During the latest decades, the Amazon has experienced a great loss of vegetation cover, in many cases as a direct consequence of wildfires, which became a problem at local, national, and global scales, leading to economic, social, and environmental impacts. Hence, this study is committed to developing a routine for monitoring fires in the vegetation cover relying on recent multitemporal data (2017–2019) of Landsat-8 and Sentinel-2 imagery using the cloud-based Google Earth Engine (GEE) platform. In order to assess the burnt areas (BA), spectral indices were employed, such as the Normalized Burn Ratio (NBR), Normalized Burn Ratio 2 (NBR2), and Mid-Infrared Burn Index (MIRBI). All these indices were applied for BA assessment according to appropriate thresholds. Additionally, to reduce confusion between burnt areas and other land cover classes, further indices were used, like those considering the temporal differences between pre and post-fire conditions: differential Mid-Infrared Burn Index (dMIRBI), differential Normalized Burn Ratio (dNBR), differential Normalized Burn Ratio 2 (dNBR2), and differential Near-Infrared (dNIR). The calculated BA by Sentinel-2 was larger during the three-year investigation span (16.55, 78.50, and 67.19 km²) and of greater detail (detected small areas) than the BA extracted by Landsat-8 (16.39, 6.24, and 32.93 km²). The routine for monitoring wildfires presented in this work is based on a sequence of decision rules. This enables the detection and monitoring of burnt vegetation cover and has been originally applied to an experiment in the northeastern Peruvian Amazon. The results obtained by the two satellites imagery are compared in terms of accuracy metrics and level of detail (size of BA patches). The accuracy for Landsat-8 and Sentinel-2 in 2017, 2018, and 2019 varied from 82.7–91.4% to 94.5–98.5%, respectively.

The fast retreat of the tropical Andean glaciers (TAGs) is considered an important indicator of climate change impact on the tropics, since the TAGs provide resources to highly vulnerable mountain populations. This study aims to reconstruct the glacier coverage of the TAGs, using Landsat time-series images from 1985 to 2020, by digitally processing and classifying satellite images in the Google Earth Engine platform. We used annual reductions of the Normalized Difference Snow Index (NDSI) and spectral bands to capture the pixels with minimum snow cover. We also implemented temporal and spatial filters to have comparable maps at a multitemporal level and reduce noise and temporal inconsistencies. The results of the multitemporal analysis of this study confirm the recent and dramatic recession of the TAGs in the last three decades, in base to physical and statistical significance. The TAGs reduced from 2429.38 km² to 1409.11 km² between 1990 and 2020, representing a loss of 42% of the total glacier area. In addition, the time-series analysis showed more significant losses at altitudes below 5000 masl, and differentiated changes by slope, latitude, and longitude. We found a more significant percentage loss of glacier areas in countries with less coverage. The multiannual validation showed accuracy values of 92.81%, 96.32%, 90.32%, 97.56%, and 88.54% for the metrics F1 score, accuracy, kappa, precision, and recall, respectively. The results are an essential contribution to understanding the TAGs and guiding policies to mitigate climate change and the potential negative impact of freshwater shortage on the inhabitants and food production in the Andean region.

In this study, we used moderate resolution imaging spectroradiometer (MODIS) satellite images to quantify the sedimentation processes in a cascade of six hydropower dams along a 700-km transect in the Paranapanema River in Brazil. Turbidity field measurement acquired over 10 years were used to calibrate a turbidity retrieval algorithm based on MODIS surface reflectance products. An independent field dataset was used to validate the remote sensing estimates showing fine accuracy (RMSE of 9.5 NTU, r = 0.75, N = 138). By processing 13 years of MODIS images since 2000, we showed that satellite data can provide robust turbidity monitoring over the entire transect and can identify extreme sediment discharge events occurring on daily to annual scales. We retrieved the decrease in the water turbidity as a function of distance within each reservoir that is related to sedimentation processes. The remote sensing-retrieved turbidity decrease within the reservoirs ranged from 2 to 62% making possible to infer the reservoir type and operation (storage versus run-of-river reservoirs). The reduction in turbidity assessed from space presented a good relationship with conventional sediment trapping efficiency calculations, demonstrating the potential use of this technology for monitoring the intensity of sedimentation processes within reservoirs and at large scale.

This study assesses the relative performance of Sentinel-1 and -2 and their combination with topographic information for plow agricultural land soil salinity mapping. A learning database made of 255 soil samples’ electrical conductivity (EC) along with corresponding radar (R), optical (O), and topographic (T) information derived from Sentinel-2 (S2), Sentinel-1 (S1), and the SRTM digital elevation model, respectively, was used to train four machine learning models (Decision tree—DT, Random Forest—RF, Gradient Boosting—GB, Extreme Gradient Boosting—XGB). Each model was separately trained/validated for four scenarios based on four combinations of R, O, and T (R, O, R+O, R+O+T), with and without feature selection. The Recursive Feature Elimination with k-fold cross validation (RFEcv 10-fold) and the Variance Inflation Factor (VIF) were used for the feature selection process to minimize multicollinearity by selecting the most relevant features. The most reliable salinity estimates are obtained for the R+O+T scenario, considering the feature selection process, with R² of 0.73, 0.74, 0.75, and 0.76 for DT, GB, RF, and XGB, respectively. Conversely, models based on R information led to unreliable soil salinity estimates due to the saturation of the C-band signal in plowed lands.

Soil moisture content (SMC) is a critical parameter for agricultural productivity, particularly in semi-arid regions, where irrigation practices are extensively used to offset water deficits and ensure decent yields. Yet, the socio-economic and remote context of these regions prevents sufficiently dense SMC monitoring in space and time to support farmers in their work to avoid unsustainable irrigation practices and preserve water resource availability. In this context, our study addresses the challenge of high spatial resolution (i.e., 20 m) SMC estimation by integrating remote sensing datasets in machine learning models. For this purpose, a dataset made of 166 soil samples’ SMC along with corresponding SMC, precipitation, and radar signal derived from Soil Moisture Active Passive (SMAP), Integrated Multi-satellitE Retrievals for GPM (IMERG), and Sentinel-1 (S1), respectively, was used to assess four machine learning models’ (Decision Tree—DT, Random Forest—RF, Gradient Boosting—GB, Extreme Gradient Boosting—XGB) reliability for SMC mapping. First, each model was trained/validated using only the coarse spatial resolution (i.e., 10 km) SMAP SMC and IMERG precipitation estimates as independent features, and, second, S1 information (i.e., 20 m) derived from single scenes and/or composite images was added as independent features to highlight the benefit of information (i.e., S1 information) for SMC mapping at high spatial resolution (i.e., 20 m). Results show that integrating S1 information from both single scenes and composite images to SMAP SMC and IMERG precipitation data significantly improves model reliability, as R² increased by 12% to 16%, while RMSE decreased by 10% to 18%, depending on the considered model (i.e., RF, XGB, DT, GB). Overall, all models provided reliable SMC estimates at 20 m spatial resolution, with the GB model performing the best (R² = 0.86, RMSE = 2.55%).

Fluvial sediment dynamics plays a key role in the Amazonian environment, with most of the sediments originating in the Andes. The Madeira River, the second largest tributary of the Amazon River, contributes up to 50% of its sediment discharge to the Atlantic Ocean, most of it provided by the Andean part of the Madeira basin, in particular the Beni River. In this study, we assessed the rainfall (R)-surface suspended sediment concentration (SSSC) and discharge (Q)-SSSC relationship at the Rurrenabaque station (200 m a.s.l.) in the Beni Andean piedmont (Bolivia). We started by showing how the R and Q relationship varies throughout the hydrological year (September to August), describing a counter-clockwise hysteresis, and went on to evaluate the R–SSSC and Q–SSSC relationships. Although no marked hysteresis is observed in the first case, a clockwise hysteresis is described in the second. In spite of this, the rating curve normally used ( $\mathrm{SSSC}={\mathrm{aQ}}^{\mathrm{b}}$ ) shows a satisfactory R² = 0.73 (p < 0.05). With regard to water discharge components, a linear function relates the direct surface flow Q_s–SSSC, and a hysteresis is observed in the relationship between the base flow Q_b and SSSC. A higher base flow index (Q_b/Q) is related to lower SSSC and vice versa. This article highlights the role of base flow on sediment dynamics and provides a method to analyze it through a seasonal empirical model combining the influence of both Q_b and Q_s, which could be employed in other watersheds. A probabilistic method to examine the SSSC relationship with R and Q is also proposed.