Summary: Leaf traits are essential for understanding many physiological and ecological processes. Partial least squares regression (PLSR) models with leaf spectroscopy are widely applied for trait estimation, but their transferability across space, time, and plant functional types (PFTs) remains unclear. We compiled a novel dataset of paired leaf traits and spectra, with 47 393 records for > 700 species and eight PFTs at 101 globally distributed locations across multiple seasons. Using this dataset, we conducted an unprecedented comprehensive analysis to assess the transferability of PLSR models in estimating leaf traits. While PLSR models demonstrate commendable performance in predicting chlorophyll content, carotenoid, leaf water, and leaf mass per area prediction within their training data space, their efficacy diminishes when extrapolating to new contexts. Specifically, extrapolating to locations, seasons, and PFTs beyond the training data leads to reduced R² (0.12–0.49, 0.15–0.42, and 0.25–0.56) and increased NRMSE (3.58–18.24%, 6.27–11.55%, and 7.0–33.12%) compared with nonspatial random cross‐validation. The results underscore the importance of incorporating greater spectral diversity in model training to boost its transferability. These findings highlight potential errors in estimating leaf traits across large spatial domains, diverse PFTs, and time due to biased validation schemes, and provide guidance for future field sampling strategies and remote sensing applications.

Rapidly and accurately estimating yields at field scales are very significant. Each type of currently yield estimation model has been well studied, yet all of them have certain limitations. Based on a coupled Carnegie-Ames-Stanford approach (CASA)-World Food Studies (WOFOST) model and time series Sentinel-2 imagery, we achieved daily crop simulations and crop yield estimations at two adjacent farms in China. The results indicated that the coupled model inherited the high computing speed of light use efficiency (LUE) models and the mechanistic advantages of crop growth models. The ${R}^{2}$ of yield simulation was 0.64 for the CASA model, 0.84 for the coupled model, and 0.86 for the WOFOST model, and the root mean square error (RMSE) values were 948.32, 792.11, and 623.64 kg/ha, respectively. The operating times of the CASA model, CASA-WOFOST coupled model, and WOFOST model over the growing season of wheat were 37 min, 48 min, and 1 day 5 h 7 min, respectively. Compared with the WOFOST model, the coupled model provided a much faster running speed in yield simulations and a similar accuracy; therefore, the proposed model can be applied for assessments at large farms with high-spatial-resolution images to obtain accurate yield simulations. Compared with the CASA model, the coupled model provided higher simulation accuracy in mountainous areas and regions of uneven terrain. It can be concluded that the proposed coupled CASA-WOFOST model can improve the precision, reliability, and stability of crop yield estimation; provide theoretical support for crop yield estimation at field scales; and promote the development of precision agriculture.

More frequent and widespread large fires are occurring in the western United States (US), yet reliable methods for predicting these fires, particularly with extended lead times and a high spatial resolution, remain challenging. In this study, we proposed an interpretable and accurate hybrid machine learning (ML) model, that explicitly represented the controls of fuel flammability, fuel availability, and human suppression effects on fires. The model demonstrated notable accuracy with a F₁‐score of 0.846 ± 0.012, surpassing process‐driven fire danger indices and four commonly used ML models by up to 40% and 9%, respectively. More importantly, the ML model showed remarkably higher interpretability relative to other ML models. Specifically, by demystifying the “black box” of each ML model using the explainable AI techniques, we identified substantial structural differences across ML fire models, even among those with similar accuracy. The relationships between fires and their drivers, identified by our model, were aligned closer with established fire physical principles. The ML structural discrepancy led to diverse fire predictions and our model predictions exhibited greater consistency with actual fire occurrence. With the highly interpretable and accurate model, we revealed the strong compound effects from multiple climate variables related to evaporative demand, energy release component, temperature, and wind speed, on the dynamics of large fires and megafires in the western US. Our findings highlight the importance of assessing the structural integrity of models in addition to their accuracy. They also underscore the critical need to address the rise in compound climate extremes linked to large wildfires.

Shoreline change detection from remote sensing imagery remains challenging in environments subject to water level fluctuations, as remotely sensed shoreline positions reflect instantaneous hydrodynamic states rather than true geomorphic change. In the Great Lakes, seasonal and short-term water level variations can produce apparent shoreline shifts unrelated to sediment dynamics. Reliable calibration with bathymetry and water level data can mitigate this effect, but such data are often unavailable or difficult to obtain for many coastal and lacustrine systems worldwide. To address this limitation, we proposed a morphology-based framework that quantifies geometric change between successive shoreline curves using a discrete Fréchet distance, a modified Euclidean distance and a Union distance metric. Rather than relying solely on cross-shore displacements, the approach leverages shape similarity to differentiate water-level-driven shifts from true morphological change. We evaluated the framework across three spatial scales (100 m, 500 m, and 1000 m) along 125 km of southwestern Lake Michigan coastline using 2010 and 2020 aerial imagery, benchmarking against water-level-calibrated DSAS erosion hotspots. The Fréchet distance improved monotonically with scale, achieving strong agreement at 1000 m (F1 = 0.84, Spearman $\rho$ = 0.79) but limited reliability at 100 m. While individual morphology-based metrics appeared competitive with or inferior to uncalibrated DSAS at each scale, the union of both distances substantially outperformed uncalibrated DSAS at management-relevant scales (F1 of 0.64 vs. 0.50 at 500 m and 0.79 vs. 0.42 at 1000 m), reflecting the complementary nature of shape-based and displacement-based detection. The Patient Rule Induction Method (PRIM) further identified gentle nearshore slopes and moderate separation from engineered structures as the geomorphic conditions under which the morphology-based and calibrated erosion indicators converged most closely (in-box F1 = 0.92 at 1000 m and 0.72 at 500 m). These results suggest that the proposed framework, particularly the complementary union of both metrics, provides a practical, calibration-free alternative for multiscale shoreline change screening in lacustrine and microtidal, data-limited environments, while local-scale applications still benefit from explicit water-level correction.

Accurate shoreline positioning is critical for coastal monitoring and management, yet deep learning shoreline products are often evaluated using conventional waterbody segmentation metrics that do not explicitly measure boundary alignment. Using 20,689 NAIP aerial images covering the Great Lakes shoreline from the Coastal Aerial Imagery Dataset (CAID), we benchmark five semantic segmentation models and quantify the inconsistency between image-level segmentation accuracy (pixel accuracy, IoU) and shoreline positioning accuracy measured by the Shoreline Intersection Ratio (SIR) and Average Eulerian Distance (AED). Although segmentation performance is consistently high (pixel accuracy typically >98% and IoU often >90%), shoreline agreement is substantially lower and strongly landscape-dependent, with the poorest results in wetlands and urban scenes. Correlation analyses across coastal types and water-surface conditions show that the correspondence between segmentation metrics and SIR varies with shoreline morphology. Multivariate regressions confirm the shoreline-to-water ratio (SWR) as the dominant predictor of both SIR and AED, while shoreline complexity (SCI) and mean water hue (MWH) have weaker, context-dependent effects. These results demonstrate that high segmentation accuracy does not guarantee precise shoreline delineation and motivate shoreline-aware evaluation protocols.

Seventy percent of global forests lie within 1 km of an edge, yet their temporal dynamics remain poorly characterized. By developing the high spatial-resolution (0.00025°) global forest edge maps for 2000-2020, we find a 24 million km (7%) decline in edge length. During 2015-2020, edge length falls while forest area slightly grows, driven by patch merging in temperate/boreal regions and tropical fragmentation. In contrast, countries such as Laos, Liberia, India, and China gain edges from new forest patches. These episodic changes closely track policy-driven land management, including logging bans, agricultural expansion, and reforestation. Given forest edges’ critical roles in biodiversity and microclimate regulation, our results highlight contrasting biome patterns and the need for adaptive policies balancing conservation, carbon storage, and land use.