Selecting samples with high representativeness is very important in the accuracy assessment of land use and land cover (LULC) classification. By optimizing sampling sites in both feature space and geographical space, this paper presented a stratified even sampling method for accuracy assessment of the GlobeLand30 dataset for Beijing, China. In this method, spatial stratification and area-weighted proportion sampling assignment method were adopted to achieve optimal coverage in feature space; spatial-simulated annealing (SSA) and the minimization of the mean of the shortest distances (MMSD) criterion were used to optimize sampling sites in geographical space. Six sample sets with sizes of 150, 300, 450, 600, 750 and 900 were drawn using the proposed method, spatial even sampling method, stratified random sampling method and simple random sampling method, and their overall accuracy (OA), root-mean-square error (RMSE) and standard deviation (STDEV) values were evaluated. The results suggested that the OA, RMSE and STDEV results of the proposed method were 71.36%-73.91% (Mean, 72.26%), 0.90% and 0.96%, respectively. Compared with the other sampling methods, the average OA of the proposed method was much closer to the true OA and corresponding RMSE and STDEV were much lower than the other three sampling methods, respectively. It can improve the representativeness of both feature and geographical space, and provide a robust operational tool for the validation of LULC datasets.

Spatial sampling design is important for accurately assessing land use and land cover (LULC) classification results from remote sensing data. Spatial stratification can dramatically improve spatial sampling efficiency by dividing the study area into several strata when classification correctness is spatially stratified heterogeneous. By integrating the LULC classification results from different sources and spatial resolutions, a spatial stratification method for spatial sampling of accuracy assessment is presented in this paper. Its efficiency is demonstrated in the case study using LULC data of Beijing, China, in the following steps. First, we standardized and reclassified multiresolution remote sensing data, including China’s land use/cover datasets (CLUDs) from 2017 (resolution: 30 m), 500 m MCD12Q1, and 10 m FROM-GLC10 data, into six classes. Second, we customized stratification rules, formulated a technical specification to realize 11 strata using CLUDs and MCD12Q1, and employed FROM-GLC10 as the reference data for accuracy assessment. Furthermore, six sample sets with sizes of 16,417; 1821; 652; 337; 198; and 142 were drawn using different methods, and their overall accuracy (OA), deviation accuracy (DA), root-mean-square error (RMSE), and standard deviation (STDEV) values were also evaluated to demonstrate the efficiency brought by spatial stratification. Compared with the spatial even sampling method, the OAs of the stratified even sampling method adopting the proposed stratification method was much closer to the true OA, and the corresponding RMSE and STDEV results decreased from 2.097% and 2.127% to 0.914% and 0.713%, respectively, due to the contribution of spatial stratification in the sampling scheme. The method can be used to distinguish the differences and improve the representativeness of samples, and it can be employed to select validation samples for LULC classification.

Soil organic matter (SOM) plays pivotal roles in characterizing dryland structure and function; however, remotely sensed spatially-detailed SOM mapping in these regions remains a challenge. Various digital soil mapping approaches based on either single-period remote sensing or spectral indices in other ecosystems usually produce inaccurate, poorly constrained estimates of dryland SOM. Here, a framework for spatially-detailed SOM mapping was proposed based on cross-wavelet transform (XWT) that exploits ecologically meaningful features from intra-annual fractional vegetation and soil-related endmember records. In this framework, paired green vegetation (GV) and soil-related endmembers (i.e., dark surface (DA), saline land (SA), sand land (SL)) sequences were adopted to extract 30 XWT features in temporally and spatially continuous domains of cross-wavelet spectrum. We then selected representative features as exploratory covariates for SOM mapping, integrated with four state-of-the-art machine learning approaches, i.e., ridge regression (RR), least squares-support vector machines (LS-SVM), random forests (RF), and gradient boosted regression trees (GBRT). The results reported that SOM maps from 13 coupled filtered XWT features and four machine learning approaches were consistent with soil-landscape knowledge, as evidenced by a spatially-detailed gradient from oasis to barren. This framework also presented more accurate and reliable results than arithmetically averaged features of intra-annual endmembers and existing datasets. Among the four approaches, both RF and GBRT were more appropriate in the XWT-based framework, showing superior accuracy, robustness, and lower uncertainty. The XWT synthetically characterized soil fertility from the consecutive structure of intra-annual vegetation and soil-related endmember sequences. Therefore, the proposed framework improved the understanding of SOM and land degradation neutrality, potentially leading to more sustainable management of dryland systems.

Identifying influencing factors of heavy metals is essential for soil evaluation and protection. This study investigates the use of a geographical detector to identify influencing factors of agricultural soil heavy metals from natural and anthropogenic aspects. We focused on six variables of soil heavy metals, i.e., As, Cd, Hg, Cu, Pb, Zn, and four influencing factors, i.e., soil properties (soil type and soil texture), digital elevation model (DEM), land use, and annual deposition fluxes. Experiments were conducted in Shunyi District, China. We studied the spatial correlations between variables of soil heavy metals and influencing factors at both single-object and multi-object levels. A geographical detector was directly used at the single-object level, while principal component analysis (PCA) and geographical detector were sequentially integrated at the multi-object level to identify influencing factors of heavy metals. Results showed that the concentrations of Cd, Cu, and Zn were mainly influenced by DEM (p = 0.008) and land use (p = 0.033) factors, while annual deposition fluxes were the main factors of the concentrations of Hg, Cd, and Pb (p = 0.000). Moreover, the concentration of As was primarily influenced by soil properties (p = 0.026), DEM (p = 0.000), and annual deposition flux (p = 0.000). The multi-object identification results between heavy metals and influencing factors included single object identification in this study. Compared with the results using the PCA and correlation analysis (CA) methods, the identification method developed at different levels can identify much more influencing factors of heavy metals. Due to its promising performance, identification at different levels can be widely employed for soil protection and pollution restoration.

The quick and accurate extraction of information on woodland resources and distributions using remote sensing technology is a key step in the management, protection, and sustainable use of woodlands. This paper presents a low-cost and high-precision extraction method for large woodland areas based on the fractal features of the Moderate Resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index (NDVI) time series data for Beijing, China. The blanket method was used for computing the upper and lower fractal signals of each pixel in the NDVI time series images. The fractal signals of woodlands and other land use/land cover types at corresponding scales were analyzed and compared, and the attributes of woodlands were enhanced at the fifth lower fractal signal. The spatial distributions of woodlands were extracted using the Iterative Self-Organizing Data Analysis technique (ISODATA), and an accuracy assessment of the extracted results was conducted using the China Land Use and Land Cover Data Set (CLUCDS) from the same period. The results showed that the overall accuracy, kappa coefficient, and error coefficient were 90.54%, 0.74, and 8.17%, respectively. Compared with the extracted results for woodlands using the MODIS NDVI time series only, the average error coefficient decreased from 30.2 to 7.38% because of these fractal features. The method developed in this study can rapidly and effectively extract information on woodlands from low spatial resolution remote sensing data and provide a robust operational tool for use in further research.

The first 10 m resolution global land-cover map, FROM-GLC10 was released in early March 2019. Due to its high spatial resolution and reliable global accuracy, we evaluated the suitability of FROM-GLC10 data for understanding agricultural ecosystems in Beijing using a comparable vector data set, Google Earth images and field survey data. The overall accuracy (OA) for FROM-GLC10 based on three data sets was 71.08%, 79.63% and 80.36%, respectively. Meanwhile, there were notable misclassifications between cropland, grassland and forest. The limited accuracy for these vegetation types might be attributed to the spatially mixed vegetation structures, seasonal variations of vegetation and the temporal inconsistence between Landsat and Sentinel data set. Due to its satisfactory OA, FROM-GLC10 data have the potential to be widely employed for evaluating the progress of large-scale ecological restoration projects. On the other hand, the data producers, who can pre-set and classify some customized land-cover types, consider time-series analysis and big data fusion methods and conduct large-scale verification, and data users, who can integrate different data sources, should work together to enhance the suitability and reliability of FROM-GLC10 data for understanding agricultural ecosystems in China.