Understanding how crop trait variability shapes genotype × environment × management (G × E × M) interactions remains a key uncertainty in predicting agricultural performance under a changing climate. Continental-scale crop models commonly rely on spatially uniform parameters, limiting their ability to represent adaptive variation in phenology, allocation, and yield formation. Here we integrate the mechanistic agroecosystem model Ecosys with a deep learning-enabled model-data fusion inversion to infer spatially explicit physiological controls (trait proxies) of U.S. winter wheat directly from observations. By constraining simulations with satellite-derived photosynthesis and county-level yield records from 2008 to 2022 across ~1000 winter wheat-producing counties, the inversion recovers coherent patterns of maturity group, reproductive capacity, harvest index, and root-shoot allocation. The optimized simulations reproduce observed carbon uptake and yield variability (gross primary productivity r = 0.76-0.88; phenology bias < 2 weeks; > 90% of yields within ±20% of reports) and reveal distinct physiological profiles that align with the geographic distributions of major winter wheat market classes. The inferred controls explain class- and region-specific climate sensitivities: warmer winters reduce vernalization success in late-maturing cultivars, while elevated vapor pressure deficit causes strong yield losses in rainfed Hard Red Winter wheat. The results demonstrate that observation-constrained trait inversion within model-data fusion framework reveals biologically meaningful crop-class variation, thereby providing a scalable, physiologically grounded framework for diagnosing adaptive diversity and climate vulnerability across agroecosystems.

Quantifying carbon outcomes from agroecosystems plays an important role in mitigating global warming and ensuring food security through sustainable production. However, high spatial-temporal-resolution (e.g., ~100m, daily), accurate, well-resolved carbon budgets and crop yield in agroecosystems are extremely challenging to quantify due to the complexity of involved processes and large variations in environmental and management drivers. Traditional process-based-modeling approaches are computationally expensive to achieve field-scale resolution and contain large uncertainty due to underdetermined model structure and parameters. Knowledge-guided machine learning (KGML) is a hybrid modeling approach that leverages recent advances in machine learning combined with known physical principles and relationships to enhance the training and application processes, which helps open the “black box” of conventional ML models, and enable better predictions that capture variability in both time and space. Here we proposed a data-efficient KGML framework that effectively predicts daily variations in agricultural CO2 emissions, crop yields, and soil carbon storage at field scale, as successfully demonstrated for the US Midwest. Multi-source data and pretraining with outputs from a well-validated agroecosystem model were incorporated into a hierarchically structured deep learning neural network that greatly outperformed both process-based and pure machine learning models, especially in data-limited cases. This work demonstrates the advantages of integrating domain knowledge with state-of-the-art artificial intelligence in agroecosystem modeling that will lead toward broader use of KGML in geoscience.

Global warming significantly impacts forest growth. However, commonly used spatially interpolated gridded air temperature datasets may not fully capture these effects due to their coarse spatial resolution and because air temperature may not accurately reflect the conditions that influence the tree growth process. Although finer spatial resolution land surface temperature (LST) datasets may capture more detailed temperature variations, their potential to assess forest growth responses to global warming has not been thoroughly explored. We evaluated the performance of air temperature and LST datasets with various spatial resolutions, including Climatic Research Unit gridded Time Series (CRU), TerraClimate, the land component of the fifth-generation European ReAnalysis (ERA5-Land), and MODIS LST (MOD11A2), in capturing the relationships between tree radial growth and temperature variations across 555 sites in the Northern Hemisphere. Our results showed that the finer spatial resolution MOD11A2 significantly outperformed the widely used CRU air temperature in modeling tree radial growth, with mean and maximum temperatures increasing the coefficient of determination (R2) by 16.32 % and 18.14 %, respectively. This improvement was especially apparent in high-elevation areas where R2 increased by 35.70 % and 36.97 %. We suggested that commonly used spatially interpolated gridded air temperature datasets (e.g., CRU and TerraClimate) may underestimate the impact of rising temperatures on forest growth. Our findings highlight the necessity of integrating high-resolution LST to accurately model forest growth responses to global warming.

Arid and semiarid forests are vulnerable to climate change. Whether rising temperatures would increase wood production by extending the growing season in these relatively dry forests remains unclear. Here, we document a decoupling of duration versus the amount of wood production. Despite the prolonged duration of xylem cell production induced by warming, we did not observe a consistent increase in the amount of xylem growth, based on a two‐year (2018–2019) monitoring of wood formation in Siberian spruce ( Picea obovata Ledeb.) along a natural hydrothermal gradient in the southern Altai Mountains of central Asia. The duration of cell production ranged from 43 to 93 days, with the total number of xylem cells between 37 and 185. Notably, the warmest Altai low‐altitude site (AL) had the longest duration (82.6 ± 13.3 days) but the lowest number of xylem cells (63.88 ± 15.6 cells) in 2018, indicating a decoupling between growing season length and xylem growth. However, this decoupling weakened in 2019 when spring temperatures were cooler and summer precipitation was higher than in 2018. Warmer spring temperatures can extend the duration of wood cell production by triggering an earlier onset of cambium activity. Additionally, sufficient summer precipitation provides the water necessary for turgor‐driven cell division and expansion, enhancing the cell production rate, which plays a dominant role in determining wood growth in arid and semiarid forests. Thus, warmer spring temperatures may amplify the observed decoupling, whereas sufficient summer precipitation may narrow it. Synthesis . This decoupling highlights the critical role of seasonal climates in regulating wood formation dynamics in arid and semiarid forests. With ongoing climate change, longer and warmer growing seasons may amplify the importance of water availability, further constraining forest growth and carbon sequestration when growing‐season water conditions are unfavourable.

Non-continuous flooding (NCF) in rice was recently reported to improve the field-scale, seasonal carbon balance. In this response article, we clarify our system boundaries (seasonal, field-scale), address the role of yield carbon and microbial indicators, and add sensitivity checks with expanded data. Across these checks, NCF consistently reduces methane without penalizing yield, supporting our original conclusions.

Tree-ring widths contain valuable historical information related to both forest disturbances and climate variability and changes within forests. However, current methods are still unable to accurately distinguish between disturbances and climate signals in tree rings, especially in the case of climate anomalies. To address this issue, we developed a novel method, called Growth Trends Clustering (GTC) that uses the distribution characteristics of tree-ring widths within a stand to distinguish the effects of climate and other forest disturbances. GTC employed a Gaussian mixture model to fit the probability density distribution of annual ring-width index (RWI) in a stand. Discriminative criteria were established to cluster diverse sub-distributions from the Gaussian mixture model into categories of growth release, suppression, or normal trends. This approach allowed us to identify the occurrence, duration, and severity of forest disturbances based on percentage changes in the growth release or suppression categories of trees. And the effect of climate on tree growth was assessed according to the mean statistics of the growth normal categories. Using common forest disturbances such as defoliating insects and thinning as examples, we validated our method using tree-ring collections from six sites in British Columbia and Quebec, Canada. We found that the GTC method was superior to traditional time-series analysis methods (e.g., Radial Growth Averaging, Boundary Line, Absolute Increase, and Curve Intervention Detection) for detecting past forest disturbances and was able to significantly enhance climate signals. In summary, the GTC method presented in this study introduces a novel statistical approach for accurately distinguishing between forest disturbances and climate signals in tree rings. This is particularly important for understanding forest disturbance regimes under climate change and for developing future disturbance mitigation strategies.

Annual tree rings are widely recognized as valuable tools for quantifying and reconstructing historical forest disturbances. However, the influence of climate can complicate the detection of disturbance signals, leading to limited accuracy in existing methods. In this study, we propose a random under-sampling boosting (RUB) classifier that integrates both tree-ring and climate variables to enhance the detection of forest insect outbreaks. The study focused on 32 sites in Alberta, Canada, which documented insect outbreaks from 1939 to 2010. Through thorough feature engineering, model development, and tenfold cross-validation, multiple machine learning (ML) models were constructed. These models used ring width indices (RWIs) and climate variables within an 11-year window as input features, with outbreak and non-outbreak occurrences as the corresponding output variables. Our results reveal that the RUB model consistently demonstrated superior overall performance and stability, with an accuracy of 88.1%, which surpassed that of the other ML models. In addition, the relative importance of the feature variables followed the order RWIs > mean maximum temperature (T-max) from May to July > mean total precipitation (P-mean) in July > mean minimum temperature (T-min) in October. More importantly, the dfoliatR (an R package for detecting insect defoliation) and curve intervention detection methods were inferior to the RUB model. Our findings underscore that integrating tree-ring width and climate variables as predictors in machine learning offers a promising avenue for enhancing the accuracy of detecting forest insect outbreaks.

Abstract. Tile drainage removes excess water and is an essential, widely adopted management practice to enhance crop productivity in the U.S. Midwest. Tile drainage has been shown to significantly change hydrological and biogeochemical cycles by lowering the water table and reducing the residence time of soil water, although such impacts and their connections are poorly understood and highly uncertain. Understanding these impacts is essential, particularly so because tile drainage has been highlighted as an adaptation under projected wetter springs and drier summers in the changing climate in the U.S. Midwest. We used the ecosys model, uniquely incorporating soil oxygen dynamics and crop oxygen uptake, to quantify the impacts of tile drainage on hydrological and biogeochemical cycles and crop growth at corn-soybean rotation fields. Tiles are represented as a water sink in the soil, characterized by tile depth and spacing in ecosys. Water flow from saturated soil layers to tiles is governed by the lateral hydraulic gradient defined by the water table depth in the field, tile depth, and tile spacing. The model was validated with data from a multi-treatment, multi-year experiment in Washington, IA. The relative root mean square error (rRMSE) for corn and soybean yield in validation is 5.66 % and 12.57 %, respectively. The Pearson coefficient (r) of the monthly tile flow during the growing season is 0.78. Model results show that tile drainage reduces soil water content and enhances soil oxygenation. It additionally increases subsurface discharge and elevates inorganic nitrogen leaching, with seasonal variations influenced by climate and crop phenology. The improved aerobic condition alleviated crop oxygen stress during wet springs, thereby promoting crop root growth during the early growth stage. The development of greater root density, in turn, mitigated water stress during dry summers, leading to an overall increase in crop yield by ~6 %. These functions indicate the potential of tile drainage in bolstering crop resilience to climate change, and the use of this modeling tool for large-scale assessments of tile drainage. The model reveals the inherent connections of tile drainage’s impacts on hydrology, soil biogeochemistry, and plant growth.

The recent increase of regional wildfire occurrences has been associated with climate change. In this study, we investigated the association between the February to March wildfire points and burned area in the southern region of China (20°N–30°N and 105°E–115°E) and the simultaneous Arctic Oscillation (AO) index during 2001–2022 and 2001–2020, respectively. After removing the El Niño-Southern Oscillation and Indian Ocean Dipole signals, time series of the regional mean fire points and burned area over the study area is significantly correlated with the AO index at − 0.37 and − 0.47, significant at the 0.1 level. Precipitation significantly affects wildfire variations. The positive AO could trigger a southeastward Rossby wave train and induce anomalous cyclone activity approximately located in the area encompassed by 15°N–27°N and 85°E–100°E. This outcome could help to enhance the southern branch trough and results in positive precipitation anomalies in southern China. This increasing moisture is conductive to reducing wildfire risks, vice versa. Our results are potentially useful for strengthening the understanding of the mechanisms of wildfire occurrences in southern China.

Climate warming and subsequent drought are predicted to alter local forest production and carbon budgets, the sensitivity of which may be site- and species-specific. Although a warmer and drier climate often constrains tree growth, whether trees grown in cool, wet regions across the Siberian forest can in turn be promoted remains unknown. Here, we aimed to investigate the historical growth tendency of Siberian forests in the Sayan Mountain areas and to explore how climate interacts to regulate their growth. We used ring-width data from Siberian pine (Pinus sibirica Du Tour) sampled from three sites in this area to establish a regional chronology and calculate percentage growth change (%GC) over the past 250 years. Bootstrapped correlation analysis between the regional ring-width chronology and climatic factors indicates the mean air temperature, though not precipitation, is more often positively linked to the radial growth of Siberian pines. The %GC series shows that, from 1966 to 2006, the rising temperatures in May resulted in a significant increase in the radial-growth rate of Siberian pines (r = 0.47, p < 0.05). Our study suggests that the positive growth rate anomaly is more likely to occur as temperatures rise in Southern Siberia.

Whether the tree growth–climate relationship is consistent in subtropical China has not yet been reported. To fill this gap, we chose Pinus taiwanensis which grow on Lushan Mountain in a subtropical region of China as the target tree species, established a standard tree-ring width chronology, and conducted a moving correlation analysis with climatic factors. The results showed that the relationship between radial growth of P. taiwanensis and climate changed significantly during 1980–1990. From 1955 to 1985, tree rings were negatively affected mainly by precipitation in September of the current growing season. From 1990 to 2014, however, a significant negative correlation appeared between tree rings and sunshine duration from March to April in the growing season. Our results suggest the need to pay attention to this growth–climate inconsistency when conducting dendroclimatology studies in subtropical China. However, the causes of the inconsistency still require further confirmation.

Fire is a principal disturbance agent in Northeastern Asia forests with the highest fire radiative power globally. The limitation of the wide network of sites with long reconstructed fire histories restricts our understanding of factors forcing regional fire dynamics. We aggregated 30 fire scars-based fire chronologies in Northeastern Asia to create a composite region chronology from 1701 to 2010. We assessed the correlation between solar activity, drought (as represented by the Self-Calibrated Palmer Drought Severity Index, scPDSI), Arctic Oscillation (AO), and forest fire. Large fires were associated with drought, positive AO, and lower sunspot numbers, as revealed by Superposed epoch analysis. The pathways of solar activity force drought and AO, hence, driving fire activity indicated by structural equation modeling. As Solar Cycle 25 (starting in December 2019 and lasting until about 2030) is projected to be as weak as the Dalton minimum (1795–1820 AD), we expect fire activity to remain high over this period. Our findings provide insights into the extraterrestrial forcing of forest fire activity and its prediction in Northeastern Asia.

Semanticsentence matching is a crucial task of natural language processing. However, semantic sentence matching is mainly used in text domain. For video clip and mixing, it explored less. Existing methods mainly focus on mapping text and video into latent spaces in video clip and mixing, but their extractor lack the ability to get effective information. So, we present a M ulti F eature F usion semantic sentence matching model (MFF), which forms the double filtering. The double filtering is designed for filtering to the similar semantic fragments in video clip and mixing, reducing the burden of heavy manual video editing. Experiments are conducted on two datasets, namely, SNLI and Quora Question Pairs, to verify that MFF can significantly improve the accuracy. Results show that MMF improves the performance of SNLI and Quora Question Pairs datasets to 75.3% and 76.7% (accuracy), respectively.

Fire is a major disturbance agent in the boreal forest, affecting the structure, dynamics and biogeochemical cycles in this biome. In the Asian section of boreal forest, the records of long-term fire history are few that limits our understanding of factors forcing regional fire dynamics. We presented an annually-resolved 352 year (1666-2017) fire chronology based on fire scars of Scots pine (Pines sylvestris L.) and Siberian larch (Larix sibirica Ledeb) from the Transbaikal area in the southeastern Siberia. Fire activity showed an increasing trend from 1720 to 1929 (R-2 = 0.80, P < 0.0001), and a significant decreasing trend from 1920 to 2010 (R-2 = 0.62, P < 0.001). We assessed the potential relationships between drought (as represented by the Palmer Drought Severity Index, PDSI, and the Monthly Drought Code, MDC), ocean-atmosphere circulation and forest fire by Superposed epoch analyses, cross-wavelet analysis and Granger causality analysis. Increased fire activity was associated with stronger drought from previous winter to current summer of fire event years and positive Arctic Oscillation (AO) before and during major fire season (February and April to May), as revealed by superposed epoch analysis. Granger causality pointed to the significant role of drought in driving forest fires. Our findings provide insights into the climate drivers of forest fire activity and its prediction in the Transbaikal region.

The high productivity in the US Corn Belt is largely enabled by the consumption of millions of tons of manufactured fertilizer. Excessive application of nitrogen (N) fertilizer has been pervasive in this region, and the unrecovered N eventually escaped from croplands in forms of nitrous oxide (N 2 O) emission and N leaching. Mitigating these negative impacts is hindered by a lack of practical information on where to focus and how much mitigation potential to expect. At a large scale, process-based crop models are the primary tools for predicting variables required by decision making, but their applications are prohibited by expensive computational and data storage costs. To overcome these challenges, we built a series of metamodels to learn the key mechanisms regarding the carbon (C) and N cycle from a well-validated process-based biogeochemical model, ecosys . The trained metamodel captures over 98% of the variability of the ecosys simulated outputs for 99 randomly selected counties in Iowa, Illinois, and Indiana. To identify hotspots with high mitigation potential, we introduce net societal benefit (NSB) as an indicator for synthesizing the loss in yield and social benefits through emissions and pollutants avoided. Our results show that reducing N fertilizer by 10% leads to 9.8% less N 2 O emissions and 9.6% less N leaching at the cost of 4.9% more SOC depletion and 0.6% yield reduction over the study region. The estimated total annual NSB is $395 M (uncertainty ranges from $114 M to $1271 M), including $334 from social benefits (uncertainty ranges from $46 M to $1076 M), $100 M from saving fertilizer (uncertainty ranges from $13 M to $455 M), and −$40 M due to yield changes (uncertainty ranges from −$261 M to $69 M). For the median scenario, we noted that 20% of the study area accounts for nearly 50% of the NSB, and thus represent hotspot locations for targeted mitigation. Although the uncertainty range suggests that developing such a high-resolution framework is not yet settled and the scenario based estimations are not appropriate to inform the management practices for individual farmers, our efforts shed light on the new generation of analytical tools for life cycle assessment.