Multi-temporal Observations Research Articles

GeoAI Dataset for Training Deep Learning-Based Optical Satellite Image Matching Model

Satellite imagery is being used to monitor the Earth, as it allows for the continuous provision of multi-temporal observations with consistent quality. To analyze time series remote sensing data with high accuracy, the process of image registration must be conducted beforehand. Image registration techniques are mainly divided into region-based registration and feature-based registration, and both techniques extract the same points based on the similarity of spectral characteristics and object shapes between master and slave images. In addition, recently, deep learning-based siamese neural network and convolutional neural network models have been utilized to match images. This has high performance compared to previous non-deep learning algorithms, but a very large amount of data is required to train a deep learning-based image registration model. In this study, we aim to generate a dataset for training a deep learning-based optical image registration model. To build the data, we acquired Satellite Side-Looking (S2Looking) data, an open dataset, and performed preprocessing and data augmentation on the data to create input data. After that, we added offsets to the X and Y directions between the master and slave images to create label data. The preprocessed input data and labeled data were used to build a dataset suitable for image registration. The data is expected to be useful for training deep learning-based satellite image registration models.

Continuous remote sensing ecological index (CRSEI): A novel approach for multitemporal monitoring of eco-environmental changes on large scale

With urbanization and anthropogenic pressures, the eco-environmental quality of East and Southeast Asian coastal regions has undergone drastic changes in the 21st century. To assess these environmental variations and realize regional sustainable development goals, a quantitative monitoring of the eco-environment is necessary, but challenging. The remote sensing ecological index (RSEI) was recently proposed and applied to quantitatively monitor the eco-environmental quality, but bias during multitemporal observations is a major limitation. Thus, a novel method, Continuous Remote Sensing Ecological Index (CRSEI), was developed in this study to enable constant and objective monitoring of large-scale and long-term environmental changes in the coastal regions of East and Southeast Asia. To reduce the multitemporal bias of pixels caused by imaging conditions and sensor status, a change detection algorithm requiring limited auxiliary data was developed. High-resolution images show that the CRSEI enables adequate multitemporal monitoring of eco-environmental changes with fewer false fluctuations on the local scale. The proposed CRSEI should deepen the current understanding of environmental changes and provide a solid scientific reference for environmental protection on multiple scales. The results of this research indicate that the eco-environmental quality of the east and southeast Asian coast remained generally stable during the studied period (2001–2019), with the mean CRSEI changed slightly from 5918 to 5856. However, on a regional level, each area revealed different conditions. First, the Korean Peninsula had relatively stable and well-maintained eco-environmental quality with a CRSEI of around 5500 during 2001–2019. Second, The North China coast had lower eco-environmental quality compared to other regions, but it showed the most improvement, with the CRSEI increasing from 2719 to 2807. Third, the eco-environmental quality of South China and Myanmar's coastal region remained relatively stable, with a CRSEI of around 5730 and 6720, respectively. Fourth, the continental coast of some Southeast Asian countries, particularly Vietnam and Cambodia, experienced a significant decline in eco-environmental quality, as indicated by a dramatic decrease in the CRSEI from 6751 to 6527. Notably, the areas with declining eco-environmental quality were concentrated in the western border regions of Cambodia, where unrestricted cultivation has led to the destruction of tropical rainforests.

Multi-LEO Satellite Stereo Winds

The stereo-winds method follows trackable atmospheric cloud features from multiple viewing perspectives over multiple times, generally involving multiple satellite platforms. Multi-temporal observations provide information about the wind velocity and the observed parallax between viewing perspectives provides information about the height. The stereo-winds method requires no prior assumptions about the thermal profile of the atmosphere to assign a wind height, since the height of the tracked feature is directly determined from the viewing geometry. The method is well developed for pairs of Geostationary (GEO) satellites and a GEO paired with a Low Earth Orbiting (LEO) satellite. However, neither GEO-GEO nor GEO-LEO configurations provide coverage of the poles. In this paper, we develop the stereo-winds method for multi-LEO configurations, to extend coverage from pole to pole. The most promising multi-LEO constellation studied consists of Terra/MODIS and Sentinel-3/SLSTR. Stereo-wind products are validated using clear-sky terrain measurements, spaceborne LiDAR, and reanalysis winds for winter and summer over both poles. Applications of multi-LEO polar stereo winds range from polar atmospheric circulation to nighttime cloud identification. Low cloud detection during polar nighttime is extremely challenging for satellite remote sensing. The stereo-winds method can improve polar cloud observations in otherwise challenging conditions.

Monitoring the population change and urban growth of four major Pakistan cities through spatial analysis of open source data

ABSTRACT Cities are complex and dynamic entities in close proximity of people, implying multi temporal observations to analyse and understand the urban context. At present, open-source data and geospatial intelligence are becoming the important means of exploring, monitoring and predicting urban status of area growth and population increase. In last few decades, unemployment and absence of infrastructures in the rural areas promoted the unplanned and haphazard urbanization across the urban centres in Pakistan. This study focuses on exploring the potential of open-source/freely available datasets for city mapping and monitoring spatially. The study gives a spatial perspective of rapidly growing cities of Pakistan using Google Earth Engine to classify Landsat images over last four decades, and discovers sprawl patterns across cities. The study works out that the built-up area is significantly increasing with population growth over four decades and there is a strong positive correlation between population growth and built-up expansion. Using Open-Source Data (Landsat images and LandScan data), this study has offered a technical solution of Google Earth Engine-supported analysis of statistics and machine learning to spatially monitoring the population change and urban growth of four major Pakistan cities. It is undoubted that our working results will provide the timely and cost-effective information to policymakers, Govt Officials and citizens for more sustainable urbanization.

Contrasting reef fish diversity and abundance in the Derawan Islands: A multi-temporal observations

Derawan Island, East Kalimantan Province, is located in the coral triangle, the center of mega biodiversity of the world's marine life, including reef fish. In this study, reef fish data were collected from 1994, 2003, and 2019 at six islands (Panjang, Derawan, Samama, Sangalaki, Kakaban, and Maratua) in Derawan Islands using the Underwater Visual Census (UVC) technique. The analysis results focused on herbivores, carnivores, and corallivores reef fish show that the total species diversity and abundance in 1994, 2003, and 2019 amounted to 116 species (5,268 fish/ha), 93 species (2,664 fish/ha), and 104 species (5,332 fish/ha), respectively. Those reef fishes decreased from 1994 to 2003 due to overexploitation but increased again in 2019, especially Acanthurid, Scarid, and Lutjanid fish taxa, possibly due to local government conservation regulations. Biodiversity and abundance were found higher in the Kakaban and Maratua Islands for herbivorous fish (>25 species, >3,500 fish/ha) followed by carnivores and corallivores (±20 species, ±2,000 fish/ha) than in other islands. The diversity of corallivores, herbivores, and carnivores was moderately correlated with live coral cover with R2 of 0.78, 0.77, and 0.61, respectively, but low for fish abundance. Long-term reef fish monitoring is necessary to know coral reef health in this area.

Dual-Task Interactive Learning for Unsupervised Spatio-Temporal–Spectral Fusion of Remote Sensing Images

Spatio-temporal-spectral fusion aims to produce high spatio-temporal-spectral resolution images by integrating the complementary spatial, temporal, and spectral advantages of multi-source remote sensing images. However, on one hand, existing spatio-temporal-spectral fusion methods are insufficient to exploit the inherent complex nonlinear spatial, temporal, and spectral relationship among multisource and multitemporal observations. On the other hand, since the unavailability of real high spatio-temporal-spectral resolution images, it is difficult to adopt deep learning methods with supervised training. In this paper, we propose an effective Unsupervised Spatio-Temporal-Spectral Fusion Model (USTSFM) with dual-task interactive learning to alleviate these problems. The proposed USTSFM has two branches: the Spatio-Temporal-Spectral Mapping (STSM) branch is to describe the temporal relationship, and the Spectral Super Resolution (SSR) branch is to model the spectral relationship. Moreover, the spatial-spectral interaction compensation block is designed to make the two branches compensate and benefited from each other. This intrinsically related and mutually facilitated strategy allows the USTSFM to sufficiently exploit the inherent spatial, temporal, and spectral relationship. In addition, a shared reconstruction module is meticulously designed for the two tasks, which not only reduces the parameters but also allows the supervised task to guide the convergence of the unsupervised task, boosting the stability of unsupervised training. The qualitative and quantitative results demonstrated the proposed USTSFM has richer spatial details and more accurate predictions than the other state-of-the-art methods.

Rice Mapping and Various Stages of Rice Growth using Sentinel-1 SAR Data-A case study of Mahabubnagar District, Telangana

Space Based technology both optical and microwave SAR Data is used to monitor the Earth at regular intervals which has its application in various fields viz., agriculture, water, urban and so on. Synthetic Aperture Radar (SAR) has its advantage to penetrate through clouds, day-night imaging, all weather conditions and high resolution making it an effective system for crop monitoring, especially during Monsoon Season, as Optical Sensors doesn’t provide cloud free data. Sentinel-1 data has been used in recent studies to map and monitor rice-growing areas. The Sentinel-1 IW GRD data were used to monitor the ricegrowing area over Mahabubnagar District of Telangana State, which was processed in SNAP Software. Few plots were selected by conducting field survey during Kharif 2021. The Radar Vegetation Index (RVI) have been used to monitor the fluctuations in various stages of rice growth. This paper explains how multi temporal SAR observations can be used to monitor rice growing stages, as well as how to interpret σ°VHand σ° VV back scattering co-efficient.

Temporal Sequence Object-based CNN (TS-OCNN) for crop classification from fine resolution remote sensing image time-series

Accurate crop distribution mapping is required for crop yield prediction and field management. Due to rapid progress in remote sensing technology, fine spatial resolution (FSR) remotely sensed imagery now offers great opportunities for mapping crop types in great detail. However, within-class variance can hamper attempts to discriminate crop classes at fine resolutions. Multi-temporal FSR remotely sensed imagery provides a means of increasing crop classification from FSR imagery, although current methods do not exploit the available information fully. In this research, a novel Temporal Sequence Object-based Convolutional Neural Network (TS-OCNN) was proposed to classify agricultural crop type from FSR image time-series. An object-based CNN (OCNN) model was adopted in the TS-OCNN to classify images at the object level (i.e., segmented objects or crop parcels), thus, maintaining the precise boundary information of crop parcels. The combination of image time-series was first utilized as the input to the OCNN model to produce an ‘original’ or baseline classification. Then the single-date images were fed automatically into the deep learning model scene-by-scene in order of image acquisition date to increase successively the crop classification accuracy. By doing so, the joint information in the FSR multi-temporal observations and the unique individual information from the single-date images were exploited comprehensively for crop classification. The effectiveness of the proposed approach was investigated using multitemporal SAR and optical imagery, respectively, over two heterogeneous agricultural areas. The experimental results demonstrated that the newly proposed TS-OCNN approach consistently increased crop classification accuracy, and achieved the greatest accuracies (82.68% and 87.40%) in comparison with state-of-the-art benchmark methods, including the object-based CNN (OCNN) (81.63% and 85.88%), object-based image analysis (OBIA) (78.21% and 84.83%), and standard pixel-wise CNN (79.18% and 82.90%). The proposed approach is the first known attempt to explore simultaneously the joint information from image time-series with the unique information from single-date images for crop classification using a deep learning framework. The TS-OCNN, therefore, represents a new approach for agricultural landscape classification from multi-temporal FSR imagery. Besides, it is readily generalizable to other landscapes (e.g., forest landscapes), with a wide application prospect.

JPPL: A joint-polarization phase linking algorithm for phase optimization of TSPolInSAR data

• As a new polarimetric optimization, the joint-polarization phase linking (JPPL) is proposed to improve the phase quality. • The used joint diagonalization can extract the ESM phases from multi-polarimetric time-series coherence matrices. • Simulated experiment analyzes the interferometric phase quality of the HH, ESPO, and JPPL algorithms quantitatively. • Real experiment demonstrates that the proposed method reduces the speckle noise and restores the interferometric phase. In the time-series interferometric SAR (TSInSAR) technology, the phase linking (PL) algorithm can exploit all interferometric combinations to reconstruct the equivalent single master (ESM) interferograms for the phase quality improvement. According to the Cramer-Rao lower bound (CRLB) for PL, besides the number of spatial samples, the time-series coherence magnitude matrix determines the reconstruction performance of the ESM phases. With the abundance of time-series polarimetric SAR data, many polarimetric optimization algorithms for distributed scatterer (DS) have been introduced, mainly the averaged coherence magnitude maximization-based exhaustive search polarimetric optimization (ESPO) algorithm. However, traditional polarimetric optimization algorithms cannot work satisfactorily because of the unstable statistical characteristics. Similarly, the multi-temporal observations of each polarimetric channel contain interferometric information. From the perspective of redundant observation, introducing full-polarization information can increase the number of observations significantly. Therefore, based on the PL theory and the polarimetric optimization, this paper proposes to perform the joint diagonalization method to simultaneously extract the ESM interferometric phases of interest from three time-series coherence matrices in the Pauli basis polarizations, called the joint-polarization phase linking (JPPL) algorithm. In the experiments, the proposed JPPL algorithm can better improve the optimization performance of time-series interferometric phases of DSs than the HH and traditional ESPO algorithms in terms of speckle noise reduction and interferometric phase restoration.

Large-Scale Detection of the Tableland Areas and Erosion-Vulnerable Hotspots on the Chinese Loess Plateau

Tableland areas, featured by flat and broad landforms, provide precious land resources for agricultural production and human settlements over the Chinese Loess Plateau (CLP). However, severe gully erosion triggered by extreme rainfall and intense human activities makes tableland areas shrink continuously. Preventing the loss of tableland areas is of real urgency, in which generating its accurate distribution map is the critical prerequisite. However, a plateau-scale inventory of tableland areas is still lacking across the Loess Plateau. This study proposed a large-scale approach for tableland area mapping. The Sentinel-2 imagery was used for the initial delineation based on object-based image analysis and random forest model. Subsequently, the drainage networks extracted from AW3D30 DEM were applied for correcting commission and omission errors based on the law that rivers and streams rarely appear on the tableland areas. The automatic mapping approach performs well, with the overall accuracies over 90% in all four investigated subregions. After the strict quality control by manual inspection, a high-quality inventory of tableland areas at 10 m resolution was generated, demonstrating that the tableland areas occupied 9507.31 km2 across the CLP. Cultivated land is the dominant land-use type on the tableland areas, yet multi-temporal observations indicated that it has decreased by approximately 500 km2 during the year of 2000 to 2020. In contrast, forest and artificial surfaces increased by 57.53% and 73.10%, respectively. Additionally, we detected 455 vulnerable hotspots of the tableland with a width of less than 300 m. Particular attention should be paid to these areas to prevent the potential split of a large tableland, accompanied by damage on roads and buildings. This plateau-scale tableland inventory and erosion-vulnerable hotspots are expected to support the environmental protection policymaking for sustainable development in the CLP region severely threatened by soil erosion and land degradation.

Multi-Season Phenology Mapping of Nile Delta Croplands Using Time Series of Sentinel-2 and Landsat 8 Green LAI.

Space-based cropland phenology monitoring substantially assists agricultural managing practices and plays an important role in crop yield predictions. Multitemporal satellite observations allow analyzing vegetation seasonal dynamics over large areas by using vegetation indices or by deriving biophysical variables. The Nile Delta represents about half of all agricultural lands of Egypt. In this region, intensifying farming systems are predominant and multi-cropping rotations schemes are increasing, requiring a high temporal and spatial resolution monitoring for capturing successive crop growth cycles. This study presents a workflow for cropland phenology characterization and mapping based on time series of green Leaf Area Index (LAI) generated from NASA’s Harmonized Landsat 8 (L8) and Sentinel-2 (S2) surface reflectance dataset from 2016 to 2019. LAI time series were processed for each satellite dataset, which were used separately and combined to identify seasonal dynamics for a selection of crop types (wheat, clover, maize and rice). For the combination of L8 with S2 LAI products, we proposed two time series smoothing and fitting methods: (1) the Savitzky–Golay (SG) filter and (2) the Gaussian Processes Regression (GPR) fitting function. Single-sensor and L8-S2 combined LAI time series were used for the calculation of key crop Land Surface Phenology (LSP) metrics (start of season, end of season, length of season), whereby the detection of cropland growing seasons was based on two established threshold methods, i.e., a seasonal or a relative amplitude value. Overall, the developed phenology extraction scheme enabled identifying up to two successive crop cycles within a year, with a superior performance observed for the seasonal than for the relative threshold method, in terms of consistency and cropland season detection capability. Differences between the time series collections were analyzed by comparing the phenology metrics per crop type and year. Results suggest that L8-S2 combined LAI data streams with GPR led to a more precise detection of the start and end of growing seasons for most crop types, reaching an overall detection of 74% over the total planted crops versus 69% with S2 and 63% with L8 alone. Finally, the phenology mapping allowed us to evaluate the spatial and temporal evolution of the croplands over the agroecosystem in the Nile Delta.

Estimation of Daily All-Wave Surface Net Radiation With Multispectral and Multitemporal Observations From GOES-16 ABI

As a vital parameter describing the Earth surface energy budget, surface all-wave net radiation ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> ) drives many physical and biological processes. Remote estimation of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> using satellite data is an effective approach to monitor the spatial and temporal dynamics of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> . Accurate daily <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> estimation typically depends on the spatio-temporal resolutions of satellite data. There are currently few high-spatial-resolution daily <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> products from polar-orbiting satellite data, and they exhibit limited accuracy due to sparse diurnal observations. In addition, traditional estimation approaches typically require cloud mask and clear-sky albedo as inputs and ignore the length ratio of daytime (LRD), which may lead to large errors. To overcome these challenges and obtain <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> data with improved spatial resolution and accuracy, an operational approach was proposed in this study to derive daily 1-km <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> , which takes the advantages from a radiative transfer model, a machine learning algorithm, and multispectral and dense diurnal temporal information of geostationary satellite observations. An improved all-sky hybrid model (AHM) coupling radiative transfer simulations with a random forest (RF) model was first developed to estimate the shortwave net radiation ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{ns}$ </tex-math></inline-formula> ). Then, another RF model was developed to estimate the daily <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> from <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{ns}$ </tex-math></inline-formula> , incorporating the LRD, which is called extended hybrid model (EHM). Data from the Advanced Baseline Imager (ABI) onboard the new-generation Geostationary Operational Environmental Satellite (GOES)-16 with a 5-min temporal resolution and a 1-km spatial resolution were used to test the proposed method. Compared to traditional lookup table (LUT) algorithms, the results show that AHM not only makes the process of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{ns}$ </tex-math></inline-formula> estimation simple and efficient but also has high accuracy in estimating instantaneous all-sky <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{ns}$ </tex-math></inline-formula> . Benefiting from high spatio-temporal resolutions, our daily <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{ns}$ </tex-math></inline-formula> estimates using GOSE-16 data exhibited superior performance compared to using the 1-km Moderate Resolution Imaging Spectroradiometer (MODIS) and 1° Clouds and the Earth’s Radiant Energy System (CERES) product. Using accurate daily <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{ns}$ </tex-math></inline-formula> estimates and LRD as inputs, the EHM model shows reasonably good results for estimating <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> , RMSE, and bias of 0.91, 20.95 W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> , and −0.05 W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> , respectively). Maps of 1-km <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{ns}$ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> exhibit similar spatial patterns to those from the 1° CERES product, but with substantially more spatial details. Overall, the proposed <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> retrieval scheme can accurately estimate all-sky 1-km <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{ns}$ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> at mid- to low-latitudes and can be easily adapted and applied to other GOES- 16-like satellites, such as Himawari-8, Meteosat Third Generation (MTG), and Fenyun-4. This study demonstrates the advantages of estimating <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R_{n}$ </tex-math></inline-formula> using geostationary satellites with improved accuracy and resolutions.

DAT-CNN: Dual Attention Temporal CNN for Time-Resolving Sentinel-3 Vegetation Indices

The synergies between Sentinel-3 (S3) and the forthcoming Fluorescence Explorer (FLEX) mission bring us the opportunity of using S3 vegetation indices (VI) as proxies of the solar-induced chlorophyll fluorescence (SIF) that will be captured by FLEX. However, the highly dynamic nature of SIF demands a very temporally accurate monitoring of S3 VIs to become reliable proxies. In this scenario, this paper proposes a novel temporal reconstruction convolutional neural network (CNN), named dual attention temporal CNN (DAT-CNN), which has been specially designed for time-resolving S3 VIs using S2 and S3 multi-temporal observations. In contrast to other existing techniques, DAT-CNN implements two different branches for processing and fusing S2 and S3 multi-modal data while further exploiting inter-sensor synergies. Besides, DAT-CNN also incorporates a new spatial-spectral and temporal attention module to suppress uninformative spatial-spectral features while focusing on the most relevant temporal stamps for each particular prediction. The experimental comparison, including several temporal reconstruction methods and multiple operational Sentinel data products, demonstrates the competitive advantages of the proposed model with respect to the state-of-the-art. The codes of this paper will be available at https://github.com/ibanezfd/DATCNN.

Combination of Aerial, Satellite, and UAV Photogrammetry for Quantifying Rock Glacier Kinematics in the Dry Andes of Chile (30°S) Since the 1950s

The diachronic analysis of aerial and satellite imagery, uncrewed aerial vehicle (UAV) andin situsurveys obtained between 1956 and 2019 are employed to analyse landform surface kinematics for the Tapado site located in the Dry Andes of Chile. A feature tracking procedure was used between series of orthorectified and co-registered images to calculate surface velocities on several ice-debris landforms, including rock glaciers and debris-covered glaciers. For the active rock glaciers, the results exhibit typical viscous flow, though local destabilisation process seems to occur, increased velocities since 2000 (&amp;gt;1 m/yr) and terminus advance. Nevertheless, the debris-covered glaciers displays heterogeneous spatial patterns of surface velocities, together with collapse (downwasting) associated with the development of thermokarst depressions and supraglacial ponds. Our findings show that surface kinematics and multitemporal observations derived from different sensors are valuable tools for differentiating between glacial and periglacial features. The pluri-decadal time series since 1956 constitute a unique dataset for documenting the surface kinematics of creeping mountain permafrost in the Southern Hemisphere. The approach developed in this work offers a way forward to reconstruct the recent behaviour of glacial and periglacial features in the Andes, where archival aerial photographs are available but have not previously been processed rigorously to obtain an accurate assessment of landform kinematics.

Modeling Vegetation Water Stress over the Forest from Space: Temperature Vegetation Water Stress Index (TVWSI)

The conventional Land Surface Temperature (LST)–Normalized Difference Vegetation Index (NDVI) trapezoid model has been widely used to retrieve vegetation water stress. However, it has two inherent limitations: (1) its complex and computationally intensive parameterization for multi-temporal observations and (2) deficiency in canopy water content information. We tested the hypothesis that an improved water stress index could be constructed by the representation of canopy water content information to the LST–NDVI trapezoid model. Therefore, this study proposes a new index that combines three indicators associated with vegetation water stress: canopy temperature through LST, canopy water content through Surface Water Content Index (SWCI), and canopy fractional cover through NDVI in one temporally transferrable index. Firstly, a new optical space of SWCI–NDVI was conceptualized based on the linear physical relationship between shortwave infrared (SWIR) and soil moisture. Secondly, the SWCI–NDVI feature space was parameterized, and an index d(SWCI, NDVI) was computed based on the distribution of the observations in the SWCI–NDVI spectral space. Finally, standardized LST (LST/long term mean of LST) was combined to d(SWCI, NDVI) to give a new water stress index, Temperature Vegetation Water Stress Index (TVWSI). The modeled soil moisture from the Australian Water Resource Assessment—Landscape (AWRA-L) and Soil Water Fraction (SWF) from four FLUXNET sites across Victoria and New South Wales were used to evaluate TVWSI. The index TVWSI exhibited a high correlation with AWRA-L soil moisture (R2 of 0.71 with p &lt; 0.001) and the ground-based SWF (R2 of 0.25–0.51 with p &lt; 0.001). TVWSI predicted soil moisture more accurately with RMSE of 21.82 mm (AWRA-L) and 0.02–0.04 (SWF) compared to the RMSE ranging 28.98–36.68 mm (AWRA-L) and 0.03–0.05 (SWF) were obtained for some widely used water stress indices. The TVWSI could also be a useful input parameter for other environmental models.

Time Series InSAR Three-Dimensional Displacement Inversion Model of Coal Mining Areas Based on Symmetrical Features of Mining Subsidence

The three-dimensional (3-D) displacements of mining areas is the basis of studying the mining subsidence law and obtaining surface movement parameters. The traditional multi-temporal interferometry synthetic aperture radar (InSAR) technology can only obtain the surface deformation in line-of-sight (LOS) direction, even if some methods can obtain the 3-D displacements of mining area based on InSAR. However, it has high data requirements for data types, which are not conducive to the inversion of 3-D displacements. In this paper, the symmetry of the surface basin caused by mining subsidence under different mining degrees is analyzed. According to the basic symmetrical features of mining subsidence—that the surface vertical displacement and horizontal displacement in near horizontal coal seam is symmetrical with respect to the main section of the basin, combined with time series InSAR technology (TS-InSAR)—a novel method for retrieving the 3-D displacement results from a single-geometry InSAR dataset based on symmetrical features (hereafter referred to as the SGI-SF method) is proposed. The SGI-SF method first generates multi-temporal observations of LOS displacement from a single-geometry InSAR dataset, and then transforms them into multi-temporal observations of 3-D displacement datasets according to symmetrical features. There is no necessity to obtain the surface movement parameters from the measured data to calculate 3-D displacement fields. Finally, the time series of 3-D displacements are estimated from multi-temporal 3-D displacements using the singular value decomposition (SVD) method. Nine descending Sentinel-1 images from the Yulin mining area of China are used to test the proposed SGI-SF method. The results show that the average root mean square errors (RMSE) in the vertical and horizontal direction of the three-dimensional deformations are approximately 9.28 mm and 13.10 mm, respectively, which are much smaller than mining-induced displacements and can provide support for deformation monitoring in mining areas.

Early detection of bark beetle infestation in Norway spruce forests of Central Europe using Sentinel-2

In the past decade, massive outbreaks of bark beetles (Ips spp.) have caused large-scale decline of coniferous-dominated, prevailingly managed forests of Central Europe. Timely detection of newly infested trees is important for minimizing economic losses and effectively planning forest management activities to stop or at least slow outbreaks. With the advancement of Copernicus services, a pair of Sentinel-2 satellites provides a unique remote sensing data source of multi-temporal observations in high spatial resolution on the scale of individual forest stands (although not allowing for individual tree detection). This study investigates the potential for using seasonal trajectories of Sentinel-2 bands and selected vegetation indices in early detection of bark beetle infestation (so–called green-attack stage detection) in Norway spruce monoculture forests in the Czech Republic. Spectral trajectories of nine bands and six vegetation indices were constructed for the 2018 vegetation season using 14 satellite observations from April to November to distinguish four infestation classes. We used a random forest algorithm to classify healthy (i.e., stands not infested) and infested trees with various trajectories of decay. The seasonal trajectories of vegetation indices separated the infestation classes better than did the individual bands. Among the most promising vegetation indices we identified the tasselled cap wetness (TCW) component and normalized difference index constructed from near and shortwave infrared bands. Analysing the inter-annual change of the indices was more promising for early detection than is single-date classification. It achieved 96% classification accuracy on day of year 291 for the tested data set.The algorithm for early detection of tree infestation based on the assessment of seasonal changes in TCW was applied on the time series of Sentinel-2 observations from 2019 and its outputs were verified using field observations of forest conditions conducted on 80 spruce forest plots (located in spruce monoculture stands). The overall accuracy of 78% was achieved for the separation of healthy and green-attack classes. Our study highlights the great potential of multi-temporal remote sensing and the use of shortwave infrared wavelengths for early detection of spruce forest decline caused by bark beetle infestation.

A Simple Approach for Mapping Forest Cover from Time Series of Satellite Data

Forest cover mapping based on multi-temporal satellite observations usually uses dozens of features as inputs, which requires huge training data and leads to many ill effects. In this paper, a simple but efficient approach was proposed to map forest cover from time series of satellite observations without using classifiers and training data. This method focuses on the key step of forest mapping, i.e., separation of forests from herbaceous vegetation, considering that the non-vegetated area can be easily identified by the annual maximum vegetation index. We found that the greenness of forests is generally stable during the maturity period, but a similar greenness plateau does not exist for herbaceous vegetation. It means that the mean greenness during the vegetation maturity period of forests should be larger than that of herbaceous vegetation, while its standard deviation should be smaller. A combination of these two features could identify forests with several thresholds. The proposed approach was demonstrated for mapping the extents of different forest types with MODIS observations. The results show that the overall accuracy ranges 91.92–95.34% and the Kappa coefficient is 0.84–0.91 when compared with the reference datasets generated from fine-resolution imagery of Google Earth. The proposed approach can greatly simplify the procedures of forest cover mapping.

Multi-Temporal Unmanned Aerial Vehicle Remote Sensing for Vegetable Mapping Using an Attention-Based Recurrent Convolutional Neural Network

Vegetable mapping from remote sensing imagery is important for precision agricultural activities such as automated pesticide spraying. Multi-temporal unmanned aerial vehicle (UAV) data has the merits of both very high spatial resolution and useful phenological information, which shows great potential for accurate vegetable classification, especially under complex and fragmented agricultural landscapes. In this study, an attention-based recurrent convolutional neural network (ARCNN) has been proposed for accurate vegetable mapping from multi-temporal UAV red-green-blue (RGB) imagery. The proposed model firstly utilizes a multi-scale deformable CNN to learn and extract rich spatial features from UAV data. Afterwards, the extracted features are fed into an attention-based recurrent neural network (RNN), from which the sequential dependency between multi-temporal features could be established. Finally, the aggregated spatial-temporal features are used to predict the vegetable category. Experimental results show that the proposed ARCNN yields a high performance with an overall accuracy of 92.80%. When compared with mono-temporal classification, the incorporation of multi-temporal UAV imagery could significantly boost the accuracy by 24.49% on average, which justifies the hypothesis that the low spectral resolution of RGB imagery could be compensated by the inclusion of multi-temporal observations. In addition, the attention-based RNN in this study outperforms other feature fusion methods such as feature-stacking. The deformable convolution operation also yields higher classification accuracy than that of a standard convolution unit. Results demonstrate that the ARCNN could provide an effective way for extracting and aggregating discriminative spatial-temporal features for vegetable mapping from multi-temporal UAV RGB imagery.

Geo-Object-Based Land Cover Map Update for High-Spatial-Resolution Remote Sensing Images via Change Detection and Label Transfer

Land cover (LC) information plays an important role in different geoscience applications such as land resources and ecological environment monitoring. Enhancing the automation degree of LC classification and updating at a fine scale by remote sensing has become a key problem, as the capability of remote sensing data acquisition is constantly being improved in terms of spatial and temporal resolution. However, the present methods of generating LC information are relatively inefficient, in terms of manually selecting training samples among multitemporal observations, which is becoming the bottleneck of application-oriented LC mapping. Thus, the objectives of this study are to speed up the efficiency of LC information acquisition and update. This study proposes a rapid LC map updating approach at a geo-object scale for high-spatial-resolution (HSR) remote sensing. The challenge is to develop methodologies for quickly sampling. Hence, the core step of our proposed methodology is an automatic method of collecting samples from historical LC maps through combining change detection and label transfer. A data set with Chinese Gaofen-2 (GF-2) HSR satellite images is utilized to evaluate the effectiveness of our method for multitemporal updating of LC maps. Prior labels in a historical LC map are certified to be effective in a LC updating task, which contributes to improve the effectiveness of the LC map update by automatically generating a number of training samples for supervised classification. The experimental outcomes demonstrate that the proposed method enhances the automation degree of LC map updating and allows for geo-object-based up-to-date LC mapping with high accuracy. The results indicate that the proposed method boosts the ability of automatic update of LC map, and greatly reduces the complexity of visual sample acquisition. Furthermore, the accuracy of LC type and the fineness of polygon boundaries in the updated LC maps effectively reflect the characteristics of geo-object changes on the ground surface, which makes the proposed method suitable for many applications requiring refined LC maps.