Search

Convolutional neural network (CNN)-based and Transformer-based methods for hyperspectral image (HSI) classification have rapidly advanced due to their unique characterization capabilities. However, the fixed kernel sizes in convolutional layers limit the comprehensive utilization of multi-scale features in HSI land cover analysis, while the Transformer's multi-head self-attention (MHSA) mechanism faces challenges in effectively encoding feature information across various dimensions. To tackle this issue, this article introduces an HSI classification method, based on multi-scale convolutional features and multi-attention mechanisms (i.e., MSCF-MAM). Firstly, the model employs a multi-scale convolutional module to capture features across different scales in HSIs. Secondly, to enhance the integration of local and global channel features and establish long-range dependencies, a feature enhancement module based on pyramid squeeze attention (PSA) is employed. Lastly, the model leverages a classical Transformer Encoder (TE) and linear layers to encode and classify the transformed spatial–spectral features. The proposed method is evaluated on three publicly available datasets—Salina Valley (SV), WHU-Hi-HanChuan (HC), and WHU-Hi-HongHu (HH). Extensive experimental results have demonstrated that the MSCF-MAM method outperforms several representative methods in terms of classification performance. [ABSTRACT FROM AUTHOR]

Hyperspectral image classification (HSIC) has garnered increasing attention among researchers. While classical networks like convolution neural networks (CNNs) have achieved satisfactory results with the advent of deep learning, they are confined to processing local information. Vision transformers, despite being effective at establishing long-distance dependencies, face challenges in extracting high-representation features for high-dimensional images. In this paper, we present the multiscale efficient attention with enhanced feature transformer (MEA-EFFormer), which is designed for the efficient extraction of spectral–spatial features, leading to effective classification. MEA-EFFormer employs a multiscale efficient attention feature extraction module to initially extract 3D convolution features and applies effective channel attention to refine spectral information. Following this, 2D convolution features are extracted and integrated with local binary pattern (LBP) spatial information to augment their representation. Then, the processed features are fed into a spectral–spatial enhancement attention (SSEA) module that facilitates interactive enhancement of spectral–spatial information across the three dimensions. Finally, these features undergo classification through a transformer encoder. We evaluate MEA-EFFormer against several state-of-the-art methods on three datasets and demonstrate its outstanding HSIC performance. [ABSTRACT FROM AUTHOR]

Despite notable advancements achieved on Hyperspectral (HS) pansharpening tasks through deep learning techniques, previous methods are inherently constrained by convolution or self-attention intrinsic defects, leading to limited performance. In this paper, we proposed an Attention-Interactive Dual-Branch Convolutional Neural Network (AIDB-Net) for HS pansharpening. Our model purely consists of convolutional layers and simultaneously inherits the strengths of both convolution and self-attention, especially the modeling of short- and long-range dependencies. Specially, we first extract, tokenize, and align the hyperspectral image (HSI) and panchromatic image (PAN) by Overlapping Patch Embedding Blocks. Then, we specialize a novel Spectral-Spatial Interactive Attention which is able to globally interact and fuse the cross-modality features. The resultant token-global similarity scores can guide the refinement and renewal of the textural details and spectral characteristics within HSI features. By deeply combined these two paradigms, our AIDB-Net significantly improve the pansharpening performance. Moreover, with the acceleration by the convolution inductive bias, our interactive attention can be trained without large scale dataset and achieves competitive time cost with its counterparts. Compared with the state-of-the-art methods, our AIDB-Net makes 5.2%, 3.1%, and 2.2% improvement on PSNR metric on three public datasets, respectively. Comprehensive experiments quantitatively and qualitatively demonstrate the effectiveness and superiority of our AIDB-Net. [ABSTRACT FROM AUTHOR]

There have been considerable efforts in generating tree crown maps from satellite images. However, tree localization in urban environments using satellite imagery remains a challenging task. One of the difficulties in complex urban tree detection tasks lies in the segmentation of dense tree crowns. Currently, methods based on semantic segmentation algorithms have made significant progress. We propose to split the tree localization problem into two parts, dense clusters and single trees, and combine the target detection method with a procedural generation method based on planting rules for the complex urban tree detection task, which improves the accuracy of single tree detection. Specifically, we propose a two-stage urban tree localization pipeline that leverages deep learning and planting strategy algorithms along with region discrimination methods. This approach ensures the precise localization of individual trees while also facilitating distribution inference within dense tree canopies. Additionally, our method estimates the radius and height of trees, which provides significant advantages for three-dimensional reconstruction tasks from remote sensing images. We compare our results with other existing methods, achieving an 82.3% accuracy in individual tree localization. This method can be seamlessly integrated with the three-dimensional reconstruction of urban trees. We visualized the three-dimensional reconstruction of urban trees generated by this method, which demonstrates the diversity of tree heights and provides a more realistic solution for tree distribution generation. [ABSTRACT FROM AUTHOR]

Strong earthquakes can not only trigger many landslides in a short period of time but can also change the stability of slopes in the earthquake area, causing them to be active for a long time after the earthquake. Research on the variation of slow-motion slopes before and after earthquakes can help us to better understand the mechanism of earthquake-affected landslides, which is also crucial for assessing the long-term landslide risk in seismically active areas. Here, L-band ALOS-2 PALSAR-2 images are utilized with the SBAS-InSAR algorithm to monitor and assess the location and activity changes of slow-moving landslides in the Iburi region (Hokkaido, Japan) before and after an earthquake occurred on 6 September 2018. Unlike previous studies, which focused on single typical landslides, we tracked all the landslides within a 33 × 55 km region close to the epicenter. According to the results, the majority of the co-seismic landslides that quickly failed during the earthquake are now stable, and a few of them are still moving. In contrast, due to near-field seismic shaking, certain slopes that did not show substantial surface changes during the earthquake period continued to move and eventually developed into slow-moving landslides. In addition, it can be seen from the spatial distribution of slow-moving landslides after the earthquake that this distribution is not only dependent on strong earthquake seismic vibration or the hanging-wall effect. Far-field weak vibrations can also accelerate landslides. Additionally, we discovered that the earthquake made the unstable slopes move more quickly but also tended to stabilize the slopes that were already in motion before the earthquake. The various response modes of slow-moving landslides to seismic events are related not only to the intensity of seismic vibration but also to the geological conditions of the region and to the size of the landslide itself. These findings are extremely valuable for studying the mechanism of earthquake-affected landslides. [ABSTRACT FROM AUTHOR]

As a by-product of Interferometric Synthetic Aperture Radar (SAR, InSAR) technique, interferometric coherence is a measure of the decorrelation noise for InSAR observation, where the lower the coherence value, the more serious the decorrelation noise. In the densely vegetated area, the coherence value could be too low to obtain any valuable signals, leading to the degradation of InSAR performance and the possible waste of expensive SAR data. Normalized Difference Vegetation Index (NDVI) value is a measure of the vegetation coverage and can be estimated from the freely available optical satellite images. In this paper, a multi-stage model is established to quantitatively estimate the decorrelation noise for vegetable areas based on Landsat-derived NDVI prior to the acquisition of SAR data. The modeling process is being investigated with the L-band ALOS-1/PALSAR-1 data and the Landsat-5 optical data acquired in the Meitanba area of Hunan Province, China. Furthermore, the reliability of the established model is verified in the Longhui area, which is situated near the Meitanba area. The results demonstrate that the established model can quantitatively estimate InSAR decorrelation associated with the vegetation coverage. [ABSTRACT FROM AUTHOR]

The excessive extraction and recharge of groundwater lead to long-time seasonal land subsidence in Los Angeles, USA, and especially in the Santa Ana basin. The rate of land subsidence in the Santa Ana basin has been rising, which could pose a danger to infrastructure and human lives. However, the most recent research on land surface deformation in the area was conducted using the traditional parameter estimation method, resulting in little understanding of the regional spatiotemporal characteristics. The parametric method consists of a least square linear inversion, using the pre-defined mathematical geometric or geophysical theoretical models to describe groundwater deformation, and it requires precise external environmental variables and accurate geophysical parameters, which are more difficult to implement. In this study, multitemporal InSAR-derived deformation time series are analyzed by using 69 descending C-band Sentinel-1A SAR scenes acquired from 2015 to 2018. A method based on independent component analysis (ICA) is applied to characterize the spatial pattern and temporal evolution of land subsidence in the Santa Ana basin. The results reveal two different spatial and temporal deformation patterns in the basin. First, a widespread seasonal deformation is identified by the first component, related to annual seasonal groundwater level changes, and the overall deformation shows a concentrated spatial pattern. The second component captures a long-term signal with a large-scale spatial pattern. For quantitative assessment, the obtained deformation time series are compared with the GNSS data, validating an accuracy of millimeters. We further calculate the cross-correlation coefficient and the elastic skeletal storage coefficient from the ICA-derived seasonal deformation and groundwater level, which reveals that the deformation responds quickly (i.e., a lag of 8 days) to the change in groundwater and the Santa Ana aquifer retains almost the same elasticity for at least 15 years. Quantifying the spatiotemporal characteristics of the deformation in the Santa Ana basin can provide a reference for the monitoring and managing of groundwater. [ABSTRACT FROM AUTHOR]

Ground deformation is the direct manifestation of the earth-rock dam's hazard potential. Therefore, it is essential to monitor deformation for dam warning and security evaluation. The Liuduzhai Dam, a clay-core dam of a large reservoir in China, was reinforced with plastic concrete cut-off walls between 13 January 2009 and 29 May 2010, as it was subject to leakage and deformation. However, the deformation development and the mechanism of the dam are still unclear. In this study, the deformation fields before and after the reinforcement of the Liuduzhai Dam were yielded by using the Interferometric Synthetic Aperture Radar (InSAR) technique. Furthermore, a numerical simulation method was employed to obtain the dynamic seepage field of the dam during the InSAR observation period. The results indicated that the average deformation velocity and maximum deformation velocity are −11.7 mm/yr and −22.5 mm/yr, respectively, and the cumulative displacement exceeds 100 mm, which shows typical continuous growth characteristics in a time series. In contrast, the dam deformation tended to be stable after reinforcement, with the average deformation velocity and maximum deformation velocity being −0.4 mm/yr and −1.2 mm/yr, respectively, behaving as cyclical deformation time series. According to the results of InSAR and seepage analysis, it is shown that: (1) dynamic seepage was the main mechanism controlling dam deformation prior to reinforcement; (2) the concentrated load caused by construction and the rapid dissipation of pore water pressure caused by the sudden drop of the infiltration line were the reasons for the acceleration of deformation during and after construction; and (3) the plastic concrete cut-off walls effectively reduced the dynamic seepage field, while the water level fluctuations were the main driving factor of elastic deformation of the dam after reinforcement. This study provides a novel approach to investigating the deformation mechanism of earth-rock dams. Furthermore, it has been confirmed that InSAR can identify the seepage deformation of dams by detecting surface movements. It is recommended that InSAR deformation monitoring should be incorporated into future dam safety programs to provide detailed deformation signals. By analyzing the temporal and spatial characteristics of the deformation signal, we can identify areas where dam performance has degraded. This crucial information aids in conducting a comprehensive dam safety assessment. [ABSTRACT FROM AUTHOR]

Soil moisture (SM) is critical in monitoring the time-lagged impacts of agrometeorological drought. In Australia and several south-west Pacific Small Island Developing States (SIDS), there are a limited number of in situ SM stations that can adequately assess soil-water availability in a near-real-time context. Satellite SM datasets provide a viable alternative for SM monitoring and agrometeorological drought provision in these regions. In this study, we investigated the performance of Soil Moisture Active Passive (SMAP), Soil Moisture and Ocean Salinity (SMOS), Soil Moisture Operational Products System (SMOPS), SM from the Advanced Microwave Scanning Radiometer 2 (AMSR-2) and SM from the Advanced Scatterometer (ASCAT) over Australia and south-west Pacific SIDS. Products were first evaluated in Australia, given the presence of several in-situ SM monitoring stations and a state-of-the-art hydrological model—the Australian Water Resources Assessment Landscape modelling system (AWRA-L). We further investigated the accuracy of SM satellite datasets in Australia and the south-west Pacific through Triple Collocation analysis with two other SM reference datasets—ERA5 reanalysis SM data and model data from the Global Land Data Assimilation System (GLDAS) dataset. All datasets have differing observation periods ranging from 1911-now, with a common period of observations between 2015–2021. Results demonstrated that ASCAT and SMOS were consistently superior in their performance. Analysis in the six south-west Pacific SIDS indicated reduced performance for all products, with ASCAT and SMOS still performing better than others for most SIDS with median R values ranging between 0.3–0.9. We conducted a case study of the 2015 El Niño and Positive Indian Ocean Dipole-induced drought in Papua New Guinea. It was shown that ASCAT is a valuable dataset indicative of agrometeorological drought for the nation, highlighting the value of using satellite SM products to provide early warning of drought in data-sparse regions in the south-west Pacific. [ABSTRACT FROM AUTHOR]

The 8 January 2022 Mw6.7 Menyuan earthquake was generated in the transition zone between the western Lenglongling fault and the eastern Tuolaishan fault, both being part of the Qilian–Haiyuan fault system with an important role in the adjustment of the regional tectonic regime. In this study, four pairs of SAR (synthetic aperture radar) data from Sentinel-1 and ALOS-2 (Advanced Land Observation Satellite-2) satellites were used to derive the surface displacement observations along the satellite line-of-sight (LOS) and azimuth directions using the differential interferometric SAR (InSAR, DInSAR), pixel offset-tracking (POT), multiple aperture InSAR (MAI), and burst overlap InSAR (BOI) methods. An SM-VCE method (i.e., a method for measuring three-dimensional (3D) surface displacements with InSAR based on a strain model and variance component estimation) was employed to combine these derived SAR displacement observations to calculate the 3D co-seismic displacements. Results indicate that the 2022 Menyuan earthquake was dominated by left-lateral slip, and the maximum horizontal and vertical displacements were 1.9 m and 0.6 m, respectively. The relative horizontal surface displacement across the fault was as large as 2–3 m, and the fault-parallel displacement magnitude was larger on the southern side of the fault compared with the northern side. Furthermore, three co-seismic strain invariants were also investigated, revealing that the near-fault area suffered severe deformation, and two obviously expanding and compressed zones were identified. We provide displacements/strains derived in this study in the prevailing geotiff format, which will be useful for the broad community studying this earthquake; in addition, the SM-VCE code used in this study is open to the public so that readers can better understand the method. [ABSTRACT FROM AUTHOR]

With the rapid increase in the costs of rural labour and the adjustment of planting structures, the phenomenon of farmland abandonment has appeared in China. It is of great significance to promptly and accurately grasp the information on dynamic temporal and spatial changes in abandoned farmland to ensure national food security and the sustainable use of cultivated land. Luquan District in Hebei, China was selected as the research area based on multispectral images from Sentinel-2A, Landsat-7, and Landsat-8 combined with methods of random forest (RF) classification and vegetation index change detection. Rules for the identification of abandoned farmland were also developed, and remote sensing monitoring of the abandonment status of the cultivated land was also carried out in the study area. We also obtained the spatial distribution of abandoned and reclaimed farmland and analysed the frequency of farmland abandonment. The results show that the overall accuracy of the land-use time-series map ranged from 90.20% to 96.92% for the study period of 2010–2020. The average rate of farmland abandonment in the study area was 10.62%, with the lowest rate (5.83%) in 2020 and the highest (14.09%) in 2012. From 2011 to 2020, the maximum farmland abandonment area was 3906.02 hm2, and the minimum area was 1618.74 hm2. The farmland abandonment area showed a trend of first increasing and then decreasing. From 2012 to 2020, the maximum area of reclaimed farmland was 291.49 hm2, and the highest rate of reclamation was 14.26%. The overall reclamation rate was low. The abandonment frequency of most of the abandoned farmland was 1–3 years, covering an area of 8193.73 hm2, which comprised 79% of the total area of abandoned farmland. The frequency of abandonment was inversely proportional to the area of abandoned farmland. Farmland abandonment mainly occurred in hilly areas. We expect that our results can provide case studies for long time series in farmland abandonment research and can provide a reference for studying the driving factors, risk assessment, and policymaking with respect to abandoned farmland. [ABSTRACT FROM AUTHOR]

Drought has significant impacts on the agricultural productivity and well-being of Pacific Island communities. In this study, a user-centred integrated early warning system (I-EWS) for drought was investigated for Papua New Guinea (PNG). The I-EWS combines satellite products (Standardised Precipitation Index and Vegetation Health Index) with seasonal probabilistic forecasting outputs (chance of exceeding median rainfall). Internationally accepted drought thresholds for each of these inputs are conditionally combined to trigger three drought early warning stages—"DROUGHT WATCH", "DROUGHT ALERT" and "DROUGHT EMERGENCY". The developed I-EWS for drought was used to examine the evolution of a strong El Niño-induced drought event in 2015 as well as a weaker La Niña-induced dry period in 2020. Examining the evolution of drought early warnings at a provincial level, it was found that tailored warning lead times of 3–5 months could have been possible for several impacted PNG provinces. These lead times would enable increasingly proactive drought responses with the potential for prioritised allocation of funds at a provincial level. The methodology utilised within this study uses inputs that are openly and freely available globally which indicates promising potential for adaptation of the developed user-centred I-EWS in other Pacific Island Countries that are vulnerable to drought. [ABSTRACT FROM AUTHOR]

Lodging is one of the main problems in maize production. Assessing the self-recovery ability of maize plants after lodging at different growth stages is of great significance for yield loss assessment and agricultural insurance claims. The objective of this study was to quantitatively analyse the effects of different growth stages and lodging severity on the self-recovery ability of maize plants using UAV-LiDAR data. The multi-temporal point cloud data obtained by the RIEGL VUX-1 laser scanner were used to construct the canopy height model of the lodging maize. Then the estimated canopy heights of the maize at different growth stages and lodging severity were obtained. The measured values were used to verify the accuracy of the canopy height estimation and to invert the corresponding lodging angle. After verifying the accuracy of the canopy height, the accuracy parameter of the tasselling stage was R2 = 0.9824, root mean square error (RMSE) = 0.0613 m, and nRMSE = 3.745%. That of the filling stage was R2 = 0.9470, RMSE = 0.1294 m, and nRMSE = 9.889%, which showed that the UAV-LiDAR could accurately estimate the height of the maize canopy. By comparing the yield, canopy height, and lodging angle of maize, it was found that the self-recovery ability of maize at the tasselling stage was stronger than that at the filling stage, but the yield reduction rate was 14.16~26.37% higher than that at the filling stage. The more serious the damage of the lodging is to the roots and support structure of the maize plant, the weaker is the self-recovery ability. Therefore, the self-recovery ability of the stem tilt was the strongest, while that of root lodging and root stem folding was the weakest. The results showed that the UAV-LiDAR could effectively assess the self-recovery ability of maize after lodging. [ABSTRACT FROM AUTHOR]

The Antarctic is one of the most sensitive areas to climate change, and ice velocity is a fundamental parameter for quantitatively assessing the glacier mass balance. Interferometric synthetic aperture radar (InSAR), a powerful tool for monitoring surface deformation with the advantages of having high precision and wide coverage, has been widely used in determining ice velocity in the Antarctic. However, the mapping of complete three-dimensional (3D) ice velocities is greatly limited by the imaging geometries and digital elevation model (DEM)-induced errors. In this study, we propose the integration of multibaseline and multiaperture InSAR measurements from the ENVISAT ASAR datasets to derive complete 3D ice velocities in the Grove Mountains area of the Antarctic. The results show that the estimated complete 3D ice velocities are in good agreement with MEaSUREs and GPS observations. Compared with the conventional 2D and quasi-3D ice velocities, the complete 3D ice velocities can effectively eliminate the effects of DEM errors and elevation changes and are also capable of retrieving the thickness change of the ice, which provides important information on the origin of mass transition. [ABSTRACT FROM AUTHOR]