The detection performance and the accuracy rates improved by 3–18% with only as few as two participant single models, regardless of the study site. However, when more than two models were included, the performance of the ensemble models only improved slightly and even dropped.

This paper presents a methodology for evaluating the uncertainties caused by the misalignment between two digital elevation models in estimating landslide kinematics. The study focuses on the earthflow near the town of Chirlești, located in the Bend Subcarpathians, Buzău County, Romania, which poses a high risk of blocking the DN10 national road. Four flights were conducted between 2018 and 2022 using a DJI Phantom 4 UAV using the same flight plan. Monte Carlo simulations were used to model uncertainty propagation of the DEM misalignments in the landslide kinematics analysis. The simulations were applied to the accuracy values of the structure from a motion process used to generate the digital elevation models. The degree of uncertainty was assessed using the displaced material’s total amount in conjunction with the spatial correlation of the displaced material between two consecutive flights. The results revealed that the increase in the RMS values did not determine an increase in the displaced earth between two UAV flights. Instead, combining the RMS values and the correlation coefficient clearly indicated the correspondence between the spatial distribution of the displaced earth material and the overall changes reported between the two UAV flights. An RMS value of up to 1 unit associated with a correlation coefficient of 0.95 could be considered the maximum allowable error for estimating landslide kinematics across space and time. The current methodology is reliable when studying slow-movement landslides and when using short intervals between UAV flights. For rapid movements or significant terrain changes, such as translational and rotational landslides, careful analysis of the correlation coefficient in conjunction with the RMS values is recommended.

The interaction between precipitation and vegetation plays a significant role in the formation of runoff, and it is still a widely discussed issue in hydrology. The difficulty lies in estimating the degree to which a forest influences runoff generation, especially flood peaks, on the one hand, due to the small amount of information regarding the evolution of the forest area and density, and, on the other hand, the correlations between these indicators and the runoff and precipitation values. The analysis focuses on a small basin in the mountain region of Romania, the upper basin of the Ruscova River located in northwestern Romania. In this river basin, there is no significant anthropic influence, other than the intense deforestation and reforestation actions. Using satellite images captured by Landsat missions 5, 7 and 8 for the period 1985–2019, the forest canopy density vegetation index was extracted. Using a gridded precipitation dataset, a hydrological model was calibrated based on three scenarios to assess the impact of forest vegetation on the runoff. Analysis of the results of these models conducted on scenarios allowed us to deduce a simple equation for estimating the influence of forest area on maximum river flows. The analysis showed that even small differences in the forest surface area exert an influence on the peak flow, varying between −5.28% and 8.09%.