Search

Extracting water-bodies accurately is a great challenge from very high resolution (VHR) remote sensing imagery. The boundaries of a water body are commonly hard to identify due to the complex spectral mixtures caused by aquatic vegetation, distinct lake/river colors, silts near the bank, shadows from the surrounding tall plants, and so on. The diversity and semantic information of features need to be increased for a better extraction of water-bodies from VHR remote sensing images. In this paper, we address these problems by designing a novel multi-feature extraction and combination module. This module consists of three feature extraction sub-modules based on spatial and channel correlations in feature maps at each scale, which extract the complete target information from the local space, larger space, and between-channel relationship to achieve a rich feature representation. Simultaneously, to better predict the fine contours of water-bodies, we adopt a multi-scale prediction fusion module. Besides, to solve the semantic inconsistency of feature fusion between the encoding stage and the decoding stage, we apply an encoder-decoder semantic feature fusion module to promote fusion effects. We carry out extensive experiments in VHR aerial and satellite imagery respectively. The result shows that our method achieves state-of-the-art segmentation performance, surpassing the classic and recent methods. Moreover, our proposed method is robust in challenging water-body extraction scenarios.

Semantic segmentation of LiDAR point clouds has implications in self-driving, robots, and augmented reality, among others. In this paper, we propose a Multi-Scale Attentive Aggregation Network (MSAAN) to achieve the global consistency of point cloud feature representation and super segmentation performance. First, upon a baseline encoder-decoder architecture for point cloud segmentation, namely, RandLA-Net, an attentive skip connection was proposed to replace the commonly used concatenation to balance the encoder and decoder features of the same scales. Second, a channel attentive enhancement module was introduced to the local attention enhancement module to boost the local feature discriminability and aggregate the local channel structure information. Third, we developed a multi-scale feature aggregation method to capture the global structure of a point cloud from both the encoder and the decoder. The experimental results reported that our MSAAN significantly outperformed state-of-the-art methods, i.e., at least 15.3% mIoU improvement for scene-2 of CSPC dataset, 5.2% for scene-5 of CSPC dataset, and 6.6% for Toronto3D dataset.

The adjacency effect and non-uniform responses complicate the precise delimitation of the surface support of remote sensing data and their derived products. Thus, modeling spatial response characteristics (SRCs) prior to using remote sensing information has become important. A point spread function (PSF) is typically used to describe the SRCs of the observation cells from remote sensors and is always estimated in a laboratory before the sensor is launched. However, research on the SRCs of high-order remote sensing products derived from the observations remains insufficient, which is an obstacle to converting between multi-scale remote sensing products and validating coarse-resolution products. This study proposed a method that combines simulation and validation to establish SRC models of coarse-resolution albedo products. Two series of commonly used 500-m/1-km resolution albedo products, which are derived from Moderate Resolution Imaging Spectroradiometer (MODIS) reflectance data, were investigated using 30-m albedo products that provide the required sub-pixel information. The analysis proves that the size of the surface support of each albedo pixel is larger than the nominal resolution of the pixel and that the response weight is non-uniformly distributed, with an elliptical Gaussian shape. The proposed methodology is generic and applicable for analyzing the SRCs of other advanced remote sensing products.

We present a novel convolutional neural network (CNN)-based change detection framework for locating changed building instances as well as changed building pixels from very high resolution (VHR) aerial images. The distinctive advantage of the framework is the self-training ability, which is highly important in deep-learning-based change detection in practice, as high-quality samples of changes are always lacking for training a successful deep learning model. The framework consists two parts: a building extraction network to produce a binary building map and a building change detection network to produce a building change map. The building extraction network is implemented with two widely used structures: a Mask R-CNN for object-based instance segmentation, and a multi-scale full convolutional network for pixel-based semantic segmentation. The building change detection network takes bi-temporal building maps produced from the building extraction network as input and outputs a building change map at the object and pixel levels. By simulating arbitrary building changes and various building parallaxes in the binary building map, the building change detection network is well trained without real-life samples. This greatly lowers the requirements of labeled changed buildings, and guarantees the algorithm’s robustness to registration errors caused by parallaxes. To evaluate the proposed method, we chose a wide range of urban areas from an open-source dataset as training and testing areas, and both pixel-based and object-based model evaluation measures were used. Experiments demonstrated our approach was vastly superior: without using any real change samples, it reached 63% average precision (AP) at the object (building instance) level. In contrast, with adequate training samples, other methods—including the most recent CNN-based and generative adversarial network (GAN)-based ones—have only reached 25% AP in their best cases.

In recent years, sequential tests for detecting structural changes in time series have been adapted for deforestation monitoring using satellite data. The input time series of such sequential tests is typically a vegetation index (e.g., NDVI), which uses two or three bands and ignores all other bands. Being limited to a vegetation index will not benefit from the richer spectral information provided by newly launched satellites and will bring two bottle-necks for deforestation monitoring. Firstly, it is hard to select a suitable vegetation index a priori. Secondly, a single vegetation index is typically affected by seasonal signals, noise and other natural dynamics, which decrease its power for deforestation detection. A novel multispectral time series change monitoring method that combines dimension reduction methods with a sequential hypothesis test is proposed to address these limitations. For each location, the proposed method automatically chooses a “suitable” index for deforestation monitoring. To demonstrate our approach, we implemented it in two study areas: a dry tropical forest in Bolivia (time series length: 444) with strong seasonality and a moist tropical forest in Brazil (time series length: 225) with almost no seasonality. Our method significantly improves accuracy in the presence of strong seasonality, in particular the temporal lag between disturbance and its detection.

Dust aerosols persistently affect nearly all landscapes worldwide, and the saltation activity caused by dusty weather (e.g., dust days) releases considerable amounts of aerosol into the atmosphere. Nevertheless, dust-induced saltation activity may also occur on non-dust days. To date, few studies have investigated the saltation activity characteristics on non-dust days. Moreover, the contribution of non-dust days to the total saltation activity remains ambiguous. This study comprehensively investigates the characteristics of saltation activity on non-dust days. Specifically, we analyze the influence of the saltation activity of non-dust days on dust aerosols by utilizing saltation, atmospheric, soil, dust aerosol (i.e., PM10 and aerosol optical depth), and weather record data obtained from the Taklimakan Desert, China, between 2008 and 2010. Our results show that lower temperature, vapor pressure, and soil moisture on non-dust days reduces the saltation threshold velocity (5.9 m/s) more compared to on dust days (6.5 m/s). Furthermore, regarding wind speed, relatively strong monthly saltation activity occurs from March to August, and daily saltation activity occurs from 9:00 to 16:00. Although non-dust days only contribute 18.5% and 7.7% to saltation time and saltation count, respectively, both significantly influence the dust aerosols. Therefore, the effect of saltation activity on non-dust days cannot be undervalued, particularly while performing dust aerosol studies. [ABSTRACT FROM AUTHOR]

A special topography and ultra-high atmospheric boundary layer conditions in the Tarim Basin (TB) lead to the unique spatial–temporal distribution characteristics of dust aerosols. A typical dust storm with persistent floating dust over the TB from 27 April to 1 May 2015 was used to investigate the characteristics of the dust radiative effect using the Weather Research and Forecasting Model with Chemistry (WRF-Chem). Based on reasonable evaluations involving in situ sounding observations, as well as remotely sensed MODIS observations of meteorology, dust aerosols, and the ultra-high atmospheric boundary layer, the simulation characterized the complete characteristics of the dust direct radiative effect (DDRE) during the dust storm outbreak stage and persistent floating dust stage over the TB. During the daytime, the shortwave (SW) radiative effect heated the atmosphere and cooled the land surface (SUR), whereas the longwave (LW) radiative effect had the opposite effect on the TB. Regarding low-level dust, the LW radiative effect was greater than the SW DDRE in the atmosphere, while for high-level dust the situation was reversed. During the nighttime, the LW DDRE at the top of the atmosphere (TOA), at the SUR, and in the atmosphere was less than that during the daytime, when the DDRE at the SUR was the most significant. In contrast to the daytime, the near-surface dust aerosols exerted an LW warming effect in the atmosphere during the nighttime; however, the dust LW radiative effect had a cooling effect from above a 100 m altitude until the top of the dust layer. In contrast, the DDRE heating rate peaked at the top of the dust layer within the TB. The event-averaged net DDRE was 0.53, −5.90, and 6.43 W m−2 at the TOA, at the SUR, and in the atmosphere over the TB, respectively. The dust SW radiative effect was stronger than the dust L4W radiative effect over the TB at the SUR and in the atmosphere. Moreover, the DDRE at the TOA was weaker than that at the SUR. Overall, the study revealed noteworthy radiative effect features of dust aerosols during typical dust storms with persistent floating dust over the TB. [ABSTRACT FROM AUTHOR]

Based on observations from the Unmanned Aerial Vehicle (UAV) together with an environmental particulate matter analyzer (Grimm-180) and Global Positioning System (GPS) sounding balloons, the vertical structure of dust with different particle sizes was explored over the Taklimakan Desert (TD) during an intensive observation from 1 July 2021 to 31 July 2021. The power functions were fitted between the particle counts and particle sizes, indicating negative correlations with an R2 higher than 0.99 under different dust pollution conditions in Tazhong (TZ). The dust concentrations show a sharp vertical increase over the TD during dust pollution; however, more particles with larger sizes are entrained into the air in TZ compared with Minfeng (MF). The total solar radiation during dust pollution days is significantly weakened, accompanied by major modifications in the temperature stratification, which were characterized by low-level cooling (with −2.71 K mean intensity) and high-level heating (with +0.70 K mean intensity). On clear days, the average convective boundary layer (CBL) heights at the TZ and MF are approximately 3.94 and 2.84 km, respectively, and the average stable boundary layer (SBL) height at the TZ and MF are approximately 0.19 and 0.14 km, respectively. With the increasing dust pollution level, the CBL height decreases rapidly while the SBL height shows the opposite trend. The unique ultra-high atmospheric boundary layer structure in daytime provides beneficial conditions for the suspension and vertical transportation of dust over TD. Moreover, a negative correlation between the CBL height and near-surface PM10/PM2.5/PM1.0 concentration in TD is revealed by power function fittings. [ABSTRACT FROM AUTHOR]

In the present study, the spatiotemporal and vertical distributions of ice cloud properties and their association with meteorological variables are analyzed for the period 2007–2016 using the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) and Modern Era Retrospective-Analysis for Research (MERRA-2) reanalysis observations. The distribution of ice cloud fraction (ICF) with its peak does not overlap with that of the ice water content (IWC) peak during daytime and nighttime due to the sampling bias. Moreover, the vertical distributions of mean IWC exhibited a vaguely "sharp thorn" at an altitude of ~4 km in all seasons at the location of about ±40°, which can be caused by the artifacts. Furthermore, it is noted that different ice cloud optical depth (ICOD) presents significant changes observed in their diurnal variations in the heights of peaks. The maximum diurnal difference of ice cloud properties occurs in the tropical regions of the North Hemisphere (NH) during summer. We also investigated the relation between ICOD and the meteorological variables and found that the ICOD values are dependent on the meteorological parameters. [ABSTRACT FROM AUTHOR]