Agricultural and Forest Meteorology最新文献

{"title":"Record-breaking high temperature amplifies the negative anomaly of tropical net land carbon sinks in the 2023-2024 El Niño","authors":"Xiaomeng Du ,&nbsp;Philippe Ciais ,&nbsp;Stephen Sitch ,&nbsp;Frédéric Chevallier ,&nbsp;Michael O’Sullivan ,&nbsp;Ana Bastos ,&nbsp;Sönke Zaehle ,&nbsp;Piyu Ke ,&nbsp;Lei Zhu ,&nbsp;Zhixuan Guo ,&nbsp;Yi Leng ,&nbsp;Wanjing Li ,&nbsp;Jefferson Goncalves de Souza ,&nbsp;Wei Li","doi":"10.1016/j.agrformet.2025.110793","DOIUrl":"10.1016/j.agrformet.2025.110793","url":null,"abstract":"<div><div>The recent El Niño event developed from May 2023 to June 2024, and it experienced record-breaking high temperatures, which is different from the previous El Niño events. Impacts of these extreme climate conditions on the tropical carbon sink in 2023 and 2024 compared to other El Niño years remain unclear. Here we used atmospheric inversions and rapidly updated dynamic global vegetation models (DGVMs) to quantify the terrestrial carbon sink anomalies in the tropics. The tropical land acted as a carbon source in both 2023 and 2024. The inversion indicated a tropical land carbon sink anomaly (after detrending) of -0.85 and -0.68 PgC yr^-1 in 2023 and 2024, respectively. Although the intensity of El Niño (Oceanic Niño Index) in 2023-2024 was lower than the previous two strong El Niño events (1997-1998, 2015-2016), the terrestrial carbon sink anomaly was comparable in magnitude to that of 2015-2016. This negative anomaly was largely contributed by carbon sources in tropical America. Reduced photosynthesis is the primary cause of the simulated reduction in the tropical carbon sink during this period. The stronger temperature sensitivity combined with large temperature anomalies contributed to the negative carbon sink anomaly. The amplifying effect of temperature in terrestrial carbon sinks in the 2023-2024 El Niño suggests that long-term warming is likely to exacerbate carbon loss in extreme climate events, increasing potential risks for ecosystem sustainability and carbon sequestration.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110793"},"PeriodicalIF":5.7,"publicationDate":"2025-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144908842","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Tracking drought in dryland vegetation through the photosynthetic afternoon depression index of Sun-induced chlorophyll fluorescence","authors":"Sicong He,&nbsp;Yanbin Yuan,&nbsp;Heng Dong,&nbsp;Yibo Geng,&nbsp;Tao Xiong,&nbsp;Feng Guo","doi":"10.1016/j.agrformet.2025.110799","DOIUrl":"10.1016/j.agrformet.2025.110799","url":null,"abstract":"<div><div>Vegetative photosynthesis is highly sensitive to water and heat stress, and the indirect monitoring of vegetative photosynthesis through Sun-induced chlorophyll fluorescence (SIF) has significant potential in global drought monitoring. However, substantial knowledge gaps remain regarding effective methods for assessing vegetation drought stress using remotely sensed SIF data. In this study, we employ GOCI geostationary satellite observations and OCO-3 SIF retrieval to drive a machine learning model for the purpose of monitoring SIF in typical drylands in China at high spatial resolution (500 m). Additionally, we investigated the spatial response patterns and quantitative metrics of drought by SIF and its decoupled components. The data-driven SIF reconstruction products successfully captured the afternoon decrease in photosynthesis in both space and time, particularly evident during the 2020 summer drought-heatwave composite event. It was observed that the disparity in photosynthetic intensity between the morning and afternoon periods was markedly diminished with the advent of drought conditions. The difference-type index, based on these observations, showed statistically significant correlation with both the soil drought anomaly indicator (SMZ; Pearson r: 0.53; <em>P</em> &lt; 0.05) and the Standardized Precipitation Evapotranspiration Index (SPEI; Pearson r: 0.71; <em>P</em> &lt; 0.01). Furthermore, it exhibited superior performance compared to the SIF and SIF yields derived from a single time observation. This study demonstrates the application of SIF for drought monitoring in drylands vegetation at a fine spatial scale, emphasizing the importance of multi-temporal remote sensing monitoring of vegetation photosynthesis for drought tracking.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110799"},"PeriodicalIF":5.7,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144895342","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Examining the direct and indirect impacts of urbanization on vegetation net primary productivity across Chinese cities","authors":"Yulong Zhao ,&nbsp;Hao Chen ,&nbsp;Wenyan Ge ,&nbsp;Shangyu Shi ,&nbsp;Rongqi Li ,&nbsp;Tamrat Sinore ,&nbsp;Fei Wang","doi":"10.1016/j.agrformet.2025.110815","DOIUrl":"10.1016/j.agrformet.2025.110815","url":null,"abstract":"<div><div>China has undergone the world's most rapid urbanization, drastically altering regional climates and urban vegetation growth environments. However, the mechanisms of urbanization on carbon sequestration in both urban green spaces and surrounding vegetation remain underexplored. This study utilized buffer zone analysis to evaluate the spatial heterogeneity of urbanization’s effects on Net Primary Productivity (NPP) across 271 cities in China, focusing on different buffer distances. Through quantitative analysis of both the direct and indirect impacts of urbanization on NPP, the findings revealed that urbanization not only indirectly affects urban green spaces but also exerts significant negative indirect effects on vegetation in surrounding buffer zones due to spatial correlations. These findings underscore the importance of selecting areas unaffected by built-up zones to accurately quantify the effects of urbanization Direct effects of urbanization significantly reduced NPP by 30 % to 80 %, with impacts increasing toward northern latitudes. Negative indirect effects were in the eastern and central regions with reductions between 50 % and 100 %. 30 % of cities showed positive indirect effects. Moreover, the urban heat island effect was found to weaken regional carbon sequestration capacity significantly. In areas with favorable water and heat conditions, NPP is less likely to recover to its original productivity levels after experiencing human disturbances. In contrast, some arid cities exhibited positive indirect effects of urbanization, highlighting the critical role of human interventions, such as irrigation and fertilization, under specific ecological conditions. This study offers valuable insights into the impacts of urbanization on carbon sequestration within urban and surrounding ecosystems.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110815"},"PeriodicalIF":5.7,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144895268","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Soil CO2 and CH4 effluxes in powerline rights-of-way and their adjacent forests","authors":"Antoine Harel ,&nbsp;Evelyne Thiffault ,&nbsp;David Paré ,&nbsp;Renée Hudon ,&nbsp;Maude Larochelle ,&nbsp;Yann Chavaillaz","doi":"10.1016/j.agrformet.2025.110801","DOIUrl":"10.1016/j.agrformet.2025.110801","url":null,"abstract":"<div><div>Global decarbonization will require a large deployment of power grids to convey electricity. The right-of-way (i.e., the cleared area below the pylons, where vegetation is periodically maintained) is a land-use change that involves changes in soil and vegetation and their carbon dynamics both within the rights-of-way and in adjacent forests, notably via an edge effect. Our main objective was to assess whether soil CO₂ effluxes (F_CO₂), soil CH₄ effluxes and microclimate (soil temperature and water content) differed between powerline rights-of-way and their adjacent forests compared to control forests over a large bioclimatic gradient of upland sites across the temperate and boreal forests of Eastern Canada. Monthly efflux measurements were carried out between May and October 2023 and 2024 in eight rights-of-way and their adjacent edge and control forests. Overall, cumulative total F_CO₂ during the snow-free period were lower (–7.57 %) in rights-of-way and higher (+11.20 %) in the edge forests compared to the control forests. However, these results were not consistent across the bioclimatic gradient: balsam fir forests, contrarily to forests from both cooler and warmer bioclimatic domains, showed enhanced soil respiration in rights-of-way. Overall, soils were warmer and wetter in rights-of-way compared to control and edge forests; however, no effects were found on the soil methane uptake. Our study indicated that the presence of a powerline right-of-way influences soil biogenic carbon emissions. Effects are related both to changes in abiotic and biotic conditions. These estimates should improve the assessment of the carbon footprint of power transmission and electricity deployment.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110801"},"PeriodicalIF":5.7,"publicationDate":"2025-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144892975","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A novel model quantifies epiphyte-mediated temperature and water dynamics in a tropical montane cloud forest","authors":"David Carchipulla-Morales ,&nbsp;Haley Corbett ,&nbsp;Damon Vaughan ,&nbsp;Sybil G. Gotsch ,&nbsp;Todd E. Dawson ,&nbsp;Nalini Nadkarni ,&nbsp;Lauren E.L. Lowman","doi":"10.1016/j.agrformet.2025.110770","DOIUrl":"10.1016/j.agrformet.2025.110770","url":null,"abstract":"<div><div>Tropical montane cloud forests (TMCFs) are ecosystems with high biodiversity that are threatened by deforestation, land use changes, and climate change. One of the unique aspects of TMCFs is the high biomass and diversity of epiphytes. Epiphytes are vascular and non-vascular plants that live in tree canopies, creating arboreal micro-ecosystems. They provide ecological services by capturing and retaining allochthonous nutrients from rain and fog, and by supporting the presence of canopy pollinators and other fauna. Predicted changes in cloudiness and land conversion threaten the abundance of epiphytes, and thus their capacity to contribute to ecosystem functions. However, how losses in epiphyte abundance will affect microclimate and host tree water status is still unclear and requires the ability to simulate the role of epiphytes in canopy water storage dynamics. We developed a water balance model for epiphytes in TMCFs. We consider epiphytes in the host tree as a water store inside the canopy that is filled via precipitation from both rain and fog, and depleted via evapotranspiration and host tree water uptake. The model was used to simulate water and energy fluxes between the epiphytes and their surroundings under idealized and real dry season conditions for TMCFs near Monteverde, Costa Rica. Results from the idealized and real simulations capture how epiphytes retain water under dry-down conditions, leading to small diurnal variability in temperature, low evapotranspiration rates, and enhanced dew deposition at night. We find that dew deposition recharges up to 34 % of epiphyte water storage lost due to evapotranspiration over a 3-day dry-down event. Our results provide the first quantitative demonstration of the importance of epiphyte water storage on temperature and dew formation in TMCFs. This work sets the foundation for developing a process-based understanding of the effects of epiphyte loss on TMCF ecohydrology.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110770"},"PeriodicalIF":5.7,"publicationDate":"2025-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144890993","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The response of ecosystem marginal water use efficiency to soil drying","authors":"Yuanzhizi Deng,&nbsp;Yao Zhang","doi":"10.1016/j.agrformet.2025.110800","DOIUrl":"10.1016/j.agrformet.2025.110800","url":null,"abstract":"<div><div>Marginal water use efficiency (λ) describes the increase in carbon assimilation per unit of water loss. The stomatal optimization theory posits that λ remains stable over short timescales, yet to what extent λ varies across space and time, and how it relates to ecosystem-level characteristics remains unclear. Here, we estimated daily λ using ecosystem-scale underlying water use efficiency (uWUE) derived from eddy covariance (EC) measurements and analyzed its trend along declining soil moisture. We found that λ generally increased with declining soil moisture, and its sensitivity to soil water deficit (λ gradients) also intensified under drier conditions. This response was more pronounced in forest and woody ecosystems compared to non-forest systems. The weakest λ gradients are observed in regions with intermediate aridity, moderate soil texture, lower tree cover, and shallower rooting depth. These different patterns of λ gradients along soil drying can inform the ecosystem-level plant water use strategies, with stronger gradients often corresponding to ecosystems dominated by more isohydric species. Our study highlights the necessity to consider spatiotemporal variations of λ to predict stomatal behaviors in response to drought.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110800"},"PeriodicalIF":5.7,"publicationDate":"2025-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144890990","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Introducing a new clustering-based method for regionalization framework for continental-scale rainfall estimates from soil moisture dynamics using machine learning methods","authors":"Mohammad Saeedi ,&nbsp;Hyunglok Kim ,&nbsp;Venkataraman Lakshmi","doi":"10.1016/j.agrformet.2025.110766","DOIUrl":"10.1016/j.agrformet.2025.110766","url":null,"abstract":"<div><div>Rainfall estimation plays a key role in various hydrological applications, ranging from flood forecasting and drought monitoring to water resource management. Traditional methods, which depend on ground-based gauges and remote-sensing products, can be expensive and limited by geography, and they often suffer from issues like sensor resolution or atmospheric interference. To tackle these problems, “bottom-up” strategies have emerged that use soil moisture as a stand-in for rainfall. By leveraging soil’s natural capacity to capture precipitation, these methods can reduce the reliance on high-resolution sensors and intricate modeling.</div><div>Nonetheless, their performance still depends heavily on careful calibration, a process that usually calls for plenty of on-site data, extended observation periods, or location-specific fine-tuning. To address these hurdles, we present a calibration parameters regionalization framework that does away with the need for a dedicated calibration phase. This framework uses both unsupervised (K-means clustering) and supervised (rainfall-intensity classification) techniques together with a genetic algorithm to automatically determine model parameters, without depending on adjustments tailored to specific regions.</div><div>We illustrate our method using the soil moisture to rainfall (SM2RAIN)-Net Water Flux (NWF) algorithm, demonstrating its ability to accurately estimate rainfall across the well-monitored contiguous United States (CONUS). Our findings indicate that SM2RAIN<img>NWF performs particularly well in areas with higher rainfall intensity, outperforming the classic SM2RAIN methods that are commonly used for estimating rainfall from soil moisture dynamics. In fact, this is the first time K-means, a genetic algorithm, and rainfall clustering have been combined to estimate rainfall without requiring a separate calibration period, achieving a 20 % improvement in Nash–Sutcliffe efficiency and a 10 % reduction in root mean square error compared to classical methods.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110766"},"PeriodicalIF":5.7,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144890994","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Different responses of Pinus sylvestris L. and Larix decidua Mill. to forest fire in Central England (UK)","authors":"Francesco Niccoli ,&nbsp;Luigi Marfella ,&nbsp;Jerzy P. Kabala ,&nbsp;Jon Rowe ,&nbsp;Rossana Marzaioli ,&nbsp;Flora A. Rutigliano ,&nbsp;Helen C. Glanville ,&nbsp;Giovanna Battipaglia","doi":"10.1016/j.agrformet.2025.110804","DOIUrl":"10.1016/j.agrformet.2025.110804","url":null,"abstract":"<div><div>The United Kingdom (UK) is facing a growing threat due to the increasing frequency of fires attributed to anthropic pressures and activities. This research analysed the impact of a human-induced 2018 wildfire in a mixed woodland of <em>Pinus sylvestris</em> L. and <em>Larix decidua</em> Mill. in The Roaches Nature Reserve (central England). Through a multidisciplinary approach integrating remote sensing, forest surveys, dendrochronology and soil analysis, we compared burned and non-burned (control) trees to assess the eco-physiological responses of two plant species. Remote sensing supported both the strategic planning of field activities and the characterization of vegetation dynamics affected by fire under pre- and post-fire trajectories, while dendrochronological and soil analyses provided crucial information on post-fire forest dynamics. Results showed that, although both species demonstrated good resistance to the immediate impact of the fire, their responses in terms of resilience and recovery were different in the medium-term (5 years). <em>P. sylvestris</em> (Scots pine) showed good resilience and recovery capacity, with surviving trees showing improved growth within five years, though full recovery may still require several years. In contrast, <em>L. decidua</em> (European larch), although a fire-adapted species, experienced nearly total mortality within three years, most likely because of secondary stressors, such as a pathogen outbreak, which potentially compromised its resilience and recovery capacity in the medium-term. According to our results, climate does not seem to have played a determining role in larch tree mortality, as weather conditions were favourable for both species over the years. Likewise, soil properties showed no variation that could decisively influence survival dynamics. This study highlights the importance of understanding species-specific responses to fire and potential secondary stress factors, emphasizing the need to implement effective management strategies for prevention and management of mixed forests in regions where fire incidence is emerging.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110804"},"PeriodicalIF":5.7,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144891027","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The climatic patterns that control regional fire activity in the Brazilian savanna","authors":"Patrícia S. Silva ,&nbsp;Renata Libonati ,&nbsp;Luiz G. Gonçalves ,&nbsp;Carlos C. DaCamara","doi":"10.1016/j.agrformet.2025.110792","DOIUrl":"10.1016/j.agrformet.2025.110792","url":null,"abstract":"<div><div>Fire activity in the Brazilian savanna (Cerrado) is heavily constrained by climate, however the climate patterns that lead to extreme fire seasons are not yet well understood. Climate conditions during the fire season determine fire weather, but climate patterns prior to the fire season months may also modulate fuel availability and condition. In the context of a changing climate, understanding the climatic patterns that lead to extreme fire events, and their mediating factors, is crucial to build resilient landscapes and inform decision-making. In this study, we propose to uncover the nature of these relationships for Cerrado. We evaluate the regional temperature and precipitation patterns that lead to severe and mild fire seasons for each of the 19 ecoregions of Cerrado. We identify two periods that show contrasting behaviours in both extremes: the concurrent climate conditions during the fire season months (August to October) and pre-conditions during the austral autumn (March to May). Despite noteworthy regional discrepancies, in general we find that severe fire seasons are preceded by hot and dry conditions during autumn and associated with hot and dry conditions during the fire season months. Mild fire seasons see the opposite pattern, with colder and wetter conditions both during and prior to the fire season. We further investigate the influence of these climatic patterns in extreme fire activity for each month of the fire season and find that, over most ecoregions, early fire season burned areas are influenced by pre-conditions during autumn, whereas late fire season burned areas rely on concurrent favourable meteorological conditions. These results contribute to the understanding of the regional fire-climate dynamics of the second largest biome in South America and provide a starting point for regional fire outlooks. We further provide regionally tailored information that, considering recent Brazilian policies, may prove useful for fire management.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110792"},"PeriodicalIF":5.7,"publicationDate":"2025-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144880313","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Variability and uncertainty in net ecosystem carbon exchange modeling: Systematic estimates at global flux sites via ensemble machine learning","authors":"Nannan Wang ,&nbsp;Zijian Yue ,&nbsp;Yaolin Liu ,&nbsp;Zhaomin Tong ,&nbsp;Yanfang Liu ,&nbsp;Yanchi Lu ,&nbsp;Yongge Shi","doi":"10.1016/j.agrformet.2025.110784","DOIUrl":"10.1016/j.agrformet.2025.110784","url":null,"abstract":"<div><div>Predicting net ecosystem carbon exchange (NEE) is crucial for understanding carbon dynamics. Machine learning (ML) has become pivotal for site-level modeling and spatial upscaling for NEE, yet spatiotemporal variability and uncertainty challenge its reliability and universality. Systematically quantifying variability and uncertainty sources in NEE modeling remains lacking due to the scale-dependent nature of carbon flux variations. Thus, this study established a systematic framework to evaluate how model construction choices and environmental predictors could impact ML-based NEE modeling across timescales with multifaceted evaluation criteria. Using observations from FLUXNET 2015, AmeriFlux, and ICOS, alongside multi-source data, this study conducted separate models for each combination of four timescales (daily, weekly, monthly, and yearly), four tree-based ensemble algorithms, and three data-splitting rules. Multi-faceted assessment included overall, across-site, seasonal, and anomaly perspectives. Key findings include: (1) <em>Model construction.</em> Boosting (LightGBM, XGBoost, and CatBoost) excelled in capturing temporal variability and anomaly, whereas bagging (Random Forest) was effective for spatial variability. Complete-random data splitting increased overfitting risks and should be avoided. (2) <em>Predictors.</em> Environmental controls on accuracy varied with timescales, data situations, and ambient conditions. Predictors for NEE modeling should be selected based on their causal importance (e.g., evapotranspiration, vapor pressure deficit, and air temperature) and statistical relationships (e.g., leaf area index, elevation, and precipitation) with NEE, tailored to specific ambient conditions. Excessive predictors may degrade NEE prediction accuracy, particularly at large scales or in regions with high environment like arid areas. (3) <em>Evaluation criteria.</em> Rigorous multi-metric accuracy assessments proved essential, as reliance on single metrics or overall accuracy could yield contradictory results. For instance, daily models achieved higher anomaly NSE (0.33 vs. 0.25) but lower overall NSE (0.54 vs. 0.59) than monthly models. NEE predictions exhibited greater challenges in accounting for spatial than temporal variability, resulting in lower accuracy for inter-annual than intra-annual predictions. This study advances ML-driven carbon flux modeling with actionable insights.</div></div>","PeriodicalId":50839,"journal":{"name":"Agricultural and Forest Meteorology","volume":"374 ","pages":"Article 110784"},"PeriodicalIF":5.7,"publicationDate":"2025-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144880058","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}