Land cover and its dynamic information is the basis for characterizing surface conditions, supporting land resource management and optimization, and assessing the impacts of climate change and human activities. In land cover information extraction, the traditional convolutional neural network (CNN) method has several problems, such as the inability to be applied to multispectral and hyperspectral satellite imagery, the weak generalization ability of the model and the difficulty of automating the construction of a training database. To solve these problems, this study proposes a new type of deep convolutional neural network based on Landsat-8 Operational Land Imager (OLI) imagery. The network integrates cascaded cross-channel parametric pooling and average pooling layer, applies a hierarchical sampling strategy to realize the automatic construction of the training dataset, determines the technical scheme of model-related parameters, and finally performs the automatic classification of remote sensing images. This study used the new type of deep convolutional neural network to extract land cover information from Qinhuangdao City, Hebei Province, and compared the experimental results with those obtained by traditional methods. The results show that: (1) The proposed deep convolutional neural network (DCNN) model can automatically construct the training dataset and classify images. This model performs the classification of multispectral and hyperspectral satellite images using deep neural networks, which improves the generalization ability of the model and simplifies the application of the model. (2) The proposed DCNN model provides the best classification results in the Qinhuangdao area. The overall accuracy of the land cover data obtained is 82.0%, and the kappa coefficient is 0.76. The overall accuracy is improved by 5% and 14% compared to the support vector machine method and the maximum likelihood classification method, respectively.

Grassland gross primary productivity (GPP) is an important part of global terrestrial carbon flux, and its accurate simulation and future prediction play an important role in understanding the ecosystem carbon cycle. Machine learning has potential in large-scale GPP prediction, but its application accuracy and impact factors still need further research. This paper takes the Mongolian Plateau as the research area. Six machine learning methods (multilayer perception, random forest, Adaboost, gradient boosting decision tree, XGBoost, LightGBM) were trained using remote sensing data (MODIS GPP) and 14 impact factor data and carried out the prediction of grassland GPP. Then, using flux observation data (positions of flux stations) and remote sensing data (positions of non-flux stations) as reference data, detailed accuracy evaluation and comprehensive trade-offs are carried out on the results, and key factors affecting prediction performance are further explored. The results show that: (1) The prediction results of the six methods are highly consistent with the change tendency of the reference data, demonstrating the applicability of machine learning in GPP prediction. (2) LightGBM has the best overall performance, with small absolute error (mean absolute error less than 1.3), low degree of deviation (root mean square error less than 3.2), strong model reliability (relative percentage difference more than 5.9), and a high degree of fit with reference data (regression determination coefficient more than 0.97), and the prediction results are closest to the reference data (mean bias is only −0.034). (3) Enhanced vegetation index, normalized difference vegetation index, precipitation, land use/land cover, maximum air temperature, potential evapotranspiration, and evapotranspiration are significantly higher than other factors as determining factors, and the total contribution ratio to the prediction accuracy exceeds 95%. They are the main factors influencing GPP prediction. This study can provide a reference for the application of machine learning in GPP prediction and also support the research of large-scale GPP prediction.

Ecological land is a type of land that has considerable ecological value. Understanding the evolution of urban ecological land in Zhuhai, China, holds great significance for revealing the evolution of ecological land in the Dawan District of southern China. We explored the temporal and spatial variation in urban ecological land in Zhuhai using the transformation matrix, equivalent ecological land, landscape index and ecological land center of gravity migration methods. Multivariate logistic regression was used to analyze the mechanism of ecological land change, and a transition probability map of the ecological land in the study area was drawn. The results showed the following. (1) From 1991 to 2018, the area of ecological land in Zhuhai city continuously decreased, with a reduction in area of 274.8 km², or 32.3%. Sharp changes mainly occurred from 1991 to 2000. (2) The ecological land in the study area has gradually become fragmented, and the degree of landscape heterogeneity has increased. Affected by the expansion of the outer edge of the city to the southwest and the construction of ecological land within the city, the center of gravity of the ecological land has shifted to the northeast by 1346 m. (3) The elevation, slope, distance from built-up land and growth rate of built-up land are important factors influencing the transformation of ecological land. In the future, rivers and shallow coastal waters, tidal flats, and grasslands in the study area have the highest probability of transformation. The Jinwan District and Xiangzhou District will face severe ecological land protection pressure. The method of spatial–temporal analysis of urban ecological land developed in this paper can be applied in similar studies on other cities, and the results obtained for Zhuhai, China, have reference value for future urban planning and ecological protection work.

Nighttime light images record the brightness of the Earth surface, indicating the scope and intensity of human activities. However, there are few studies on the long-term changes in global nighttime lights. In this paper, the authors constructed a long time series (1992~2017) nighttime light dataset combining the Defense Meteorological Satellites Program/Operational Linescan System (DMSP-OLS) and the Suomi National Polar-Orbiting Partnership Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) data sources and observed the following: (1) Global nighttime lights have become brighter. The global nighttime brightness in 2017 was 2.2 times that of 1992. Approximately 40.3% of the lighted area was significantly brightened, and an area of 1.3 × 10⁷ km² transitioned from an unlighted area to a lighted area. (2) Approximately 85.7% of the nighttime light increase occurred in the low-brightness zone (LBZ). Therefore, global brightness has become more uniform than before. (3) China, India, and the United States have led the global lighting trend. The increase in Chinese nighttime lights is the largest, with an average annual growth of 6.48%, followed by the light growth in India, while the United States has the largest brightened area. (4) The changes in nighttime lights in developing countries (e.g., China and India) are closely and positively related to their electricity consumption, industrial added value and gross domestic product (GDP). The shift of the LBZ center from Asia to Africa indicates the intercontinental transition of poverty.

The farming–pastoral zone in northern China is one of the most ecologically sensitive areas globally, having experienced extensive cropland abandonment and land reclamation over decades, primarily influenced by policy adjustment and global warming. However, the spatiotemporal patterns and suitability of long-term cropland change remain poorly understood. Using the annual China land cover dataset (CLCD), we provide a cropland abandonment and land reclamation mapping approach based on actual land use processes (rather than land cover conditions) to investigate spatiotemporal features of abandonment and reclamation and evaluate the rationality. Our findings show that: (1) Returning farmland to forest and grassland has been a clear trend in the study area over the past 30 years. Specifically, cropland use has undergone three phases of change, i.e., cropland contraction and expansion alternately (before 2000), followed by substantial abandonment (after 2000), and low-intensity reclamation (after 2010). (2) In the last decade, the intensity of the abandonment of cropland with high and moderate suitability is low. The rate of abandonment decreased, while the intensity of land reclamation was relatively high. The rate of the reclamation increased, and the spatial distribution of cropland tended to be reasonable. Our study emphasizes the importance of monitoring actual cropland changes based on land use processes, and this method can be effectively extended to regional or global long-term cropland monitoring.