Quantifying the global land carbon sink is essential for understanding carbon-climate feedbacks and developing effective mitigation strategies. Despite decades of research using eddy covariance, remote sensing, and atmospheric observations, carbon sink estimates remain highly uncertain across different approaches. The key challenge is to obtain the right global land carbon sink estimates for the right reasons. In this review, we first provide an overview of current methodologies for estimating terrestrial carbon sinks, including FLUXNET observations, top-down atmospheric inversions, and bottom-up approaches. We then leverage insights learned from FLUXNET to evaluate the observed range of annual net ecosystem exchange (NEE), gross primary production (GPP)-ecosystem respiration (Reco), and GPP-evapotranspiration (ET) relationships, and assess whether current models capture these observational constraints. Because FLUXNET is widely used as a benchmark for evaluation and as training data for model calibration, we also identify opportunities to strengthen the utility of FLUXNET for global land carbon sink inference.

High-resolution gross primary productivity (GPP) estimation is crucial for ecological and agricultural applications that require fine spatial details to capture GPP heterogeneity. Satellite-based GPP estimation usually relies on land cover and meteorological data. However, the misclassification of land cover data and coarse resolution of meteorological data greatly increase the uncertainty. Here, we propose a robust high-resolution GPP estimation deep learning (DL) network, named GPP-net, using only satellite surface reflectance (SR) from Sentinel-2 and photosynthetically active radiation (PAR). Specifically, GPP-net is based on a fully 1-D convolutional encoder-decoder network combined with a spectral band importance estimation module. To enhance the generalization of GPP-net, we ran the soil-canopy energy balance radiative transfer (SCOPE) model, and then combined these SCOPE-simulated reflectance data with GPP and PAR data extracted from FLUXNET2015 to pre-train GPP-net. Compared to benchmark models including near-infrared reflectance of vegetation multiplied by incoming sunlight (NIRvP), partial least squares (PLS) and random forest (RF), GPP-net improved half-hourly and daily GPP retrieval across seven plant functional types (PFTs) including four forest types, cropland, grassland and wetland. Owing to its robust nonlinear feature learning capabilities, GPP-net also facilitated robust GPP estimation across both C3 and C4 vegetation. We found that GPP-net could reliably estimate GPP under drought and heatwave conditions, with minimal improvement from including vapor pressure deficit (VPD) as a predictor. Furthermore, GPP-net demonstrated great robustness to soil effects in GPP mapping, and had strong ability in capturing inter-annual variability of GPP. The pretraining paradigm enabled us to fully leverage historical data, and the DL framework ensured that the model generalization continually improves as new data is integrated. Our model dispenses with land cover data and minimizes the requirements of coarse-resolution meteorological data for high-resolution GPP estimation, which could support future efforts in global high-resolution GPP mapping.

Photosynthesis governs terrestrial carbon uptake and tightly couples carbon, energy and water exchange. However, any single observation has limited spatiotemporal coverage. Eddy-covariance CO2 exchange measurements, for instance, are still underrepresented in the tropics. At the global scale, model-based estimates of photosynthesis, often quantified as gross primary productivity (GPP), remain highly uncertain. Solar-induced chlorophyll fluorescence (SIF), carbonyl sulfide (COS or OCS), and carbon isotope discrimination (Δ¹³C) provide complementary windows into photosynthesis. They offer partially independent constraints on energy partitioning, conductance limitations, and diffusion–carboxylation controls. Breathing Earth System Simulator (BESS) is a remote-sensing-driven, process-based model that couples canopy carbon assimilation, evapotranspiration, and surface energy balance. Building on BESS, we (1) incorporate the Johnson–Berry model to provide a mechanistic yet parsimonious description of energy conversion within the electron transport system, enabling SIF simulation while accounting for photosynthetic control, cyclic electron flow, and non-photochemical quenching; (2) couple OCS exchange to BESS through shared conductance pathways (stomatal and boundary-layer) and biochemical capacity (Vcmax25℃), and implement an explicit mesophyll conductance scheme so that net CO₂ assimilation is computed from chloroplastic CO₂ concentration (Cc); (3) integrate a ¹³C discrimination module that mechanistically estimates Δ¹³C along the explicitly simulated CO₂ diffusion pathway from the atmosphere to the chloroplast, accounting for fractionation during boundary layer, stomatal, and mesophyll diffusion, as well as Rubisco carboxylation. By coupling SIF, OCS exchange, and Δ¹³C within a shared canopy gas-exchange and energy-balance framework, BESS is extended into a multi-tracer forward framework that generates internally consistent predictions of these tracers together with carbon-water fluxes. Based on this framework, we aim to: (1) evaluate whether multi-tracer integration improves simulations of carbon-water fluxes; (2) explore multi-constraint parameter optimization or data assimilation using independent observations to reduce uncertainty in photosynthesis estimates; and (3) quantify relationships between tracer signals and fluxes (e.g., GPP–SIF, GPP–OCS, SIF–OCS, Δ¹³C–GPP) and their responses to environmental variability.

Global greening and browning, as evidenced by changes in leaf area index (LAI) derived from satellite observations, indicate how ecosystems respond to rising atmospheric CO 2 , climate change, and human interventions. However, uncertainties in satellite LAI records have led to conflicting conclusions about global trends and their drivers. Here, by developing a refined Advanced Very High-Resolution Radiometer LAI dataset with reduced radiometric and geometric uncertainties, we find a sustained global greening during 1982–2021 (0.035 m 2 m –2 decade –1 ). Global greening is largely driven by the continuous CO 2 fertilization (1982–2001: 72%; 2002–2021: 77%), while land-use management determines regional patterns. An ensemble of 15 Earth system models also simulates persistent greening (0.050 ± 0.044 m 2 m –2 decade –1 ), but fails to reproduce the observed spatial patterns of greening and drivers. These findings provide observational evidence for sustained global greening, with no clear signs of slowing CO 2 fertilization effect under current CO 2 and climatic conditions.

Robust estimation of the temperature sensitivity of ecosystem respiration (ER) is critical for projecting terrestrial carbon dynamics under climate change. While ER is traditionally modeled as an exponential function of temperature, recent evidence suggests a unimodal response, with respiration peaking at an optimum temperature (Topt) and declining thereafter. However, confounding variables like soil moisture, vegetation status, and vapor pressure deficit complicate isolating the effect of temperature. In this study, we combined long-term eddy covariance data from the flux tower in Sor & oslash;, Denmark (DKSor) with an explainable machine learning framework to disentangle the direct effects of air temperature (Tair) on nighttime ER. Using an optimized machine learning model and Causal Shapley value, we quantified the marginal contributions of environmental predictors while accounting for known causal dependencies. Then we employed a new concept of Topt, which is a temperature that maximizes the SHAP). Our results revealed a clear unimodal temperature response (Topt 15.99 degrees C), suggesting that ER may begin to decline at a cooler temperature than previously assumed (Topt = 18.25 degrees C). A binning analysis further revealed that unimodal response could be observed in normalized difference vegetation index ranges where high temperature (>= 15 degrees C) occurred, suggesting that unimodal temperature response was not biased by vegetation phenology. These findings underscore the importance of incorporating causal inference and interpretable machine learning in capturing the temperature response of ER.

The strong correlation between gross primary production (GPP) and sun-induced chlorophyll fluorescence (SIF) has been reported in many studies and is the basis of the SIF-based GPP estimation. However, GPP and SIF are not fully synchronous under various environmental conditions, which may destroy a stable GPP–SIF relationship. Therefore, exploring the difference between responses of GPP and SIF to the environment is essential to correctly understand the GPP–SIF relationship. As the common driver of GPP and SIF, the incident radiation could cause GPP and SIF to have similar responses to the environment, which may obscure the discrepancies in the responses of GPP and SIF to the other environmental variables, and further result in the ambiguity of the GPP–SIF relationship and uncertainties in the application of SIF. Therefore, we tried to exclude the dominant role of radiation in the responses of GPP and SIF to the environment based on the binning method, in which continuous tower-based SIF, satellite SIF, and eddy covariance GPP data from two growing seasons were used to investigate the differences in the responses of GPP and SIF to radiation, air temperature (Ta), and evaporation fraction (EF). We found that the following: (1) At both the site and satellite scales, there were divergences in the light response speeds between GPP and SIF which were affected by Ta and EF. (2) SIF and its light response curves were insensitive to EF and Ta compared to GPP, and the consistency in GPP and SIF light responses was gradually improved with the improvement of Ta and EF. (3) The dynamic slope values of the GPP–SIF relationship were mostly caused by the different sensitivities of GPP and SIF to EF and Ta. Our results highlighted that GPP and SIF were not highly consistent, having differences in environmental responses that further confused the GPP–SIF relationship, leading to complex SIF application.

The slow temperature acclimation of photosynthesis has been confirmed through early field experiments and studies. However, this effect is difficult to characterize and quantify with some simple and easily accessible indicators. As a result, the impact of slow temperature acclimation of photosynthesis on gross primary production (GPP) estimation has often been overlooked or not integrated into most GPP models. In this study, we used a theorical variable-state of acclimation (S), to characterize the slow temperature acclimation. This variable represents the temperature to which the photosynthetic machinery adapts and is defined as a function of air temperature (Ta) and time constant (tau) required for vegetation to respond to temperature, to discuss its impact on GPP simulation. We used FLUXNET2015 dataset to calculate S and established a GPP model using S and shortwave radiation (SW) based on random forest algorithm (S model). As a comparison, we directly used Ta and SW to build the other GPP model (Ta model). Moreover, the divergent temperature acclimation capacities of plants are crucial to predict and make preparations for likely temperature stress in the future. Therefore, the spatial distribution of tau values was also mapped using satellite sun induced chlorophyll fluorescence (SIF) and Ta datasets. The results indicated that: (1) taking into account the slow temperature acclimation of photosynthesis led to a more precise estimation of GPP which mainly reflected in reduction of excessive fluctuations in GPP predictions; (2) considering the slow temperature acclimation of photosynthesis can reduce the sensitivity of vegetation to temperature; (3) the improvement of S model in GPP estimations was different in different vegetation growth stages which was more significant in the springtime recovery stage; (4) tau values had significant spatial distribution which was strongly affected by the determinants of vegetation growth and seasonal variations in temperature.

Forests are facing various threats, such as drought, in the context of global climate change. Canopy water content (CWC) is a crucial indicator of forest water stress, mortality, and fire monitoring. However, previous studies on CWC have not adequately simulated forests with heterogeneous and discontinuous canopy structures. At the same time, there is a lack of field validation. This study retrieved the forest CWC across the contiguous U.S. (CONUS) with coupled radiative transfer models (RTMs) and the random forest (RF) algorithm. A Gaussian copula and prior knowledge were used for model parameterization. The results indicated that more accurate simulations of leaf trait dependencies and canopy structure characteristics lead to better CWC inversion. In addition, GeoSail, coupled with PROSPECT-5B, showed good performance (R2 = 0.68, RMSE = 0.15 kg m−2, MAE = 0.12 kg m−2, rRMSE = 12.78%, Bias = −0.036 kg m−2) for forest CWC retrieval. Large variation existed in forest CWC, spatiotemporally, and evergreen needle forest (ENF) showed strong CWC capacity. This study underscores the suitability of 3D RTMs for inversing the parameters of forest canopies.

Although the light use efficiency (LUE) models are widely employed to estimate ecosystem gross primary pro-duction (GPP), the majority of these models inadequately consider the effects of environmental and biological factors on GPP, resulting in considerable uncertainty. In addition, most developed LUE models have assumed that the maximum LUE (epsilon max) is a fixed value for different vegetation types, while epsilon max should be dynamic under environmental changes. The canopy nitrogen (N) concentrations were considered to have a significant linear relationship with epsilon max and could be estimated using various vegetation indices. In this study, we selected a vegetation index to characterize the canopy N concentrations and further simulate the dynamic epsilon max. We then developed an improved LUE model that simultaneously integrated the effects of canopy N concentrations, temperature, water, atmospheric carbon dioxide (CO2) and radiation components on the GPP estimates. Different forms of LUE models that partially integrate the above factors were also constructed for comparison. Our results showed that (1) the green chlorophyll index (CIgreen) correlated well with measured canopy N concentrations (R2 = 0.68), and the model using the CIgreen to characterize canopy N concentrations performed the best; (2) the GPP estimated using the improved model gave the best accuracy (R2 = 0.69, RMSE = 2.13 gC/m2/d, MAE=1.36 gC/ m2/d, IOA = 0.915) and performed well for different vegetation types when validated against the FLUXNET GPP; and (3) the estimated GPP had the best accuracy compared with MOD17 GPP and the revised EC-LUE GPP on a both daily and yearly scale. Overall, this study was an attempt to integrate N into the LUE model to obtain the spatiotemporally dynamic epsilon max while simultaneously taking into account the impacts of multiple environmental variables on the GPP estimates. The proposed model has the potential for satisfactory GPP simulations on a global or regional scale.

The terrestrial gross primary productivity (GPP) plays a crucial role in regional or global ecological environment monitoring and carbon cycle research. Many previous studies have produced multiple products using different models, but there are still significant differences between these products. This study generated a global GPP dataset (NI-LUE GPP) with 0.05° spatial resolution and at 8 day-intervals from 2001 to 2018 based on an improved light use efficiency (LUE) model that simultaneously considered temperature, water, atmospheric CO2 concentrations, radiation components, and nitrogen (N) index. To simulate the global GPP, we mapped the global optimal ecosystem temperatures (Topteco) using satellite-retrieved solar-induced chlorophyll fluorescence (SIF) and applied it to calculate temperature stress. In addition, green chlorophyll index (CIgreen), which had a strong correlation with the measured canopy N concentrations (r = 0.82), was selected as the vegetation index to characterize the canopy N concentrations to calculate the spatiotemporal dynamic maximum light use efficiency (εmax). Multiple existing global GPP datasets were used for comparison. Verified by FLUXNET GPP, our product performed well on daily and yearly scales. NI-LUE GPP indicated that the mean global annual GPP is 129.69 ± 3.11 Pg C with an increasing trend of 0.53 Pg C/yr from 2001 to 2018. By calculating the SPAtial Efficiency (SPAEF) with other products, we found that NI-LUE GPP has good spatial consistency, which indicated that our product has a reasonable spatial pattern. This product provides a reliable and alternative dataset for large-scale carbon cycle research and monitoring long-term GPP variations.

Vegetation productivity is an important parameter for estimating carbon stocks in terrestrial ecosystems and is important for monitoring regional and global ecological changes. In this study, gross primary productivity (GPP) and net primary productivity (NPP) products with a spatial resolution of 500 m and a temporal resolution of 8 days from 2000 to 2019 were produced based on Global land surface satellite (GLASS) leaf area index (LAI) and the fraction of absorbed photosynthetically active radiation (FPAR) products, and an improved light use efficiency (LUE) model that introduced clearness index (CI) to represent the effect of radiation on LUE. Validated by FLUXNET GPP data, Bigfoot NPP and EMID NPP data, the GPP and NPP products have high accuracy. The dataset has the potential to monitor global and regional ecology and vegetation growth conditions.

Previous studies have indicated that gross primary production (GPP) and solar-induced chlorophyll fluorescence (SIF) have a strong linear relationship, and usually exhibit similar spatial and temporal patterns. However, the responses of GPP and SIF to the environment may be different, which will lead to a variant GPP-SIF relationship. To better investigate the impact of the dynamics in GPP-SIF relationship on GPP estimation, we established two GPP models. An inconstant GPP/SIF ratio model (Dynamic-Ratio model, DR model) was first established using meteorological variables and leaf area index (LAI) based on random forest regression algorithm. The model was then used to estimate GPP (referred to as GPP_DR) with different satellite SIF datasets i.e., downscaled fine resolution SIF from the Orbiting Carbon Observatory-2 (GOSIF) and Global Ozone Monitoring Experiment-2 SIF (downscaled GOME-2 SIF). The second model (SIF-Climate-LAI model, SCL model) was also based on the random forest algorithm but was directly driven by meteorological variables, LAI and SIF data, and no GPP/SIF ratio was used in the model. As a comparison, the linear relationship between GPP and SIF was also established using eddy covariance tower GPP (GPP_EC) and SIF datasets based on linear regression without considering variations of GPP-SIF relationship (Fixed-Ratio model, FR model). Considering the spatio-temporal variations of GPP-SIF relationship can improve the GPP simulation to a certain extent by mitigating the underestimation of peak GPP values. This improvement was found for both DR and SCL models. Owing to the dynamic variations of GPP/ SIF ratio and associated uncertainties, the performance of DR model was not as good as that of SCL model. GPP estimation derived from GOSIF matched better with GPP_EC than that from downscaled GOME-2 SIF for DR, SCL and FR models. Our findings suggested that GPP can be better derived from satellite SIF by considering the variations of GPP-SIF relationship.

The scaling effect in remote sensing limits the estimation accuracy and application of remote sensing products, such as leaf area index (LAI). At present, the studies on scaling effect mostly focus on developing the algorithm for scaling bias correction, and rarely consider the model parameter type, which plays a critical role during the scaling bias calculation of LAI. In addition, some studies have suggested that the nonlinearity of normalized difference vegetation index (NDVI) equation affects the scaling bias calculation of LAI. However, few studies have been performed to clarify its effects in a quantitative way. In this paper, at two VALERI sites, discuss on the scaling bias calculation of LAI based on the Taylor series expansion method (TSEM) from two aspects: i) the nonlinearity of NDVI equation and ii) the model parameter selection (reflectance, NDVI and directional gap probability). The results indicate that the nonlinearity of NDVI equation has little effects on the LAI scaling bias calculation when NDVI was included in the retrieval model. On the other hand, when directional gap probability was considered in the retrieval model, it had effects. In comparison, when directional gap probability was used in the retrieval model, the scaling bias calculation of LAI showed higher quality.

The rapid urbanization process has threatened the ecological environment. Net primary productivity (NPP) can effectively indicate vegetation growth status in an urban area. In this paper, we evaluated the change in NPP in China and China’s urban lands and assessed the impact of temperature, precipitation, the sunshine duration, and vegetation loss due to urban expansion on NPP in China’s three fast-growing urban agglomerations and their buffer zones (~5–20 km). The results indicated that the NPP in China exhibited an increasing trend. In contrast, the NPP in China’s urban lands showed a decreasing trend. However, after 1997, China’s increasing trend in NPP slowed (from 9.59 Tg C/yr to 8.71 Tg C/yr), while the decreasing trend in NPP in China’s urban lands weakened. Moreover, we found that the NPP in the Beijing–Tianjin–Hebei urban agglomeration (BTHUA), the Yangtze River Delta urban agglomeration (YRDUA), and the Pearl River Delta urban agglomeration (PRDUA) showed a decreasing trend. The NPP in the BTHUA showed an increasing trend in the buffer zones, which was positively affected by temperature and sunshine duration. Additionally, nonsignificant vegetation loss could promote the increase of NPP. In the YRDUA, the increasing temperature was the main factor that promoted the increase of NPP. The effect of temperature on NPP could almost offset the inhibition of vegetation reduction on the increase of NPP as the buffer zone expanded. In PRDUA, sunshine duration and vegetation loss were the main factors decreasing NPP. Our results will support future urban NPP prediction and government policymaking.

推动区域均衡协调与可持续发展是我国的重大战略之一,植被净初级生产力(NPP)对生态环境是否可持续发展起着重要作用.以"胡焕庸线"为界把我国分为东、西部,从像元尺度和县级行政单元研究分析我国NPP、人口以及人均NPP时空变化,尤其是东、西部的区域差异.结果 表明:我国人口增长较快,从1982年的10.05亿增长到2017年的13.95亿,以"胡焕庸线"为界的西部占比由5.91％增长到6.42％;我国NPP整体呈现增长的趋势,总量由1982年的2.69 Pg C增长到2015年的3.24 Pg C,增长率为16.60 Tg C/a,其中东部增长率12.30 Tg C/a是西部(4.30 Tg C/a)的近3倍;西部人均NPP远大于东部与全国,1982、2000、2010、2017年西部与全国人均NPP持续处于下降的状态,但下降速率略有放缓,东部人均NPP则在2017年首次出现增长.据此可知我国整体生态环境处于恢复的状态,但不同区域之间差异较大,因此在相关政策制定方面应该充分考虑区域差异性,以实现我国生态环境的区域协调发展.

Numerous validation campaigns have been conducted over the last decade to assess the accuracy of the global leaf area index (LAI) products. Accurate and comprehensive validations for coarse-resolution LAI products are still very difficult due to lack of enough high-quality field measurements. Here we developed a fine resolution LAI dataset, consisting of 80 sample plots with an area of 3 km × 3 km in four major agricultural regions in China collected from 2003 to 2017. Instead of the indirect optical measurement method employed in most validation campaigns, the direct destructive method was employed to measure LAI of cropland for all the field experiments to avoid the measurement uncertainties, especially for crops at early growth stages with low height. Fine resolution reference LAI maps were derived from Landsat-5 TM and Landsat-8 OLI surface reflectance products based on the semi-empirical inversion model, which were calibrated using field measurements for each growth stage with an RMSE ranging from 0.22 to 0.95, and a relative root mean square error (RRMSE) ranging from 7.58% to 44.42%. Then, 80 sample plots with an area of 3 km × 3 km were selected as the fine resolution validation dataset from the fine resolution reference LAI maps with a proportion of cropland larger than 75% and one or more in-situ samples were contained in each 3 km × 3 km reference map.

由于茶园大多分布在地形复杂的山区,地块破碎,分布零散,形状差异大、植被混杂且茶园所处环境长期受到云雨的影响,增加了茶园遥感识别的难度与不确定性,针对这一问题,该研究提出了利用高分1号(GF-1)和哨兵2号(Sentinel-2)时序数据提取茶园的方法,以浙江省武义县王宅镇为研究区,采用GF-1号为主要数据源,并利用MODIS地表反射率产品和Sentinel-2反射率数据,基于时空融合算法得到时间分辨率5 d的10 m Sentinel-2完整的时序数据.综合利用GF-1在空间细节方面的优势和重建的Sentinel-2高观测频率时序数据在反映茶树生长过程方面的优势,分别基于GF-1的光谱和纹理特征及GF-1的光谱、纹理特征和Sentinel-2时序特征两种特征组合方式,采用随机森林算法提取茶园.结果表明,GF-1光谱、纹理信息结合Sentinel-2时序信息分类结果的准确率、错误率、精确率、召回率和F1分数分别为96.91％、3.09％、89.00％、83.09％和0.86,仅基于GF-1光谱和纹理信息的分类准确率、错误率、精确率、召回率和F1分数分别为94.72％、5.28％、73.09％、84.61％和0.78,添加时序信息分类结果总体优于未添加时序信息的分类结果.表明高空间分辨率结合高频率时序遥感数据是提高茶园分类精度的有效手段.

Long time series of vegetation productivity products are significant for the research of global carbon cycle and climate change. In this article, the 0.05° global gross primary productivity (GPP) and net primary productivity (NPP) products from 1981 to 2018 were estimated by using the improved multisource data synergized quantitative (MuSyQ) NPP algorithm. The model was based on the fraction of absorbed photosynthetically active radiation (FPAR) and leaf area index (LAI) data from the global land surface satellite (GLASS) dataset, the light use efficiency (LUE) from the parameterization approach with the clearness index (CI), the ERA-Interim meteorological data, and other environmental factors. The results suggested that the accuracy of the MuSyQ GPP product was slightly higher than that of the MOD17 GPP product when compared with the FLUXNET GPP, especially for the evergreen broadleaf forest (EBF), deciduous broadleaf forest (DBF), wetland (WET), cropland (CRO), woody savanna (WSAV), and closed shrubland (CSH) land types. MuSyQ NPP product also has higher accuracy [ R 2 = 0.81, RMSE = 214.6 gC/(m 2 year)] than MOD17 NPP [ R 2 = 0.55, RMSE = 214.7 gC/(m 2 year)] when compared with the BigFoot NPP, which indicated the reliability of the improved MuSyQ-NPP algorithm in estimating global NPP. Our results showed a significant upward trend in global NPP, which was most affected by FPAR, followed by LUE, temperature, and PAR. The average NPP declined significantly in Asia and Amazon tropical rainforests and increased significantly in Africa tropical rainforest, which were affected by the local deforestation or the forest expansion, and also the climate factors.

The accuracy assessment of the global leaf area index (LAI) products is an indispensable step before applications. Four popular LAI products, namely MCD15A2H, GLASS, GEOV2, and GLOBMAP, were assessed over croplands in China using field measurements. All of these four products were validated with upscaled reference LAI maps. The validation results revealed uncertainties in these products for crops with RMSE ranging from 0.49 to 1.37 over the validation sites in Beijing, in Zhoukou and Jiaozuo counties in Henan province, Youyi farm in Heilongjiang province, and in Longkang farm in Anhui province. GEOV2 gave the highest accuracy (R2 = 0.85, RMSE = 0.49, relative bias = −7.2%) compared to MCD15A2H (R2 = 0.54, RMSE = 0.91, and relative bias = −24.4%), GLASS (R2 = 0.80, RMSE = 0.73, and relative bias = −23.3%), and GLOBMAP (R2 = 0.25, RMSE = 1.37, and relative bias = −55.4%). The LAI products overestimated over the croplands with dark soil-backgrounds in Youyi farm while they underestimated over other validation sites. The seasonal variation of these products was assessed with continuous in situ measurements at Daman Station, Gansu province. All of these four products showed good temporal consistency at Daman Station, while GEOV2 showed the highest accuracy (R2 = 0.92, RMSE = 0.41, relative bias = −10.3%). The influence of scaling effect on the products’ accuracy was also investigated, and the scaling differences of these LAI products contributes overestimation at coarse-resolution for croplands with dark soil-backgrounds, while underestimation at coarse-resolution in other validation sties.