Geographically weighted regression (GWR) is a classic local linear method for modeling spatial non-stationarity that is applied in various geographical scenarios. The modeling of spatial proximity in traditional GWR and its variants is usually based on various forms of spatial distances to construct spatial weights, overlooking the potential effect of multidimensional attribute similarity of physical entities. Therefore, we proposed the geographically spatial-attribute weighted regression (GSAWR) method with convolutional neural networks to account for spatial non-stationarity based on spatial proximity and attribute similarity. An attribute fusion convolutional neural network (AFCNN) considers the differential effects of attribute variables by assessing similarities among multiple variables. A spatial-attribute joint proximity neural network (SAJPNN) combines attribute similarity and spatial proximity to generate a proximity measure adaptive to both spatial proximity and attribute similarity. A spatial-attribute weighted convolutional neural network (SAWCNN) and ordinary linear regression (OLR) use the spatial-attribute joint proximity to make final predictions. We validated the GSAWR approach on simulated dataset and two real-world datasets: the PM2.5 and HIV datasets. The results revealed that the GSAWR model outperformed the other baseline models in terms of fitting and prediction performance. Ablation experiments and coefficient visualization further determined the effectiveness and interpretability of GSAWR model.

The accurate prediction of sea surface temperature (SST) is highly important for climate change research and the management of marine ecosystems. Traditional numerical models rely on complex physical processes and precise initial conditions, resulting in high computational costs and limited generalizability. Although deep learning methods can improve the physical consistency and interpretability of ocean processes by incorporating physical constraints, their performance may degrade under anomalous or extreme SST conditions, where rigid constraints limit the model’s adaptability to complex variations. In contrast, incorporating motion-aware mechanisms enables models to flexibly capture dynamic patterns from data, thus increasing their responsiveness to nonstationary processes. Therefore, we propose a novel SST prediction model that integrates optical flow guidance with spatiotemporal awareness, aiming to enhance the modeling of motion and spatiotemporal features in SST evolution. The proposed model consists of three key modulesa spatiotemporal information extraction module (SIEM), a motion trend extraction module (MTEM), and a spatiotemporal feature fusion module (SFFM). First, the SIEM, composed of multiple layers of SwinLSTM, captures spatial and temporal dependencies in the SST time series. The MTEM then estimates the optical flow to extract motion information from the SST data. Finally, the motion information is used to dynamically adjust the spatiotemporal features, which are then fused in the SFFM. To evaluate the performance of the proposed model, we conducted SST prediction experiments at both daily (1–10 days) and weekly (1–10 weeks) scales over the South China Sea (SCS) and the Pacific Ocean, and compared the results with several existing models. The experimental results demonstrate that our model outperforms existing methods across multiple evaluation metrics, with superior prediction accuracy and robustness.

Location prediction based on social media information is highly valuable in human mobility research and has multiple real-life applications. However, existing research methods often ignore social influences, largely ignoring implicit information regarding interactions between users and geographical locations. Additionally, they generally employ single modeling structures, which restricts the effective integration of complex spatiotemporal characteristics and factors influencing user mobility. In this context, we propose a novel network with physical and social awareness that expresses both physical and social influences of user mobility from a global perspective based on a heterogeneous graph constructed using users and spatial locations as nodes and relationships between them as edges. This graph enables the model to leverage information from connected nodes and edges to infer missing or unobserved data. The model predicts future locations of users by effectively integrating the temporal and spatial features of user trajectory series. The proposed model is validated using three social media datasets. The experimental results demonstrate that the proposed method outperforms the state-of-the-art baseline models. This indicates the importance of considering complex interactions between users and locations, as well as the various influences of physical and social spaces.

Vegetation is a key component of biodiversity and ecosystem stability. The normalized difference vegetation index (NDVI) is widely used to monitor the vegetation growth status. Timely spatiotemporal prediction of NDVI is of great significance to agriculture and the ecological environment. Traditional deep learning models neglect the local spatial relationships and the effect of spatial autocorrelation in vegetation, failing to capture the neighboring contributions to the forecast target. In this study, a geographically weighted convolutional long short-term memory (GWConvLSTM) model is coupled with the spatial weight and the convolutional long short-term memory (ConvLSTM) model by the Hadamard product is proposed to predict NDVI. In the GWConvLSTM model, a spatiotemporal memory flow with the ConvLSTM unit is designed to calculate the spatial weight of each step, which is iteratively updated with the training of the model. We use the proposed GWConvLSTM model to forecast the monthly NDVI of Huangpi District and the experiment results indicate that the developed GWConvLSTM model outperforms five other baseline models. The predictive coefficient of determination ($R^{2}$) reached up to 0.8427, 0.8209, and 0.8149 for one-, three-, and five-month lead times, respectively. The mean absolute error (MAE) decreased by 14.26%, 10.04%, and 16.75% for GWConvLSTM versus the baseline ConvLSTM model in one-, three-, and five-month NDVI predictions, respectively. Therefore, the developed GWConvLSTM model could be an effective geospatial deep learning model for spatiotemporal NDVI prediction at region scales.

From a process perspective, a flood includes several phases with distinguishable features. Fine-grained multisource data for different flood phases can be used to inform decision-making as flooding progresses. Therefore, the aim of this study was to develop an integrated framework based on human perceptions to progressively profile floods, including flood process segmentation rules (FPSR), flood severity index (FSI) and flood process perception ontology (FPPO). FPSR identifies flood phases based on specific signals in multisource data and provides spatiotemporal process information to FPPO consistent with flood perception. FSI follows FPSR to evaluate flooding throughout its evolution process. The comparison between FPSR and the flood monitoring index (IF) demonstrates that FPSR can detect flood events and segment the flooding process into latency, onset, development and recovery phases. The correlations between the standardized antecedent precipitation index (SAPI) and FSI show that FSI can assess flood severity with both natural and social effects in every flooding phase (R2 = 0.726 and 0.673 for the 2016 and 2020 floods, respectively). An experiment finds that flood events in Wuhan, China, usually begin in mid-to-late June and are the most severe in July, when more caution is needed for flood prevention and mitigation.

Early warning of cyanobacteria bloom is very important for water ecological management of lakes. Thus, Based on Landsat8 OLI data, combining the trophic state index (TSI), cyanobacteria and macrophytes index (CMI) and floating algae index (FAI), we proposed a risk zoning framework of cyanobacteria bloom, and applied it in Lake Longgan (China). Eutrophication frequency of Lake Longgan has worsened since 2017. But the maximum eutrophication proportion occurred on 16 February 2016. With a deeper analysis on 16 February 2016, we find that the risk area of cyanobacteria bloom in Lake Longgan is distributed horizontally from the east bank to the center of the lake, and mostly is mid risk. High risk areas are mainly distributed in the east coast, showing a southwest direction. Therefore, the proposed risk zoning framework provides a useful approach to obtain the pre-warning information of cyanobacteria bloom in some Eutrophic and macrophytic lakes.

The sensor web is a new method in the earth observation field, comprising sensors that can sense, compute, and correspond with the World Wide Web. The Chinese Integrated Earth Observation Sensor Web (CIEOSW) project is a five-year national basic research program conducted by Wuhan University since 2011 and is aimed to address the following major challenges: (1) the lack of collaboration within the satellite observation system, (2) the absence of a coupling mechanism in heterogeneous spaceborne -- airborne -- ground sensors, and (3) the limited connection between monitoring and decision-support services. For two years, the CIEOSW project has strived to achieve the following: (1) a theory on the coupling and modeling of the Earth observation sensor web (EOSW), (2) an event-driven multi-sensor collaborative observation method, (3) an EOSW fusion and assimilation method, (4) an EOSW-based information extraction and rapid change detection method, (5) a task-oriented EOSW focusing service model, and (6) an epicontinental environment observation and analysis of typical areas in China.