Tropical dry forests (TDFs) are ecosystems that are highly vulnerable to drought and face threats from human activities. However, long-term assessments of ecological drought dynamics across their geographic extent remain limited. In this study, we quantify spatial and temporal drought patterns in TDFs across the Americas using the Standardized Precipitation Evapotranspiration Index (SPEI). We examined SPEI at 3-, 6-, and 12-month timescales to capture short-term moisture deficits and longer-term water stress affecting ecosystem functioning. Drought trends from 1950 to 2024 were analyzed, focusing on drought occurrences, their severity, duration, and spatial extent.Our results reveal heterogeneity in drought dynamics across the Neotropical dry forests. While some areas have experienced increases in drought severity and persistence in recent decades, longer SPEI timescales indicate an intensification of prolonged water deficits, particularly in regions with higher precipitation regimes. Overall, this study provides a continental-scale perspective on ecological drought in TDFs in the Neotropics. It highlights emerging hotspots of vulnerability and emphasizes the importance of incorporating evaporative demand into drought assessments. Our findings have direct implications for understanding ecosystem resilience, guiding conservation strategies, and predicting climate change impacts on tropical dry forest structure and function.

Land classification from satellite imagery is important for land management, environmental monitoring, and urban planning. Machine learning methods such as random forests and multilayer perceptrons have shown strong performance on multispectral data, while the Kolmogorov-Arnold network has emerged as an alternative architecture with compact model structures. This study evaluates the Kolmogorov-Arnold network for land classification using Landsat 8 imagery and compares it with random forest and multilayer perceptron models. The models were trained and tested on data from Edmonton, Alberta and evaluated on an independent dataset from Calgary, Alberta across five land classes: agriculture, urban, water, forest, and bare ground. For the Calgary dataset, the Kolmogorov-Arnold network matched the accuracy of the random forest and outperformed the multilayer perceptron, while requiring substantially fewer trainable parameters and providing greater interpretability.

Coastlines serve as dynamic interfaces between terrestrial and marine ecosystems. While advances in satellite remote sensing have promoted coastline monitoring, no comprehensive global coastline dataset derived from Sentinel-2 imagery has been produced, despite its 10-m resolution and frequent revisit capabilities. To address this, we propose S2Coast, a knowledge-based framework based on the Google Earth Engine (GEE) platform, designed to automatically detect the unified High Water Line (HWL) from annually composited Sentinel-2 imagery, termed HWLSentinel-2. This method integrates multi-temporal observational information, spectral characteristics, and spatial features to delineate the stable extent of high seawater submergence captured in cloud-free satellite images over a year. The boundaries in the resultant "Land-Water" binarization image represent HWLSentinel-2, derived through integrated threshold segmentation for three decision layers. Subsequently, raster-to-vector conversion and optimization steps were performed. Following the sequential execution of 12,275 sub-tasks, the resultant S2Coast-2023 dataset includes approximately 2.17 million kilometers of coastline for 2023, covering all continents and most islands larger than 100 m2, excluding Antarctica and remote polar islands. Global validation based on 1146 samples demonstrates that the developed tool, S2Coast, exhibits robust stability and universality, with average 88 % of sampled coastline segments falling within a 10-m buffer when comparing repeatedly generated coastlines from consecutive years (2021 to 2023). For positional accuracy, using 532 coastline samples with very high resolution image-based OpenStreetMap coastlines as reference data reveal an average positional deviation of-1.10 m (95 % CI:-2.06 to-0.15 m) and an average root mean square error (RMSE) of 17.40 m (95 % CI: 16.23 to 18.65 m). As the first global coastline dataset with 10-m resolution and a unified coastline indicator, it will serve as a crucial foundational resource for future coastal research.

Non-native and invasive species are among the leading causes of global biodiversity loss and could therefore compromise the recovery of native forests after disturbance, such as on abandoned agricultural lands. Here we evaluated how the relative density and richness of non-native woody species (NNS) change across secondary tropical forest succession, determined whether they vary between dry and moist forests and identified the underlying environmental and social drivers of these changes. We used data from 1,561 forest plots and 58 chronosequences from ten neotropical countries. We classified 3,735 woody species by origin and invasiveness. Our analyses and conclusions focus on NNS, whereas native (potentially) invasive groups were examined separately. NNS were widespread, occurring in 81% of the chronosequences and comprising 18% of dry and 41% of moist forest plots. We recorded 11 non-native invasive species, most of which were multifunctional trees associated with human activity. In early succession (the first 10-20 years), NNS reached high relative density and richness, accounting for 28% of stems and 22% of species in moist forests, and 9% of stems and species in dry forests. Both metrics declined considerably during the same period but were still present in late succession, mirroring the successional trajectory of native pioneer species, probably due to canopy closure and increased shading. Spatially, NNS richness increased with the Human Development Index. However, both density and richness were negatively affected by increasing surrounding forest cover, agricultural proximity and precipitation, while soil organic carbon generally favoured NNS retention. Our findings suggest that naturally regrowing forests and maintaining relatively intact forest landscapes provide nature-based solutions to control NNS, thereby protecting native biodiversity, ecosystem integrity and local livelihoods.

Terrestrial carbon flux dynamics are strongly influenced by climate variability, particularly in tropical dry forests (TDFs), which are drought-adapted ecosystems characterized by pronounced seasonality. However, the impacts of extreme climate events on carbon flux in TDFs remain poorly understood. In this study, we investigate the sensitivity of carbon flux to climate extremes and essential climate variables (ECVs) by integrating 11 years of net ecosystem exchange (NEE) data with 17 extreme climate indices and 8 ECVs, using a Random Forest model interpreted through Shapley Additive exPlanations (SHAP). We find that carbon flux dynamics in the SRNP-EMSS are governed primarily by ECVs rather than short-term extreme events. Results indicate that soil temperature (27.1 Carbon flux in TDFs is mainly driven by ECVs; extreme events have limited effect. Soil temperature, VPD, and moisture shape carbon flux thresholds and dynamics. RF, PDPs and SHAP reveal nonlinear responses and sink–source transitions. ENSO and seasonality regulate carbon flux dynamics in SRNP-EMSS. TDFs show resilience to short extremes but are vulnerable to long-term shifts.

Frequent droughts increasingly threaten ecosystem stability and agricultural production. Soil moisture is a key indicator of drought, but its spatial coverage remains limited. Remote sensing drought indices provide higher spatial resolution, yet their ability to reflect soil moisture variability has not been systematically assessed. This study evaluates the Vegetation Condition Index, Vegetation Water Index, and Temperature Condition Index by combining Pearson correlation analysis with a Copula-based conditional probability framework to assess their long-term and threshold-based relationships with soil moisture across multiple temporal scales in China. The Vegetation Condition Index shows the strongest correlation with soil moisture at the annual scale and remains dominant during spring, summer, and autumn at shorter time scales. Ecosystem-dependent patterns emerge in summer, with the Vegetation Water Index performing better in forests and the Temperature Condition Index in grasslands, reflecting differences in vegetation density and surface energy processes. The Copula-based analysis reveals a contrasting pattern: across regions, remote sensing indices are less likely to reach extreme or severe drought thresholds than to indicate general drought, suggesting weaker vegetation and temperature responses under extreme soil moisture deficits. Under soil moisture drought conditions, the Vegetation Condition Index shows the highest conditional drought probability in forested regions, whereas the Vegetation Water Index is more responsive in northern arid regions and grasslands, and the Temperature Condition Index shows clearer responses during the growing season. These results indicate that vegetation regulation and legacy effects can weaken synchronous responses during extreme droughts, and that dominant drought signals vary among ecosystems.

Gravel beaches serve as natural defenses against wave energy, storm surges, and coastal hazards. Understanding their sediment dynamics requires analyzing gravel grain-size and grain-shape parameters, yet traditional field sampling, laboratory analysis, and transport-direction determination methods have long presented challenges. To overcome these limitations, this study proposes an integrated technical framework that employs machine learning algorithms to derive sediment characteristics and transport directions from multisource uncrewed aerial vehicle (UAV) datasets. Using spatial and spectral UAV datasets collected at the Fengmenkou gravel beach (Nantian Island, China), the framework derives four key grain-size and grain-shape parameters: mean grain size, sorting coefficient, skewness, and roundness. Results demonstrate that oblique quadrat images collected in the field can be orthorectified for reliable digital analysis, yielding grain-size and grain-shape parameters within acceptable error margins (mean grain-size errors <0.4 Phi; sorting coefficient errors <0.3 Phi). The four derived parameters exhibit consistently high accuracy on validation datasets (R-2 >0.80, root mean square error [RMSE] <0.04), enabling high-resolution and spatially continuous characterization of gravel beach sediments. The resulting spatially continuous parameter fields overcome discrete sampling constraints and substantially reduce field costs, while providing improved input data for the Gao-Collins model. Enhanced data resolution and reduced edge effects extend the applicability of the model for grain-size trend analysis on gravel beaches. In addition, a UAV elevation-based approach was developed to infer sediment transport directions from surface elevation changes. This independent analysis shows potential for identifying net sediment movement and provides complementary support for grain-size trend interpretations. Collectively, this work enables spatially explicit and cost-effective characterization of gravel beach sediment dynamics, thereby improving the assessment of coastal processes using UAV data.

Abstract Tropical dryland ecosystems are highly biodiverse and fragmented and are experiencing significant anthropogenic and climatic changes. With increasing extremes in temperature and precipitation, coupled with significant alteration, these ecosystems are at greater risk of increased exposure and vulnerability to climatic change; however, little work has quantified the climatic shifts occurring within these ecosystems globally. Here, we aim to fill this gap by using the ERA-5 reanalysis and CHIRPS precipitation data to quantify changes in essential climatic variables in tropical drylands since 2000. Overall, we find that regional pressures differ, with tropical dry forests, savannas, and shrublands becoming hotter and drier in the Neotropics and parts of the Afrotropics and Australasia. By contrast, the tropical dry forests in the Indomalayan, Oceania, and Nearctic are experiencing hotter and wetter conditions. Globally, though, these ecosystems are experiencing more change than the global average, suggesting they may be approaching tipping points in their resilience, ultimately shrinking the area where they can survive.

Traditional Land Use and Cover Change (LUCC) studies have predominantly focused on attribute changes (e.g., forest area, forest change rates, forest types) with limited consideration of the spatial features embedded in dynamic processes. However, LUCC varies across spatial and temporal dimensions in the real world, and these spatial dynamic features are essential in LUCC modeling. Here, we used the spatiotemporal features of LUCC to reproduce and forecast tropical forest changes from 1979 to 2100 in the Guanacaste region using a proposed Cellular Automata-Agent Based Model (CA-AB). The pilot model was validated against historical forest change scenarios. Additionally, this model was used to forecast future scenarios under different assumptions: current trend scenarios, economy-development-driven scenarios, and ecology-protection-driven scenarios. Our results demonstrated that historical simulations for the second period (1997-2015) were more accurate than those for the first period (1979-1997), likely because the extent of forest change (both loss and gain) was greater in the first period, increasing the likelihood of forest changes in random locations. In addition, the simulations of the CA-AB model more closely match the actual forest cover in the second period compared to the CA model alone. This improvement is primarily due to the inclusion of spatial and temporal disparities, particularly those driven by agents' decision behaviors. Besides, simulated forest areas are largest under ecology-protection-driven scenarios, followed by current trend scenarios, with economy-development-driven scenarios having the smallest forest areas. Notably, all three scenarios share a common feature: forest cover concentrates in the peripheral areas of the Guanacaste region, while it declines in the central area over time due to land resource limitations, population growth, and other potential factors.

This study introduces an Entropy-based index: the Lorenz-entropy (LE) index, which we have developed by integrating Light Detection And Ranging (LiDAR), econometrics, and forest ecology. The main goal of the LE is to bridge the gap between theoretical entropy concepts and their practical applications in monitoring vertical structural complexity of tropical forest ecosystems. The LE index quantifies entropy by analyzing Relative Height (RH) metrics (representing a one-dimensional (1D) canopy structure metric) distributions from full-waveform LiDAR across successional stages in a tropical dry forest (TDF) and a tropical rainforest. To validate the LE trends derived from LiDAR, we extended the analysis using inventory-based two-dimensional (2D) and three-dimensional (3D) metrics, specifically basal area and biomass. The consistency of trends between the 1D LiDAR-derived LE and the inventory-based 2D and 3D metrics reinforces the LE's ability to capture and monitor structural complexity reliably across different measurement dimensions. Our findings demonstrated that LE captures the changes in entropy as a function of successional stages, reflecting how canopy structure evolves towards homogeneity and complexity. Our statistical analysis revealed significant differences between successional stages (ANOVA, alpha = 0.05, p < 2e-16), with LE increasing substantially from early to late stages and plateauing at climax, where vertical structure (entropy) stabilizes. The mean LE increased by 1.70x10(-2) between late and climax stages, with a small effect size (Cohen's d = 0.25), indicating minor differences in complexity. The LE index, calculated from biomass and basal area, confirming that as forests mature, entropy and vertical structural complexity increase. Furthermore, the sensitivity analysis showed that LE is most responsive to RHs variability during intermediate stages, suggesting that structural development is most dynamic during this phase. These results demonstrate the potential of the LE index as a tool for ecological analysis and monitoring forest dynamics.

Tropical dry forests (TDFs) are sensitive ecosystems projected to experience significant warming due to global climate change, potentially disrupting their ecological functions. Accurate and low-uncertainty climate projections are critical for understanding monthly temperature trends in these regions. This study employs NASA’s Earth Exchange Global Daily Downscaled Projections (NEX-GDDP) based on the Coupled Model Intercomparison Project Phase 6 (CMIP6) to analyze monthly mean air temperature changes in a TDF across historical (1960–2014), near-term (2015–2040), mid-term (2040–2060), and far-term (2080–2100) periods. We conduct this analysis under three shared socio-economic pathways (SSP1-2.6, SSP3-7.0, and SSP5-8.5). We identified statistically significant positive temperature trends for historical and projected periods (α = 0.05, p < 0.001). SSP5-8.5 exhibited the steepest increase, with a slope of 85.8 × 10⁻⁶ °C/month across all terms (2015–2100). Monthly results show projected air temperature increases of 1.5 ̊ ± 0.9 °C (5.1

Terrestrial ecosystems are crucial in mitigating global climate change, and dynamic global vegetation models (DGVMs) have become essential tools for simulating these ecosystems. However, uncertainties remain in DGVM simulations for China, highlighting the need for systematic evaluations of their dynamics across various timescales to enhance model performance. As such, we utilize reprocessed monthly MODIS leaf area index (LAI) and contiguous solar-induced fluorescence (CSIF) data as observational references to assess the long-term trends and seasonal variations of LAI and gross primary production (GPP) simulated by 14 models (CABLE-POP, CLASSIC, CLM5.0, DLEM, IBIS, ISAM, ISBA-CTRIP, JULES, LPJ-GUESS, LPX, OCN, ORCHIDEEv3, SDGVM, and VISIT) in China from 2003 to 2019. Additionally, we evaluate the trends and seasonal variations of simulated LAI and GPP in response to environmental and climatic factors. Our findings indicate the following. (1) While the overall trend of simulated LAI is captured, the spatial performance of simulated LAI and GPP is poor, with underestimation in forested areas, overestimation in grasslands, and misestimation in croplands. (2) The models misestimate the simulated LAI and GPP responses to changes in environmental factors, as well as their inaccuracy in capturing anthropogenic impacts on vegetation dynamics. We indicate that the main reason for the model's misestimation is that the model's representation of the CO2 fertilization effect is inadequate and thus fails to simulate the vegetation response to CO2 concentration. (3) Despite these issues, the models can effectively capture the seasonality of LAI and GPP in China, largely due to their robust representation of seasonal responses to climate factors.

Lianas, woody vines acting as structural parasites of trees, have profound effects on the composition and structure of tropical forests, impacting tree growth, mortality, and forest succession. Remote sensing could offer a powerful tool for quantifying the scale of liana infestation, provided the availability of robust detection methods. We analyze the consistency and global geographic specificity of spectral signals-reflectance across wavelengths-from liana-infested tree crowns and forest stands, examining the underlying mechanisms of these signals. We compiled a uniquely comprehensive database, including leaf reflectance spectra from 5424 leaves, fine-scale airborne reflectance data from 999 liana-infested canopies, and coarse-scale satellite reflectance data covering 775 ha of liana-infested forest stands. To unravel the mechanisms of the liana spectral signal, we applied mechanistic radiative transfer models across scales, establishing a synthesis of the relative importance of different mechanisms, which we corroborate with field data on liana leaf chemistry and canopy structure. We find a consistent liana spectral signal at canopy and stand scales across globally distributed sites. This signature mainly arises at the canopy level due to direct effects of more horizontal leaf angles, resulting in a larger projected leaf area, and indirect effects from increased light scattering in the near and short-wave infrared regions, linked to lianas' less costly leaf construction compared with trees on average. The existence of a consistent global spectral signal for lianas suggests that large-scale quantification of liana infestation is feasible. However, because the traits responsible for the liana canopy-reflectance signal are not exclusive to lianas, accurate large-scale detection requires rigorously validated remote sensing methods. Our models highlight challenges in automated detection, such as potential misidentification due to leaf phenology, tree life history, topography, and climate, especially where the scale of liana infestation is less than a single remote sensing pixel. The observed cross-site patterns also prompt ecological questions about lianas' adaptive similarities in optical traits across environments, indicating possible convergent evolution due to shared constraints on leaf biochemical and structural traits.

Analyzing the vertical structural complexity of tropical forests is essential for understanding their ecological functions and biodiversity. Given this significance, an indicator that quantifies entropy, representing heterogeneity and disorder in structural complexity, plays a significant role in forest ecological studies. This study explored the potential of the Lorenz-entropy (LE) index as an innovative metric for classifying tropical forest types. Using spaceborne LiDAR data from the Global Ecosystem Dynamics Investigation (GEDI) mission from April 2019 to March 2023, we integrated the LE index with supervised machine learning algorithms to evaluate its effectiveness in distinguishing vertical structural complexity across the three tropical forest ecosystems. In addition to the LE index, forest structural variables such as Above Ground Biomass Density (AGBD), Plant Area Index (PAI), and Relative Height 98 (RH98) were included. The results revealed that incorporating the LE index into remote sensing-based forest monitoring frameworks can significantly improve the classification of tropical forest types. Paired t-tests confirmed statistically significant improvements (p < 0.05) in the classification metrics with substantial effect sizes measured using Cohen's d. Moreover, ensemble methods, mainly Random Forest (RF) and Gradient Boosting classifiers, exhibited the highest accuracy, with RF showing 90 +/- 2 % overall accuracy and XGBoost 90 +/- 2.5 % upon incorporating the LE index. These findings highlight the utility of the LE index as a metric of vertical structural complexity and underscore its value in improving tropical forest discrimination using GEDI data. Future research should explore integrating additional remote sensing data to refine the application of the LE index for forest classification.

Lianas play a crucial role in shaping the vertical structural complexity of tropical forests; however, their impact is not well understood. This study evaluates the influence of lianas on the Lorenz-entropy (LE) index, a measure of canopy heterogeneity, in a Neotropical tropical dry forest (TDF) in Santa Rosa National Park, Costa Rica. Using full-waveform LiDAR data and simulated Global Ecosystem Dynamic Investigation (GEDI) waveforms, we compared liana-infested and non-infested plots across early, intermediate, and late successional stages. Using Mann-Whitney U test with Benjamini-Hochberg (BH) correction (95 % confidence level) our findings indicate that liana-infested plots exhibit significantly higher Lorenz-entropy index values in intermediate (P< 0.05, P < BH-threshold), and late (P < 3.33 x 10(-2), P < BH-threshold) successional stages, while early-stage in noninfested plots showed significantly higher entropy values (P< 1.67 x 10(-2), P < BH-threshold). Effect size analysis showed a moderate impact in the intermediate stage (Cliff's delta = 0.35, 95 % CI: 0.07-0.61) and a moderate to large impact in the late stage (Cliff's delta = 0.41, 95 % CI: 0.11-0.67). In early-stage plots, lianainfested stands had significantly lower LE index values than non-infested plots (Cliff's delta = -0.50, 95 % CI: -0.78 to -0.19). These results demonstrate that the LE index effectively captures liana-driven increases in vertical canopy stratification and heterogeneity, particularly in more mature forest stages. The integration of airborne LiDAR and GEDI simulations offers an approach for assessing structural complexity at the plot level. These findings highlight the need for further research to understand the long-term ecological consequences of liana abundance, in the context of forest monitoring.

Long-term eddy covariance (EC) data are crucial for understanding the impact of global change on ecosystem functions. However, EC data often contain long gaps, particularly in tropical dry forests (TDF) due to seasonality and El Ni & ntilde;o-Southern Oscillation (ENSO) phases. These factors create high variability, complex dependencies, and dynamic flux footprints. No current gap-filling method adequately addresses long gaps in TDFs. This study introduces a novel framework for addressing this issue by (a) defining gap sizes by their relative percentages, (b) training, tuning, and evaluating two machine learning (ML) models: MissForest for short gaps and Prophet for intermediate and long gaps, and (c) predicting half-hourly EC data from 2013 to 2022 for six EC variables, where actual gap data sets ranged from 26.6% to 28.4%, at TDF in Costa Rica. Results indicate that MissForest excelled at filling short gaps (<= 5%, R2 = 0.76 and Nash-Sutcliffe efficiency (NSE) = 0.71), while Prophet performed exceptionally well for gaps between 5% and 10% (R2 = 0.72 and NSE = 0.67). However, both models struggled with gaps between 10% and 13%. Validation showed R2 values of 0.79, 0.88, and 0.77 for CO2 flux, sensible heat flux, and latent heat flux, respectively, with corresponding NSE values of 0.78, 0.86, and 0.72, and normalized root mean squared error (NRMSE) around 2E-4. Additionally, to validate our results, we applied our approach at three EC sites with different ecological conditions, demonstrating robust performance. This study presents a reliable ML approach for imputing long gaps in EC data, which can be applied to sites with strong variability.

According to the Paris Climate Change Agreement, all nations are required to submit reports on their greenhouse gas emissions and absorption every two years by 2024. Consequently, forests play a crucial role in reducing carbon emissions, which is essential for meeting these obligations. Recognizing the significance of forest conservation in the global battle against climate change, Article 5 of the Paris Agreement emphasizes the need for high-quality forest data. This study focuses on enhancing methods for mapping aboveground biomass in tropical dry forests. Tropical dry forests are considered one of the least understood tropical forest environments; therefore, there is a need for accurate approaches to estimate carbon pools. We employ a comparative analysis of AGB estimates, utilizing different discrete and full-waveform laser scanning datasets in conjunction with Ordinary Least Squares and Bayesian approaches SVM. Airborne Laser Scanning, Unmanned Laser Scanning, and Space Laser Scanning were used as independent variables for extracting forest metrics. Variable selection, SVM regression tuning, and cross-validation via a machine-learning approach were applied to account for overfitting and underfitting. The results indicate that six key variables primarily related to tree height: Elev.minimum, Elev.L3, lev.MAD.mode, Elev.mode, Elev.MAD.median, and Elev.skewness, are important for AGB estimation using ALSD and ULSD , while Leaf Area Index, canopy coverage and height, terrain elevation, and full-waveform signal energy emerged as the most vital variables. AGB values estimated from ten permanent tropical dry forest plots in Costa Rica Guanacaste province ranged from 26.02 Mg/ha to 175.43 Mg/ha . The SVM regressions demonstrated a 17.89 error across all laser scanning systems, with SLSF W exhibiting the lowest error 17.07 in estimating total biomass per plot.