Urban building segmentation is a prevalent research domain for very high resolution (VHR) remote sensing; however, various appearances and complicated background of VHR remote sensing imagery make accurate semantic segmentation of urban buildings a challenge in relevant applications. Following the basic architecture of U-Net, an end-to-end deep convolutional neural network (denoted as DeepResUnet) was proposed, which can effectively perform urban building segmentation at pixel scale from VHR imagery and generate accurate segmentation results. The method contains two sub-networks: One is a cascade down-sampling network for extracting feature maps of buildings from the VHR image, and the other is an up-sampling network for reconstructing those extracted feature maps back to the same size of the input VHR image. The deep residual learning approach was adopted to facilitate training in order to alleviate the degradation problem that often occurred in the model training process. The proposed DeepResUnet was tested with aerial images with a spatial resolution of 0.075 m and was compared in performance under the exact same conditions with six other state-of-the-art networks—FCN-8s, SegNet, DeconvNet, U-Net, ResUNet and DeepUNet. Results of extensive experiments indicated that the proposed DeepResUnet outperformed the other six existing networks in semantic segmentation of urban buildings in terms of visual and quantitative evaluation, especially in labeling irregular-shape and small-size buildings with higher accuracy and entirety. Compared with the U-Net, the F1 score, Kappa coefficient and overall accuracy of DeepResUnet were improved by 3.52%, 4.67% and 1.72%, respectively. Moreover, the proposed DeepResUnet required much fewer parameters than the U-Net, highlighting its significant improvement among U-Net applications. Nevertheless, the inference time of DeepResUnet is slightly longer than that of the U-Net, which is subject to further improvement.

Change detection (CD) with hyperspectral images (HSIs) can be effectively performed using deep learning networks (DLNs) by taking advantage of HSIs for their abundant spectral and spatial information and the excellent performance of DLN in machine learning. By modeling the temporal dependence of multiscale representative features, more discriminative information reflecting land use and land cover (LULC) changes can be obtained by suppressing less correlated information while improving the robustness of pseudo-changes caused by imaging noises. However, preserving time-dependent multiscale representative features while extracting spatial–spectral features based on conventional DLN is difficult, mainly due to the structural limitation of conventional DLN. A multipath convolutional long short-term memory (LSTM) multipath convolutional long short-term memory neural network (MP-ConvLSTM) taking advantage of LSTM and convolutional neural network (CNN) through the designed parallel architecture to learn multilevel temporal dependencies of bitemporal HSIs, therefore, was proposed for extracting multiscale temporal–spatial–spectral features by combining hidden states from different paths of ConvLSTM in the present study. In the proposed MP-ConvLSTM, the efficient channel attention (ECA) module was introduced to refine features of different paths, and Siamese CNN was adopted to reduce HSIs’ dimensionality and extract preliminary features to build up an end-to-end trainable model for CD with HSIs. The validity of the MP-ConvLSTM was evaluated using the binary and multiclass CD datasets. The CD accuracy of the proposed MP-ConvLSTM was visually and statistically evaluated by different criteria and compared with those derived from several state-of-the-art (SOTA) CD algorithms. The experiments demonstrated that our proposed model not only outperformed those SOTA CD models but also exhibited better tradeoff between complexity and accuracy in general.

A M_w=6.5 earthquake struck the Jiuzhaigou region of Sichuan Province, China, at 21:19 LT on Tuesday, 8 August 2017, and triggered a large number of landslides. For mitigating the damages of earthquake-triggered landslides to individuals and infrastructures of the earthquake-affected region, a comprehensive landslide susceptibility mapping was attempted with an integrated weighted index model by combining the frequency ratio and the analytical hierarchy process approaches under a GIS-based environment in the heavily earthquake-affected Zhangzha town of the Jiuzhaigou region. For this purpose, a total number of 842 earthquake-triggered landslides were visually interpreted and located from Sentinel-2A images acquired before and after the earthquake at first, and then the recognized landslides were randomly split into two groups to establish the earthquake-triggered landslide inventory, among which 80 % of the landslides were used for training the integrated model and the remaining 20 % for validation. Nine landslide controlling factors were considered including slope, aspect, elevation, lithology, distance from faults, distance from rivers, land use–land cover, normalized difference vegetation index and peak ground acceleration. The frequency ratio was utilized to evaluate the contribution of each landslide controlling factor to landslide occurrence, and the analytical hierarchy process was used to analyse the mutual relationship between landslide controlling factors. Finally, the landslide susceptibility map was produced by using weighted overlay analysis. Furthermore, an area under the curve approach was adopted to comprehensively evaluate the performance of the integrated weighted index model, including the degree of model fit and model predictive capability. The results demonstrated the reliability and feasibility of the integrated weighted index model in earthquake-triggered landslide susceptibility mapping at a regional scale. The generated map can help engineers and decision makers assess and mitigate hazards of the earthquake-triggered landslides to individuals and infrastructures of the earthquake-affected region.

Landslide susceptibility mapping is an effective approach for landslide risk prevention and assessments. The occurrence of slope instability is highly correlated with intrinsic variables that contribute to the occurrence of landslides, such as geology, geomorphology, climate, hydrology, etc. However, feature selection of those conditioning factors to constitute datasets with optimal predictive capability effectively and accurately is still an open question. The present study aims to examine further the integration of the selected landslide conditioning factors with Q-statistic in Geo-detector for determining stratification and selection of landslide conditioning factors in landslide risk analysis as to ultimately optimize landslide susceptibility model prediction. The location chosen for the study was Atsuma Town, which suffered from landslides following the Eastern Iburi Earthquake in 2018 in Hokkaido, Japan. A total of 13 conditioning factors were obtained from different sources belonging to six categories: geology, geomorphology, seismology, hydrology, land cover/use and human activity; these were selected to generate the datasets for landslide susceptibility mapping. The original datasets of landslide conditioning factors were analyzed with Q-statistic in Geo-detector to examine their explanatory powers regarding the occurrence of landslides. A Random Forest (RF) model was adopted for landslide susceptibility mapping. Subsequently, four subsets, including the Manually delineated landslide Points with 9 features Dataset (MPD9), the Randomly delineated landslide Points with 9 features Dataset (RPD9), the Manually delineated landslide Points with 13 features Dataset (MPD13), and the Randomly delineated landslide Points with 13 features Dataset (RPD13), were selected by an analysis of Q-statistic for training and validating the Geo-detector-RF- integrated model. Overall, using dataset MPD9, the Geo-detector-RF-integrated model yielded the highest prediction accuracy (89.90%), followed by using dataset MPD13 (89.53%), dataset RPD13 (88.63%) and dataset RPD9 (87.07%), which implied that optimized conditioning factors can effectively improve the prediction accuracy of landslide susceptibility mapping.

Shortly after the outbreak of the novel infectious disease (COVID-19) started at the end of 2019, it turned into a global pandemic, which caused the lockdown of many countries across the world. Various strict measures were adopted to reduce anthropogenic activities in order to prevent further spread and infection of the disease. In this study, we utilized continental scale remotely sensed data along with city scale in situ air quality observations for 2020 as well as data from the baseline period (2015–2019) to provide an early insight on air pollution changes in response to the COVID-19 pandemic lockdown, by combining both continental and city scales. For the continental scale analysis, data of NO₂, SO₂, and O₃ were acquired from the ozone monitoring instrument (OMI) and data of aerosol optical depth (AOD) were collected from the moderate resolution imaging spectroradiometer (MODIS). For city scale analysis, data of NO₂, CO, PM2.5, O₃, and SO₂ were derived from ground-based air quality observations. Results from satellite observations at the continental scale showed that concentrations of NO₂, SO₂, and AOD substantially dropped in 2020 during the lockdown period compared to their averages for the baseline period over all continents, with a maximum reduction of ~33% for NO₂ in East Asia, ~41% for SO₂ in East Asia, and ~37% for AOD in South Asia. In the case of O₃, the maximum overall reduction was observed as ~11% in Europe, followed by ~10% in North America, while a slight increase was found in other study regions. These findings align with ground-based air quality observations, which showed that pollutants such as NO₂, CO, PM2.5, and SO₂ during the 2020 lockdown period decreased significantly except that O₃ had varying patterns in different cities. Specifically, a maximum reduction of ~49% in NO₂ was found in London, ~43% in CO in Wuhan, ~38% in PM2.5 in Chennai, and ~48% in SO₂ in Beijing. In the case of urban O₃, a maximum reduction of ~43% was found in Wuhan, but a significant increase of ~47% was observed in Chennai. It is obvious that restricted human activities during the lockdown have reduced the anthropogenic emissions and subsequently improved air quality, especially across the metropolitan cities.