The Normalized Difference Vegetation Index (NDVI) is widely used for vegetation monitoring, phenological analysis, and ecosystem assessment. However, obtaining continuous daily high-resolution NDVI time series remains challenging because high-resolution satellite observations are often affected by long revisit intervals and cloud contamination. Existing reconstruction methods generally rely on dense observations or auxiliary datasets, limiting their applicability in data-sparse regions. In this study, we propose a method by extracting phenological archetypes from tower observations to reconstruct GF NDVI time series. The proposed method successfully reconstructed daily NDVI for 2024 over a forested area. The results show that the reconstructed NDVI achieves an RMSE of 0.100 at 16 m resolution (R2=0.942) and 0.158 at 1.5 m resolution (R2=0.789), compared with tower observations. The method enhances spatial detail by more than five times relative to the original GF NDVI while maintaining temporal consistency, requiring only tower observations and limited satellite data. These results demonstrate the potential of the proposed approach for high-resolution daily NDVI reconstruction in data-sparse regions.

Leaf area index (LAI) and leaf chlorophyll content (LCC) are two key vegetation traits affecting vegetation growth and function. They have often been retrieved individually from multi-spectral remote sensing data by ignoring their mutual influence on spectral signals. In this study, we developed a machine learning algorithm for synergetically retrieving both traits at the same time through training the algorithm using simulations of a radiative transfer model PROSAIL. The algorithm determines LAI based on visible, red-edge, near-infrared and shortwave infrared bands, while it infers LCC mostly from visible and red-edge bands. In this way, the mutual influences of LAI and LCC on visible and red-edge bands are considered in the algorithm. The algorithm is applied to a rice field over multiple growing seasons (2017-2018, 2024) using Sentinel-2 data. It is found that synergetically retrieved LAI and LCC are more accurate (R2 = 0.82 and 0.85, RMSE = 0.83 and 5.00 mu g/cm2, respectively) than individually retrieved LAI (R2 = 0.54, RMSE = 1.49) and LCC (R2 = 0.85, RMSE = 9.35 mu g/cm2). Our results suggest that this synergetic retrieval principle and methodology may be utilized to support the development and improvement of LAI and LCC products for rice paddies, while further cross-site validation is required to assess broader generalizability.

Context: Currently, most ET related models usually set the light extinction coefficient (C) as a constant, even though it actually changes with season even within a day. Previous studies proved that daily transpiration modelling was significantly improved by using varied C. Objective: We aimed to explore whether considering C changes can improve transpiration simulation at hourly scale, and whether the dynamic C calculation method at daily scale can be used at hourly scale. Methods: A quantitative method that was previously suggested at daily scale was used to capture C variations in an apple orchard and an orange orchard. Furthermore, the method was developed to a new equation for better capturing C changes at hourly scale. The constant C method and different dynamic C methods were integrated into the Penman-Monteith model separately for hourly transpiration modelling based on Bayesian analysis. Results and conclusions: The results show that: i) Compared with the fixed C, dynamic C significantly improved transpiration simulation at hourly scale, with the Coefficient of Determination increased (around 0.69 and 0.68 for apple and orange orchards respectively) and the Mean Relative Error dropped (around 16 % and 19 % for apple and orange orchards respectively) for both calibration and validation periods. ii) When capturing C dynamics, different factors should be considered for different orchards. iii) The new dynamic C method we proposed achieved better transpiration estimation accuracy at both hourly and daily scales than the one previously proposed focusing on daily scale. Significance: These findings deepened the understanding of evapotranspiration processes and are vital for more precision water management in agriculture.

Fractional vegetation cover(FVC)serves as a crucial indicator for assessing the health status of terrestrial vegetation ecosystems.Currently,most global FVC products derived from satellite remote sensing have a temporal resolution of 8-10 days,which proves insufficient for capturing vegetation dynamics during rapid growth periods.Based on the gap probability theory,this study utilized the global GLASS leaf area index(LAI)product with a spatial resolution of 5 km and a temporal resolution of 8 days,along with a priori clumping index information,to generate a global FVC product at 5 km/8-day resolution from 1981 to 2000.Furthermore,the global MuSyQ LAI product with a spatial resolution of 500 m and a temporal resolution of 4 days,together with clumping index products,was used to generate a global FVC product at 500 m/4-day from 2001 to 2020.Validation using ground-based measurement from the VALERI and IMAGINES projects showed that the product achieved a root mean square error(RMSE)of 0.15 and a mean relative error of 12.56%.This product provides data support for long-term global vegetation dynamics monitoring.

Stem diameter is a key agronomic trait in sorghum (Sorghum bicolor L.), influencing biomass accumulation and stalk strength. It also affects practical outcomes such as mechanical harvesting efficiency and end-use quality. Despite its importance, the genetic basis of this trait remains unclear. To address this, we performed a genome-wide association study (GWAS) on a diverse panel, which identified a candidate locus containing Sobic.010G085400 (SbMADS52), a gene encoding a MADS-box transcription factor. Within this gene, we identified a putative functional G185S substitution associated with both stem thickness and plant height, suggesting a role in stalk architecture. To leverage this variant for breeding, we developed and validated a putative functional KASP marker. Population genetic analyses further revealed that the favorable allele has been under selection since domestication but is less frequent in modern breeding, highlighting an underused genetic resource. Collectively, our study characterizes a candidate regulator of stem morphology and provides a practical marker that may be used for sorghum breeding.

Temporal similarity is critical for retrieving and integrating geoscientific data. Existing approaches often neglect the effect of time scale and lack efficient models for handling composite temporal objects composed of multiple points and intervals. To address these limitations, we propose a unified temporal similarity framework that explicitly incorporates both time-scale effects and composite temporal topology.To evaluate the scale-sensitive model, we compile the Geoscientific Data Retrieval and Ranking Dataset (GDRRD), containing 105 datasets organized into themed query groups with expert rankings. On GDRRD, our method improves the Spearman correlation between system ranking and expert ranking (from 0.905 to 0.976 compared with a representative baseline) and shows superior discrimination for disjoint relations under decade-scale and century-scale settings.To extend the framework to composite time, we introduce a sparse matrix representation based on temporal transitivity and a most-similar matching model that aggregates topological and metric similarities with length-weighted integration. Experiments on composite-time cases derived from gridded population density datasets (2000–2015, 1 km resolution) and remote sensing image sequences demonstrate that our method reduces redundant relation storage by over 40%, while yielding more precise similarity scores for complex temporal queries such as “2000, 2005, 2010” versus “2000–2010”.By jointly achieving scale-awareness and form-completeness, the proposed framework provides a robust foundation for temporal similarity computation, enabling more accurate and rational retrieval in large-scale geoscientific data repositories.

Global Aboveground forest Biomass(AGB)products have become increasingly abundant in recent years,providing valuable data for assessing carbon stocks and fluxes.However,substantial spatiotemporal inconsistencies among these products have led to large uncertainties in the estimation of global carbon storage and carbon sink strength.This study aims to evaluate systematically the interannual consistency of major global AGB products and identify their strengths and limitations for long-term biomass monitoring and carbon accounting.We integrated multi-source forest biomass data of AGB data,including satellite-derived products,e.g.,European Space Agency Climate Change Initiative(CCI),NASA Jet Propulsion Laboratory(JPL),Dynamic Global Vegetation Model(DGVM)simulations(net biome poductivity,carbon in vegetation),and ground-based validation datasets.A multidimensional evaluation framework was designed from three perspectives:(1)spatial consistency,assessed using correlation coefficients and spatial agreement metrics among products;(2)interannual variability,analyzed through temporal correlation and trend consistency;(3)ground validation,performed using field observations to quantify product accuracy in regions of biomass increase and decrease.The results show the following:(1)different remote sensing products exhibit pronounced differences in spatial consistency.Single-epoch products based on Global Ecosystem Dynamics Investigation(GEDI)and Ice,Cloud,and Land Elevation Satellite-2(ICESat-2)show relatively high spatial consistency(ρc>0.7),with the most significant consistency found in tropical forest regions of South America and Africa.Long-timeseries remote sensing products(CCI and JPL)demonstrate higher consistency in interannual variability compared with other products.(2)For long-timeseries biomass products,interannual consistency improves to a certain extent as time span increases.However,CCI and JPL show fewer regions of consistency in high-latitude areas(40°—60° N/S).JPL exhibits higher interannual variability values in Asia compared with in other regions.By contrast,DGVM data indicate substantially higher interannual variability in tropical regions(20°S—20°N),with an overall tendency toward carbon sink estimates(proportion of pixels with increasing trends>80%).Nevertheless,in high-latitude regions,interannual variability estimated by DGVM diverges strongly from that of remote sensing products.(3)Ground validation indicates that CCI performs better in regions of biomass increase(r=0.36,RMSE(root mean square error)=8.54 Mg/hm2),but performs poorly in regions of biomass decrease(r<0.15,RMSE>14 Mg/hm2).Both result show some degree of underestimation,though the bias is smaller than that of DGVM simulations.This study provides a comprehensive,multidimensional assessment of global AGB product consistency from spatial,interannual variations,and ground-based perspectives.The results highlight that although GEDI-based and ICESat-2-based products ensure reliable spatial distribution,long-timeseries products,such as CCI and JPL,offer better interannual stability for tracking biomass dynamics.The DGVM outputs complement remote sensing data in capturing large-scale carbon flux trends but require further calibration in high-latitude regions.Overall,the findings provide a scientific basis for selecting,integrating,and applying multisource AGB datasets to improve the accuracy and reliability of carbon monitoring and ecological assessment at the global scale.

The rapid growth of Earth science observation and simulation data has made efficient data classification increasingly challenging, particularly under conditions of limited annotation resources and continuously evolving data semantics. Conventional classification methods rely heavily on large-scale labeled datasets, which are costly to construct and difficult to adapt to dynamic classification systems. This paper proposes a hierarchical classification framework for Earth science data that leverages large language models (LLMs) and explicitly incorporates hierarchical label relationships to constrain model inference and enhance classification consistency across complex, domain-specific semantic spaces. The framework further integrates retrieval-augmented generation (RAG) and knowledge graph (KG) techniques to introduce external domain knowledge and explicit semantic constraints, enhancing contextual understanding, interpretability, and adaptability to semantic evolution. A benchmark dataset with a two-level hierarchical label structure is constructed based on official NASA metadata. Experimental results demonstrate that by integrating few-shot learning and label space optimization strategies, the proposed framework steadily outperforms various baseline methods in hierarchical classification tasks. Compared with the Bert-BiLSTM model, it achieves an absolute improvement of 8.68% in Micro-F1 and 29.92% in Macro-F1 on the overall hierarchical paths. The framework demonstrates clear advantages in long-tailed data distributions, particularly for minority classes, highlighting its potential for scalable annotation and efficient management of large-scale Earth science datasets.

Rapid urbanization exerts profound pressure on urban biodiversity, yet long-term assessments integrating multi-source remote sensing data remain scarce. Objective: Focusing on the Hangzhou Bay Urban Agglomeration, a rapidly developing region in China’s Yangtze River Delta, this study aims to construct a remote sensing-based Biodiversity Index (BI) and analyze its spatiotemporal evolution and underlying drivers. Six Essential Biodiversity Variables derived from satellite observations (2000–2024) were integrated using Principal Component Analysis. Spatial autocorrelation and Geodetector models were then applied to examine BI dynamics and driving factors. The regional BI declined gradually from 0.80 in 2000 to 0.72 in 2024, with the rate of decline slowing after 2020 and a partial recovery observed in Zhoushan. Marked inter-city heterogeneity exists: Huzhou retains the highest and most stable BI due to extensive forest cover, whereas Jiaxing exhibits the lowest BI and the most pronounced decline, driven by rapid expansion of construction land. Land use/cover (LULC) and fractional vegetation cover (FVC) emerge as the dominant drivers (average q-values of 0.196 and 0.208, respectively), and their interaction explains over 46% of the spatial variance in BI. Road density shows a consistently increasing influence over time. This study demonstrates the utility of remote sensing-based frameworks for monitoring urban biodiversity dynamics and provides actionable insights for evidence-based land use planning and ecological restoration.

Study region: Wei River Basin, a typical erosion prone basin on the Chinese Loess Plateau, northwestern China. Study focus: Based on eight watersheds within Wei River Basin, the separate and joint contributions of climate, land surface composition (e.g. vegetation coverage, land use and land cover change) and landscape configuration were quantified across different spatial scales. Subsequently, their sensitivity were evaluated and threshold behaviors were identified. New hydrological insights for the region: Runoff and sediment reductions since the early 21st century were mainly associated with human activities. Across spatial scales, landscape configuration exhibited stronger independent contributions (approximately 15 similar to 21%) than precipitation (3%similar to 7%) and land surface composition (3%similar to 5%). Clear scale dependence was observed, with the highest explanatory power at the 500 m buffer scale, where combined factors explained up to 84.8% of runoff and sediment. Furthermore, sediment load showed stronger sensitivities than runoff to key factors. Among the tested factors, LSI was identified as the most sensitive factor, with a 1% increase being associated with relative increases of 1.60% in runoff and 2.46% in sediment load, respectively. Moreover, runoff and sediment load may undergo sudden changes once key landscape metrics exceed critical values. These findings reveal the importance of spatial structure and scale effects in regulating runoff and sediment processes in erosion dominated watersheds.

Leaf chlorophyll content (LCC) is a crucial parameter reflecting vegetation's photosynthetic activity. Many LCC inversion algorithms based on satellite and unmanned aerial vehicle (UAV) data have been developed in recent decades. The one-dimensional radiative transfer model, like PROSAIL (1D model), has been a classic tool for LCC inversion. In recent years, three-dimensional radiative transfer models (3D model) have been developed rapidly. However, studies on 3D models for LCC inversion are limited, and their impact on inversion accuracy across different sensor resolutions remains unclear. This study focuses on winter wheat and integrates the DART, AdelWheat, and PROSPECT models to construct the 3D-model-derived look-up table (LUT). The 3D-model-based LUT and 1D-model-based LUT were applied to Sentinel-2 (S2) and UAV data to retrieve LCC. Validation results demonstrate that the 3D-model-based algorithm significantly improves LCC inversion accuracy for both S2 and UAV images. For UAV data, the root mean square error (RMSE) decreases from 9.90 mu g/cm2 to 7.97 mu g/cm2, and the coefficient of determination (R2) improves from 0.70 to 0.79. For S2 data, the RMSE decreases from 12.40 mu g/cm2 to 8.68 mu g/cm2, while R2 increases from 0.66 to 0.85. Additionally, overestimation at low LAI levels and underestimation at high LCC levels are effectively reduced. The high accuracy achieved under varying LAI and LCC conditions allows the 3D model to capture temporal trends throughout the growing season better. The 3Dmodel-based LCC inversion algorithm can better utilize the high spatial resolution advantages, thereby playing a significant role in vegetation physiological monitoring and crop phenotyping.

The leaf chlorophyll content (LCC) is a crucial parameter indicating vegetation's photosynthetic activity. However, the resolution of current global LCC products ranges from 300m to 500m, and the existing 30m-resolution Multi-source data Synergized Quantitative remote sensing production system LCC (MuSyQ LCC) product is only available in China region, resulting in a lack of global high-resolution LCC products. This study used an empirical relationship method based on the chlorophyll sensitive index to produce a high-resolution global LCC product (MuSyQ Global LCC) using the Google Earth Engine platform. The validation result shows the accuracy of 500m-resolution MuSyQ Global LCC is slightly higher than the MODIS LCC and the accuracy improves when the resolution increases. The 10m-resolution LCC product has an RMSE of 15.33 mu g/cm(2), R-2 of 0.27. The high-resolution MuSyQ Global LCC product can show more detailed spatial distribution than the existing MODIS LCC product, indicating its ability in precision agriculture, forestry monitoring, and related research.

The ratio of leaf carotenoid to chlorophyll content (LCar/LCC) is a crucial indicator of vegetation physiological status. However, limited studies have investigated the retrieval of LCar/LCC at the canopy scale. Based on the spectral invariant theory, this study proposed a leaf-scale carotenoid-to-chlorophyll index (RCCI) for LCar/LCC estimation, which can be directly derived from canopy reflectance. Validation results showed that RCCI achieved an estimation accuracy of R2 = 0.57 and RMSE = 0.053 using ground-measured spectra, and R2 = 0.64 and RMSE = 0.118 using Sentinel-2 data. Compared to other VIs, RCCI reduced the RMSE of retrievals by 0.017-0.142. Moreover, RCCI consistently exhibited high and stable retrieval accuracy under varying LAI, LCC, and vegetation growth stages. These findings highlight the potential of RCCI for improving the accuracy of LCar/LCC estimation at the canopy scale.

The spectral invariants theory (p-theory) has received much attention in the field of quantitative remote sensing over the past few decades and has been adopted for modeling of canopy solar-induced chlorophyll fluorescence (SIF). However, the spectral invariant properties (SIP) in simple analytical formulae have not been applied for modeling canopy fluorescence anisotropy primarily because they are parameterized in terms of leaf total scattering, which precludes the differentiation between forward and backward leaf SIF emissions. In this study, we have developed the canopy-SIP SIF model by combining geometric-optical (GO) theory to account for asymmetric leaf SIF forward and backward emissions at the first-order scattering and by modeling multiple scattering based on thep-theory, thus avoiding the dependence on radiative transfer models. The applicability of the model simulations especially over 3D heterogeneous canopies was improved by incorporating canopy structure through multi-angular clumping index, and by modeling single scattering from the four components of the scene in view according to the GO approach. The results show good consistency with both the state-of-the-art SIF models and multi-angular field SIF observations over grass and chickpea canopies. The coefficient of determination (R2) between the simulated SIF and field measurements was 0.75 (red) and 0.74 (far-red) for chickpea, and 0.65 (both red and far-red) for grass. The average relative error was approximately 3 % for 1D homogeneous scenes when comparing the canopy-SIP SIF model simulations to the SCOPE model simulations, and around 4 % for the 3D heterogeneous scene when comparing to the LESS model simulations. The results indicate that the proposed approach for separating asymmetric leaf SIF emissions is a robust way to keep a balance between satisfactory simulation accuracy and efficiency. Model simulations suggest that neglecting the leaf SIF asymmetry can lead to an underestimation of canopy red SIF by 6.3 % to 42.6 % for various leaf biochemical and canopy structural parameters. This study presents a simple but efficient analytical approach for canopy fluorescence modeling, with potential for large-scale canopy fluorescence simulations.

Leaf chlorophyll content (LCC) is a crucial biochemical parameter for monitoring the plant's nutritional status and photosynthetic capacity. However, retrieving LCC from canopy reflectance is challenging due to the coupling influence of LCC and canopy structure, particularly leaf area index (LAI). The isolation of leaf-scale information from canopy signals is therefore essential to improve the LCC estimation. This study proposed an approach for deriving the leaf-scale chlorophyll index (CIleaf) from the canopy bidirectional reflectance factor (BRF) based on the spectral invariant theory (p-theory). Six widely used canopy-scale chlorophyll indices (CIcanopy) were selected to derive the corresponding CIleaf. The CIleaf is expressed as the product of its original CIcanopy and a scale conversion factor (SCF) (CIleaf = CIcanopy x SCF). The SCF is determined by two spectral invariants of p-theory (recollision probability p and directional area scattering factor DASF), as well as canopy BRFs at specific wavelengths, and it corrects for the contribution of canopy multiple scattering to CIcanopy. The analysis through radiative transfer model simulations showed that CIleaf exhibited more unified relationships with LCC across LAI conditions than the original CIcanopy and substantially eliminated the influence of LAI on the CI-based model. Validation results demonstrated that CIleaf improved the accuracy of LCC estimation compared to CIcanopy. The leaf-scale MERIS terrestrial chlorophyll index (MTCIleaf) exhibited the most prominent improvements, reducing the root-mean-square error (RMSE) by 6.68 mu g/cm2 for ground spectra and 2.33-4.21 mu g/cm2 for Sentinel-2 images with multi-ecosystem datasets. Additionally, the influence of vegetation types on the CI-based model was mitigated by CIleaf. MTCIleaf reduced the RMSE values by 3.8 %-34.0 % for different plant functional types, giving more consistent accuracies across species than MTCIcanopy. Our results show that the proposed CIleaf combines the robustness of the physically-based method with the simplicity of the CI-based method, thus providing a practical approach for large-scale high-resolution LCC mapping. Moreover, the method holds promise for designing leaf-scale vegetation indices sensitive to various leaf biochemical parameters beyond LCC, extending its utility to broader leaf-scale remote sensing retrieval (e.g., leaf carotenoid content and leaf dry mass).

Purpose Food security, of utmost global significance, is indelibly intertwined with national stability and citizen living quality. China has been resolutely committed to augmenting food security within host countries through outward foreign direct investment (OFDI). This paper aims to examine the impact of China’s OFDI on the food security of host countries. Design/methodology/approach Using panel data from 46 countries spanning from 2010 to 2021, the study establishes a food security index using the entropy method, encompassing three dimensions: food supply stability, food availability and the food production foundation, followed by an empirical analysis of the effect of China’s OFDI on the index. Findings The findings indicate that, for every one-unit increase in China’s OFDI, the food security index of the host country increases by 0.0078 units significantly. This impact is modulated by the degree of corruption prevailing in the host countries. Those countries with relatively lower levels of corruption are capable of optimizing the efficiency of foreign investment utilization, thereby magnifying the salutary impact of foreign investment on food security. Additionally, the trade scale and agricultural infrastructure of the host country are identified as the two major channels through which China’s OFDI promotes food security. Originality/value The paper concludes with some policy recommendations aimed at enhancing food security by enhancing the host country’s food security governance capacity and strengthening China’s OFDI and foreign cooperation.

Multimodal data fusion in Unmanned Aerial Vehicle (UAV) remote sensing (RS) has transformed many scientific research and industrial application domains, like environmental monitoring, precision agriculture, and urban planning, by integrating sensor data’s modalities such as RGB, Light Detection and Ranging (LiDAR), Synthetic Aperture Radar (SAR) and hyperspectral (HS). However, this is still a research area in its infancy. This paper systematically reviews strategies and deep learning (DL) architectures for UAV RS multimodal data fusion, focusing on analyzing sensor characteristics, fusion strategies, DL architectures, and practical challenges. The review critically analyses recent advancements and identifies future research directions to tackle issues like real-time processing, data heterogeneity, and fusion strategies. The findings aim to enhance the robustness and scalability of UAV-based RS applications, pushing the boundaries of their performance and applicability.