Qiaomei Su

Crop identification with remote sensing images has currently emerged as prominent research topic with a focus on flat regions dominated by crop cultivation. However, farmland is also distributed in vast complex terrain areas where the study on crop identification is necessary and of importance. This study aims to propose innovative method to address the challenges by investigating what complex terrain posed on the feature extraction and classification algorithms. Shouyang Loess Hill and Jinzhong Basin were chosen as the study area. Three indices, topographic indices, vegetation indices and texture indices were extracted from ALOS DEM and time-series Sentinel-2 spectral bands, respectively. Various feature combinations of spectral bands and three indices were ingested into the algorithm of Random Forest (RF), and then five models were built to investigate the impact of various feature combinations on the accuracy of crop identification in both hilly and flat areas. Results indicate that the combination of multiple features can significantly improve the accuracy of crop identification in complex terrain area while the differences in flat region were less pronounced. This study revealed that the combination of diverse features with topographic features is an effective method to improve the accuracy of crop identification in complex terrain area. The high accuracy crop maps can provide valuable insights for optimizing agricultural practices and resource management in complex terrain area. The future study should extend this method to other more complex terrain area with efficient optimization of topographic features.

Surface subsidence induced by underground coal mining is a primary driver of ecological degradation. The traditional Remote Sensing Ecological Index (RSEI), however, struggles to capture surface deformation constraints and vegetation response lags. To address this, we developed a Mining Deformation–Ecology Coupling Index (MDECI). This index integrates Interferometric Synthetic Aperture Radar (InSAR)-monitored surface stability with multi-spectral indicators via Principal Component Analysis (PCA). We applied this method to the Datong Coalfield, China, using 231 Sentinel-1A SAR scenes and 8 Landsat images (2017–2024) to validate the effectiveness of the index. Meanwhile, we systematically analyzed non-linear response mechanisms, the Ecological Turning Point (ETP), and spatial clustering characteristics. The results demonstrate the following: (1) InSAR and MDECI effectively identified patterns of surface subsidence and ecological decline. Subsidence centers expanded to a maximum of −2085 mm, causing the mean MDECI in these areas to drop to 0.185 (<−1800 mm). This represents a 57.4% decrease relative to the regional average (0.434). (2) MDECI outperformed traditional models with a stable Average Correlation Coefficient (ACC) (0.63–0.75) and high cross-correlation coefficients with RSEI (0.906) and the Mine-specific Eco-environment Index (MSEEI) (0.931). During the 2018 drought, MDECI maintained a robust ACC of 0.628 while RSEI dropped to 0.482. (3) Multi-scale analysis revealed a unimodal MDECI response with an ETP at −100 mm. Initial ‘micro-disturbance gain’ (0.371 to 0.471) is followed by a progressive decline to a minimum of 0.185 under severe deformation. (4) Local Indicators of Spatial Association (LISA) spatial clustering characterized the distribution patterns of ecological damage and localised high-maintenance areas. High–Low damaged areas accounted for 5.09%, while High–High high-maintenance areas reached 9.00%. The scale of High–High areas was approximately 1.77 times that of the damaged areas. The MDECI addresses the deficiencies of traditional indices in high-disturbance areas and isolates the impact of mining on the ecology, providing a quantitative basis for risk identification and differentiated restoration.

Global climate change is increasing extreme rainfall events, and severe floods are becoming more frequent. Flood storage and detention basins (FSDBs) are an important part of the flood control system in China. They play a key role in regional flood emergency response and regulation. Therefore, accurate simulation of flood evolution after the activation of FSDBs is urgently needed. This study proposes a high-accuracy flood evolution simulation method that combines terrain clustering and physical propagation constraints. We first build a 2 m resolution digital elevation model (DEM) using GF-7 stereo imagery and laser altimetry data. We then introduce an improved superpixel segmentation algorithm (TSLIC). This method reduces the number of computational units while preserving key micro-topographic features. It groups high-resolution grids into terrain units with similar elevation characteristics and continuous spatial structure. Based on these terrain units, we develop a flood evolution model called RS-CFPM. The model combines flow velocity estimated from the Manning equation with flood propagation speed derived from radar remote sensing. It uses a water balance framework and includes a propagation time delay constraint. This design helps overcome the limitation of traditional static inundation methods that ignore flood travel time. We apply the proposed method to simulate the flood inundation process during the “23·7” extreme basin-scale flood event in the Haihe River Basin. Comparison with multi-temporal radar observations shows that the errors of simulated water level and inundation extent in the Dongdian FSDB are both within 10%. The computational efficiency is also improved by more than 60% compared with traditional methods. This study provides a new approach for rapid and accurate simulation of flood inundation processes in FSDBs under emergency conditions. The method can support flood emergency operation and decision-making.

Under the influence of extreme climatic conditions, landslide disasters occur frequently in the Loess Plateau due to complex geological structures, loose soil, and frequent intense rainfall. These events are often concealed, posing significant challenges for disaster prevention. High-resolution optical remote sensing combined with field surveys can improve identification accuracy; however, concerns persist regarding issues such as omission and misidentification during hazard identification and monitoring processes. To address these challenges, this study proposes an integrated remote-sensing identification approach, focusing specifically on the central region of Tianshui, a typical landslide-prone area within the Loess Plateau. Utilizing Sentinel-1 and JL1LF01A remote-sensing imagery collected from 2022 to 2023, we conducted ground deformation monitoring through the Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) technique. By integrating deformation results with optical imagery features indicative of potential landslide sites, a comprehensive identification method was developed to precisely detect potential landslide hazards. Verification of the identified sites was subsequently performed using the Google Earth platform, resulting in the establishment of a final dataset of potential landslide hazards within the study area. This outcome clearly demonstrates the high applicability and accuracy of the integrated remote-sensing identification method in the context of landslide hazard assessment. Furthermore, this research provides a solid scientific foundation for geological hazard identification efforts and plays a critical guiding role in disaster prevention and mitigation in Tianshui City, thereby enhancing the region’s capacity to withstand disaster risks and effectively safeguarding local lives and property.

Soil moisture (SM) data can provide guidance for decision-makers in fields such as drought monitoring and irrigation management. Soil Moisture Active Passive (SMAP) satellite offers sufficient spatial resolution for global-scale applications, but its utility is limited in regional areas due to its lower spatial resolution. To address this issue, this study proposed a downscaling framework based on the Stacking strategy. The framework integrated extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), and categorical boosting (CatBoost) to generate 1 km resolution SM data using 15 high-resolution factors derived from multi-source datasets. In particular, to test the influence of terrain partitioning on downscaling results, Anhui Province, which has diverse terrain features, was selected as the study area. The results indicated that the performance of the three base models varied, and the developed Stacking strategy maximized the potential of each model with encouraging downscaling results. Specifically, we found that: (1) The Stacking model achieved the highest accuracy in all regions, and the performance order of the base models was: XGBoost > CatBoost > LightGBM. (2) Compared with the measured SM at 87 sites, the downscaled SM outperformed other 1 km SM products as well as the downscaled SM without partitioning, with an average ubRMSE of 0.040 m³/m³. (3) The downscaled SM responded positively to rainfall events and mitigated the systematic bias of SMAP. It also preserved the spatial trend of the original SMAP, with higher levels in the humid region and relatively lower levels in the semi-humid region. Overall, this study provided a new strategy for soil moisture downscaling and revealed some interesting findings related to the effectiveness of the Stacking model and the impact of terrain partitioning on downscaling accuracy.

Irrigation is one of the key agricultural management practices of crop cultivation in the world. Irrigation practice is traceable on satellite images. Most irrigated area mapping methods were developed based on time series of NDVI or backscatter coefficient within the growing season. However, it has been found that winter irrigation out of growing season is also dominating in north China. This kind of irrigation aims to increase the soil moisture for coping with spring drought and reduce the wind erosion in spring. This study developed a remote sensing-based classification approach to identify irrigated fields out of growing season with Radom Forest algorithm. Four spectral bands and all Normalized Difference Vegetation Index (NDVI) like indices computed from any two of these four bands for each of the seven scenes of GF-1 satellite data were used as the input features in the building of separated RF models and in applying the built models for the classification. The results showed that the mean of the highest out-of-bag accuracies for seven RF models was 94.9% and the mean of the averaged out-of-bag accuracies in the plateau for seven RF models was 94.1%; the overall accuracy for all seven classified outputs was in the range of 86.8–92.5%, Kappa in the range of 84.0–91.0% and F1-Score in the range of 82.1–90.1%. These results showed that the classification was neither overperformed nor underperformed as the accuracies of all classified images were lower than the model ones. This study also found that irrigation started to be applied as early as in November and irrigated fields were increased and suspended in December and January due to freezing conditions. The newly irrigated fields were found again in March and April when the temperature rose above zero degrees. The area of irrigated fields in the study area were increasing over time with sizes of 98.6, 166.9, 208.0, 292.8, 538.0, 623.1, 653.8 km² from December to April, accounting for 6.1%, 10.4%, 12.9%, 18.2%, 33.4%, 38.7%, and 40.6% of the total irrigatable land in the study area, respectively. The results showed that the method developed in this study performed well. This study found on the satellite images that 40.6% of irrigatable fields were already irrigated before the sowing season and the irrigation authorities were supposed to improve their water supply capacity in the whole year with this information. This study may complement the traditional consideration of retrieving irrigation maps only in growing season with remote sensing images for a large area.

Irrigation is one of key agricultural management practices of crop cultivation in the world. Irrigation practice is traceable on satellite image. Most irrigated area mapping methods were developed based on time series of NDVI or back scatter coefficient within the growing season. However, it found the winter irrigation out of growing season is also dominating in north China. This kind of irrigation aims to increase the soil moisture for coping with spring drought, and also reduce the wind erosion with the wet soil in the field surface as the strong wind happens in spring. This study developed a remote sensing-based classification approach to identify irrigated fields with Radom Forest algorithm out of growing season. The results showed that the mean of the highest accuracies of 7 RF models was 94.9% and the mean of the averaged accuracies of 7 RF models was 94.1%; the overall accuracy for all 7 outputs was in the range of 86.8-92.5%, Kappa in the range of 84.0-91.0% and F-1 score in the range of 82.1-90.1%. These results showed that the classification was acceptable and not over performed as the accuracies of all classified images were lower than the models. This study also found that irrigation started to apply in early November and irrigated fields were increased and suspended in December and January due to freeze. The newly irrigated fields were found again in March and April when the temperature goes up above zero degree. The area of irrigated fields in the study area were increasing over time with sizes of 98.6, 166.9, 208.0, 292.8, 538.0, 623.1, 653.8 km2 from December to April, accounting for 6.1%, 10.4%, 12.9%, 18.2%, 33.4%, 38.7%, and 40.6% of the total irrigatable land in the study area, respectively.

The main purpose of this study is to establish an effective landslide susceptibility zoning model and test whether underground mined areas and ground collapse in coal mine areas seriously affect the occurrence of landslides. Taking the Fenxi Coal Mine Area of Shanxi Province in China as the research area, landslide data has been investigated by the Shanxi Geological Environment Monitoring Center; adopting the 5-fold cross-validation method, and through Geostatistics analysis means the datasets of all non-landslides and landslides were divided into 80:20 proportions randomly for training and validating models. A set of 15 condition factors including terrain, geological, hydrological, land cover, and human engineering activity factors (distance to road, distance to mined area, ground collapse density) were selected as the evaluation indices to construct the susceptibility assessment model. Three machine learning algorithms for landslide susceptibility prediction (LSP) including C5.0 Decision Tree (C5.0), Random Forest (RF), and Support Vector Machine (SVM) have been selected and compared through the Areas under the Receiver Operating Characteristics (ROC) Curves (AUC), and several statistical estimates. The study revealed that for these three models the value range of prediction accuracies vary from 83.49 to 99.29% (in the training stage), and 62.26–73.58% (in the validation stage). In the two stages, AUCs are between 0.92 to 0.99 and 0.71 to 0.80 respectively. Using Jenks Natural Breaks algorithm, three LSPs levels are established as very low, low, medium, high, and very high probability of landslide by dividing the indices of the LSP. Compared with RF and SVM, C5.0 is considered better in five categories according to quantities and distribution of the landslides and their area percentage for different LSP zones. Four factors such as distance to road, lithology, profile curvature, and ground collapse density are the most suitable condition factors for LSP. The distance to mine area factor has a medium contribution and plays an obvious role in the occurrence of landslides in all the models. The result reveals that C5.0 possesses better prediction efficiency than RF and SVM, and underground mined area and ground collapse sifnigicantly affect significantly the occurrence of landslides in the Fenxi Coal Mine Area.

Landslide susceptibility mapping is the first and most important step involved in landslide hazard assessment. The purpose of the present study is to compare three nonlinear approaches for landslide susceptibility mapping and test whether coal mining has a significant impact on landslide occurrence in coal mine areas. Landslide data collected by the Bureau of Land and Resources are represented by the X, Y coordinates of its central point; causative factors were calculated from topographic and geologic maps, as well as satellite imagery. The five-fold cross-validation method was adopted and the landslide/non-landslide datasets were randomly split into a ratio of 80:20. From this, five subsets for 20 times were acquired for training and validating models by GIS Geostatistical analysis methods, and all of the subsets were employed in a spatially balanced sample design. Three landslide models were built using support vector machine (SVM), logistic regression (LR), and artificial neural network (ANN) models by selecting the median of the performance measures. Then, the three fitted models were compared using the area under the receiver operating characteristics (ROC) curves (AUC) and the performance measures. The results show that the prediction accuracies are between 73.43% and 87.45% in the training stage, and 67.16% to 73.13% in the validating stage for the three models. AUCs vary from 0.807 to 0.906 and 0.753 to 0.944 in the two stages, respectively. Additionally, three landslide susceptibility maps were obtained by classifying the range of landslide probabilities into four classes representing low (0–0.02), medium (0.02–0.1), high (0.1–0.85), and very high (0.85–1) probabilities of landslides. For the distributions of landslide and area percentages under different susceptibility standards, the SVM model has more relative balance in the four classes compared to the LR and the ANN models. The result reveals that the SVM model possesses better prediction efficiency than the other two models. Furthermore, the five factors, including lithology, distance from the road, slope angle, elevation, and land-use types, are the most suitable conditioning factors for landslide susceptibility mapping in the study area. The mining disturbance factor has little contribution to all models, because the mining method in this area is underground mining, so the mining depth is too deep to affect the stability of the slopes.