Protected forests in the Brazilian Amazon safeguard remnants of undisturbed forests. However, some of these forests are legally managed by sustainable logging practices and may also be exposed to improper or illegal logging, increasing forest degradation in these areas. This study assesses the potential and limitations of REDD+AI - a deep learning-based approach that uses high-resolution Planet NICFI imagery - to map sustainable forest management in the Jamari National Forest (south-western Brazilian Amazon) and compare its relative performance to other forest disturbance datasets. To validate the satellite-based estimates, we used a stratified sampling design to estimate REDD+AI statistical accuracy of logging detections based on visual interpretation of time series of high spatial resolution imagery. We also used field data provided by the Brazilian Forest Service to estimate the percentage of detections inside management units. Three datasets of forest disturbance and degradation detection were compared: DETER - INPE; GFC - UMD; and TMF - JRC. Our findings showed that REDD+AI had higher overall accuracy (95.65% +/- 1 [95% CI]) than DETER (87.40%), TMF (86.77%) and GFC (86.55%) to map logging over the Jamari Forest. Furthermore, a higher spatial agreement with the production units was found for REDD+AI (83.07%), followed by DETER (21.89%), TMF (4.47%) and GFC (2.36%). We estimate that 13.74% of the Jamari forest has undergone some level of logging between 2016 and 2022, of which approximately 71% were located within the management zones and 29% outside the management zones, strongly indicating anthropism-related and illegal logging in this protected area. This study highlights REDD+AI as a tool for monitoring logging in tropical forests under management which can provide vital support for government oversight efforts to combat illegal logging and ensure compliance within sustainable forest management practices using modern remote sensing applications.

Selective logging often marks the beginning of forest degradation, but remote sensing alert systems typically detect it long after it starts. This type of disturbance produces a subtle canopy signal not easily recognised in the early stages, causing alerts in some areas of the Brazilian Legal Amazon to lag behind the first visible signs by several months. This study explores whether a geospatial foundation model can help reduce this time gap. For each documented disturbance site, we compile Harmonized Landsat and Sentinel-2 (HLS) image time series spanning October 2024 to May 2026 into a datacube and use it to pretrain, without labels, a compact spatiotemporal masked autoencoder (ST-MAE). This model learns to embed the dynamics of forest degradation into latent-space representations, the so-called "embeddings". Following this, a lightweight downstream model — trained with few labelled samples — interprets ST-MAE's embeddings to evaluate every new HLS observation for recent logging activity. To effectively translate the evaluation scores into production-ready alerts, we apply a simple change-detection approach: an edge filter for sudden logging-level changes, paired with a persistence filter for sustained detections. We conducted a leave-one-out cross-validation assessment across 20 well-documented logging sites: yielding no false alarms prior to the first disturbance evidence, the change detector confirmed the onset of the 20 logging activities with a median delay of 15 days after the first post-disturbance image. The downstream model's scores remain low during the pre-disturbance period but increase once logging begins, confirming that it responds to logging-related events rather than to other landscape dynamics or calendar artefacts. This label-free pretraining enables effective performance with minimal annotations, providing a practical solution to reduce the temporal gap in monitoring forest degradation.

Between 2023 and 2024, Brazil experienced one of the most severe droughts in the century, with temperatures in the Amazon reaching 2-4 degrees C above average. Amplified by El Nino and North Atlantic warming, the drought disrupted the hydrological cycle and essential ecosystem services, affecting agriculture, transportation, and the livelihoods of urban, rural, and Indigenous communities. This event revealed substantial shortcomings in Brazil's fire governance framework. Moreover, it demonstrated that drought conditions exacerbate Brazil's susceptibility to wildfires, underscoring the urgent need for coordinated and immediate policy action. This study maps and quantifies hydrological stress, forest degradation, and fire dynamics associated with the drought, identifying critical implications for environmental management. Our analysis reveals the intensification of a drought-fire-degradation feedback loop. Wildfire-affected areas increased by 9%, and degradation alerts rose by 19% in relation to long-term average levels (2016-2024). At the drought's peak (in 2024), 4.2 Mha of fire-affected areas were detected. Maximum Cumulative Water Deficit anomalies exceeded historical thresholds, signaling a continuous dry increase in Brazil. These results highlight the urgent need for integrated fire governance. This includes comprehensive environmental education, enhanced early warning systems, targeted investments in ecosystem restoration, and stronger coordination between federal and state policies. By linking scientific evidence to actionable policy recommendations, this work guides decision-makers seeking to reduce socio-environmental risks and strengthen climate adaptation in the Amazon.

Understanding future changes in cumulative water deficit (CWD) is essential for assessing the vulnerability of Amazonian ecosystems to climate change. This study evaluates the performance of CMIP6 models in simulating CWD in the southwestern Amazon from 1985 to 2024 and projects future changes through 2100 under three emission scenarios (SSP1-2.6, SSP3-7.0, SSP5-8.5). CWD was calculated using a fixed evapotranspiration threshold of 100 mm/month, and the maximum cumulative water deficit (MCWD) was derived to quantify overall water stress over time. Model performance was assessed using statistical metrics (MAE, RMSE, bias and Pearson correlation), and the best-performing models were selected for future projections. Additionally, time series decomposition with breakpoint detection (BFAST) was applied. Results show that most models reproduce the seasonal cycle but diverge during peak dry months. Under SSP5-8.5, deficits intensify significantly, exceeding 150 mm between July and September. CAS-ESM2-0 and CESM2-WACCM exhibit large negative biases, while BCC-CSM2-MR and IPSL-CM6A-LR perform best. SSP1-2.6 shows minimal changes, whereas SSP3-7.0 and SSP5-8.5 exhibit significant negative trends, with depletion rates reaching -21 mm/month by 2100. Temporal decomposition reveals increased hydrological instability, structural breaks after 2040 (SSP3-7.0) and 2050 (SSP5-8.5), and amplified seasonality under high emissions. Residual variability also increases, indicating a higher frequency of non-seasonal anomalies. These findings highlight growing water stress under higher warming scenarios and emphasise CWD as a key indicator of ecological risk. Reliable projections depend on careful model evaluation, underscoring the need for model selection to guide climate adaptation and forest conservation strategies.

The Amazon Rainforest, crucial for climate regulation, carbon and water cycles, and biodiversity preservation, faces escalating threats from heightened forest degradation, including disturbances from fire and logging. In 2020, Brazil was responsible for a concerning 70% of the active fire hotspots detected in the Amazon, signaling a notable 60% increase compared to 2019. This surge has pushed the region into an extreme fire situation. Urgent and effective interventions are imperative to mitigate these extremes, ensuring the preservation of the Amazon and global climate stability. The study focuses on the Boca do Acre region in the southwest Amazon, one of the most recent hotspots of deforestation and forest degradation in the Amazon. We project the suitability of fire for 2030, following the timeframe set by the United Nations for the implementation of actions aimed at creating a better world for all peoples and nations through the Agenda 2030. Using the MAXENT algorithm within the R software, we conducted a detailed analysis exclusively within the non-forest land-use class on a 5 km x 5 km grid. Burned area data from products Fire CCI (250m), MapBiomas Fire (30m), and MODIS MCD64 (500m) were used to study fire occurrence across the study area. The chosen baseline year is 2014, representing the last year of historical data before the influence of different Shared Socioeconomic Pathways (SSPs) on IPCC models (1-2.6 and 3-7.0). The statistic involves the use of specifically selected variables, determined by their performance in correlation tests and principal component analysis. These variables encompass the percentage of forested areas, agriculture, pasture, and a 1000 m buffer along the region's roads. Additionally, factors such as the percentage of conservation unit occupancy, indigenous lands, and medium-sized properties (400-1000 ha) in the Rural Environmental Registry (CAR), along with precipitation values during dry months, are taken into account. Model validation incorporates AUC analysis, where the model must exhibit performance greater than 0.7, background analysis with the same curve behavior, false positive rate (FPR), accuracy evaluation, and sensitivity analysis. Following this process, we project the feasibility of fire for 2030. Results consistently demonstrate high performance, with AUC values surpassing 0.7 and pixel-to-pixel accuracy ranging from 60% to 90%, lower FRP values, and higher sensitivity values. Projected results indicate an increased susceptibility to fires that spread in the region, especially under less sustainable scenarios, emphasizing the urgency of preventive measures before 2030. Projections reveal an advancement in fire suitability, particularly in the SSP 3-7.0 scenario, with a significant increase in non-forest areas. However, as the scenario worsens, areas prone to fires that spread decrease due to the advancement of agricultural and pasture areas, underscoring the need for more sustainable practices. In conclusion, this study holds promise as a management tool for decision-makers, offering valuable insights for the development of mitigation and adaptation measures to climate change in the Boca do Acre region. These contributions are essential for preserving this vital ecosystem, highlighting the importance of implementing effective strategies.

Anthropogenic disturbances stand as the primary driver of degradation in the remaining Amazon forests, posing a significant threat to their future. Notable among these disturbances are edge effects, timber extraction, fire, extreme droughts and temperatures, which have been intensified by human-induced climate change. A pilot study aiming to integrate forest fire occurrence, timber extraction and climate change scenarios was developed for a new deforestation frontier in southwestern Amazonia. We integrated a series of remote sensing fire products, spatialized land tenure information, selective logging mapping techniques and Global Climate Models (GCMs) simulated projections of three SSPs (SSP climate forcing scenarios) for 2015–2100 period. The results showed that the increased deforestation trend occurred between 2003 and 2019 predominantly on public lands, following the implementation of the new forest code. This surge contributed to a spike in fires, escalating from 66% to 84% in 2019. Over the period from 2007 and 2019, 2.4% of the primary forest was logged. By 2022, precipitation values aligned closely with SSP 5-8.5, and temperature values neared SSP 3-7.0. Projections for 2100 indicated an alarming increase of 5.19 ºC in overall temperature and a reduction of 55 mm in annual precipitation compared to 2003 baseline. The results indicate that the study region is already heading towards a less sustainable future. Logging activities, as well as agricultural production, are threatened by both increase in economic losses by fires and temperatures, and rainfall reduction. Implementing mitigation measures, such as fire-free land management, traceability controls for all wood production from logged forests, and addressing issues of land tenure and regulation are pivotal in steering the current development pathway towards a more sustainable pathway.

The Amazon is home to a vast biodiversity and plays a crucial role in the maintenance and conservation of the planet. The progress of anthropogenic activities in the region is one of the drivers of land cover change. Studying these changes through remote sensing enables the monitoring of degradation and deforestation in the Amazon rainforest. Using bibliometric tools this study analyses the different remote sensing approaches for land use and land cover monitoring in the Amazon. The methodology includes 1) data search and collection, 2) tool selection and data processing, and 3) data and trend analysis. The results indicate that studies on forest loss and biodiversity are trending in the field. Emerging topics focus on new tools for environmental impact assessment and evaluating new computational technologies for satellite data management.

Gaps are openings within tropical forest canopies created by natural or anthropogenic disturbances. Important aspects of gap dynamics that are not well understood include how gaps close over time and their potential for contagiousness, indicating whether the presence of gaps may or may not induce the creation of new gaps. This is especially important when we consider disturbances from selective logging activities in rainforests, which take away large trees of high commercial value and leave behind a forest full of gaps. The goal of this study was to quantify and understand how gaps open and close over time within tropical rainforests using a time series of airborne LiDAR data, attributing observed processes to gap types and origins. For this purpose, the Jamari National Forest located in the Brazilian Amazon was chosen as the study area because of the unique availability of multi-temporal small-footprint airborne LiDAR data covering the time period of 2011–2017 with five data acquisitions, alongside the geolocation of trees that were felled by selective logging activities. We found an increased likelihood of natural new gaps opening closer to pre-existing gaps associated with felled tree locations (<20 m distance) rather than farther away from them, suggesting that small-scale disturbances caused by logging, even at a low intensity, may cause a legacy effect of increased mortality over six years after logging due to gap contagiousness. Moreover, gaps were closed at similar annual rates by vertical and lateral ingrowth (16.7% yr−1) and about 90% of the original gap area was closed at six years post-disturbance. Therefore, the relative contribution of lateral and vertical growth for gap closure was similar when consolidated over time. We highlight that aboveground biomass or carbon density of logged forests can be overestimated if considering only top of the canopy height metrics due to fast lateral ingrowth of neighboring trees, especially in the first two years of regeneration where 26% of gaps were closed solely by lateral ingrowth, which would not translate to 26% of regeneration of forest biomass. Trees inside gaps grew 2.2 times faster (1.5 m yr−1) than trees at the surrounding non-gap canopy (0.7 m yr−1). Our study brings new insights into the processes of both the opening and closure of forest gaps within tropical forests and the importance of considering gap types and origins in this analysis. Moreover, it demonstrates the capability of airborne LiDAR multi-temporal data in effectively characterizing the impacts of forest degradation and subsequent recovery.

Tropical rainforests from the Brazilian Amazon are frequently degraded by logging, fire, edge effects and minor unpaved roads. However, mapping the extent of degradation remains challenging because of the lack of frequent high-spatial resolution satellite observations, occlusion of understory disturbances, quick recovery of leafy vegetation, and limitations of conventional reflectance-based remote sensing techniques. Here, we introduce a new approach to map forest degradation caused by logging, fire, and road construction based on deep learning (DL), henceforth called DL-DEGRAD, using very high spatial (4.77 m) and bi-annual to monthly temporal resolution of the Planet NICFI imagery. We applied DL-DEGRAD model over forests of the state of Mato Grosso in Brazil to map forest degradation with attributions from 2016 to 2021 at six-month intervals. A total of 73,744 images (256 x 256 pixels in size) were visually interpreted and manually labeled with three semantic classes (logging, fire, and roads) to train/validate a U-Net model. We predicted the three classes over the study area for all dates, producing accumulated degradation maps biannually. Estimates of accuracy and areas of degradation were performed using a probability design-based stratified random sampling approach (n = 2678 samples) and compared it with existing operational data products at the state level. DL-DEGRAD performed significantly better than all other data products in mapping logging activities (F-1-score = 68.9) and forest fire (F-1-score = 75.6) when compared with the Brazil's national maps (SIMEX, DETER, MapBiomas Fire) and global products (UMD-GFC, TMF, FireCCI, FireGFL, GABAM, MCD64). Pixel-based spatial comparison of degradation areas showed the highest agreement with DETER and SIMEX as Brazil official data products derived from visual interpretation of Landsat imagery. The U-Net model applied to NICFI data performed as closely to a trained human delineation of logged and burned forests, suggesting the methodology can readily scale up the mapping and monitoring of degraded forests at national to regional scales. Over the state of Mato Grosso, the combined effects of logging and fire are degrading the remaining intact forests at an average rate of 8443 km(2) year(-1) from 2017 to 2021. In 2020, a record degradation area of 13,294 km(2) was estimated from DL-DEGRAD, which was two times the areas of deforestation.

The balance between environment and economic stability is key to sustainable rural development. This is particularly true in the agricultural frontier areas along the Transamazon Highway and Southern Par & PRIME;a in the Brazilian Amazon, where thousands of migrant families have settled in the forests over the past 50 years in search of a better life. To better understand the extent to which sustainable development is possible in such a context, this study examined the standard of living that smallholders who grow cocoa can achieve compared to those who raise cattle, and what this means for forest conservation. An analysis of 95 households revealed that both livelihood strategies may generate an acceptable standard of living despite significant logistical and environmental challenges. This was observed even more so for families who combined both production systems. The availability of technology and the size of landholdings had the greatest impact on the standard of living expressed in income and housing conditions. The majority of the analyzed households, especially those involved in cattle ranching, converted their forests for economic success. The pure cocoa farmers behaved differently, but also cleared large areas of forest and may continue to do so. The findings suggest that achieving sustainable local development in Amazonian agricultural frontiers requires large and well-coordinated investments by competent public and private actors not only in building and optimizing sustainable production systems such as cocoa agroforestry but also in significantly improving social rural infrastructure.

. The Amazon has the largest remnant of tropical forests in the world and is being threatened daily by deforestation and forest degradation. The estimated area of forest degradation is underestimated, which is problematic for both sustainable policy enforcement, environmental oversight and national carbon emission inventories. Therefore, a deep learning model (U-Net) was trained to map forest degradation using Planet imagery (4.77 m spatial resolution) in the Jamari National Forest at the Brazilian Amazon. Preliminary results showed an overall accuracy of 67%. Our approach is promising to monitor forest concessions in Amazonia. Resumo. A Amazônia possui

Restoration projects designed to promote one ecosystem service may have synergistic benefits to other services. Therefore, bundling them can be an effective way to maximize the return to the investments in programs of payments for ecosystem services (PES). Here, we investigated the additional gain of restoration actions—which were implemented as part of a PES program to protect a key watershed for water supply—on increasing functional landscape connectivity in the Atlantic Forest region of southeastern Brazil. Using a landscape ecology approach, we estimated the amount of forest cover before (2006) and after (2012) restoration activities by the PES program and changes in structural and functional landscape connectivity for birds with varying gap-crossing capabilities. Forest cover increased from 42.5 to 86.1 ha after the implementation of restoration projects by the PES program. In the simulated scenarios of landscape connectivity, the mean patch size of functionally connected forest increased by 1,034%, 392%, 248%, and 94% for species with gap-crossing capabilities of 0, 20, 40, and 60 m, respectively. Our results highlight the potential for incorporating biodiversity conservation objectives into PES projects primarily designed to enhance water-related ecosystem services.

Agroforestry systems (AFS) are important agricultural land use in synergy with socio-environmental aspects, especially with cacao (Theobroma cacao L.) crop, a commodity mainly produced by smallholders in the humid tropics. In southern Pará, Brazilian Amazon, farmers manage native shade trees growing with cacao, but species selection may be not appropriate to AFS maintenance over time. The objective of this study was to understand the shade trees transition between successional management phases of cacao-AFS, considering its initial shade (IS) and secondary shade (SS). It was sampled 10 plots in each situation (20,000 m2 in total) identifying individuals with CBH ≥ 15 cm. As expected, floristic composition was different and SS had greater species richness and diversity than IS, where only 17% of species were the shared among them. Musa sp. and Carica papaya L. were found only in IS and were dominant species, representing almost a half of the individuals. Although there was increase of late succession species from IS to SS, this still keeps high abundance of early succession species, such as Cecropia sp. The result shows an unexploited potential products and gap of services provision, such as N-fixing. The conclusion highlights the necessity of long-term succession planning and management practices to guarantee cacao crop maintenance and improve diversification with other income sources, such as fruits and wood. The role of biodiversity conservation, provided by shade trees, should be the target of political strategies to encourage its maintenance, such as payment for ecosystems services or other economic incentives.