Remote Sensing of Environment最新文献

Land Surface Temperature (LST) and Land Surface Emissivity (LSE) are key properties of natural materials essential for scientific analysis. Existing retrieval techniques, however, impede the automatic retrieval of LST and LSE across various observational contexts due to the frequent unavailability of in-situ atmospheric data or blackbody references for atmospheric compensation. To address this, we propose the Inter-Channel Correlated Automatic Atmospheric Compensation (ICCAAC) method, seamlessly integrated with the ASTER-TES approach, enabling direct retrieval of LST and LSE from at-sensor radiance without prerequisite atmosphere and land surface information. ICCAAC innovatively models the inter-channel relationships among atmospheric constituents to streamline the Radiative Transfer Equation (RTE), which includes the transmittance reconstruction with neighboring channels and the atmospheric upwelling radiance simplification with the atmospheric vertical characteristics, tackling the complex challenge of retrieving ground-leaving radiance. Coupled with ASTER-TES and under the LSE smoothness constraint, along with a lookup table for atmospheric downwelling radiance, this method facilitates accurate LST and LSE retrieval. In controlled simulations, ICCAAC shows a maximum brightness temperature error of 1.6 K for Water Vapor Content (WVC) under 1.5 $g/{\mathrm{cm}}^2$. A spectral sensitivity analysis suggests the optimal performance of ICCAAC for full width at half maximum (FWHM) range extending beyond the channel interval. Applied to real-world airborne data from Hypercam-LW and HyTES, ICCAAC-TES validates its accuracy with an LST and LSE error margin of 1.2 K and 0.014, respectively, corroborated by seventeen distinct ground validations in Hypercam-LW imagery. Comparative analysis with five varied HyTES observation scenes reveals an LST discrepancy of around 1.2 K and notable emissivity textures in LSE images, particularly in mineral terrains. These outcomes underscore the efficacy of ICCAAC-TES, advocating its suitability for automated LST and LSE retrieval in airborne survey applications.

The Tibetan Plateau and surroundings, commonly referred to as the Third Pole region, has the largest ice store outside the Arctic and Antarctic regions. Glacial lakes in the Third Pole region are expanding rapidly as glaciers thin and retreat. The Landsat satellite series is the most popular for mapping glacial lakes, benefiting from long-term archived data and suitable spatial resolution (30 m since ∼1990). However, the homogeneous mapping of high-quality, large-scale, and multi-temporal glacial lake inventories using Landsat imagery relies heavily on visual inspection and manual editing due to mountain shadows, wet ice, frozen lakes, and snow cover on lake boundaries, which is time consuming and labour-intensive. Deep learning methods have been applied to glacial lake extraction in the Third Pole and other regions, yet these methods are either concentrated on small test sites without large-scale applications or in polar regions. In this study, several classical deep convolutional neural networks were evaluated, and the DeepLabv3+ with Mobilenetv3 backbone performed best, with a high accuracy of mean intersection over union (mIoU) of 94.8 % and a low loss error of 0.4 %. The proposed method demonstrated robustness in challenging conditions such as mountain shadows, frozen or partially frozen lakes, wet ice and river contact, all without requiring extensive manual correction. Compared with manual delineation, the model's prediction has a precision rate of 86 %, recall rate of 85 %, and F1-score of 85 %. The area extracted by the model shows a strong correlation with the manual delineation (r² = 0.97, slope = 0.94) and a high intersection over union (IoU > 0.8) of the predicted areas. A test of large-scale glacial lake mapping based on the developed automated model in 2020 across the Third Pole region shows the robust performance with 29,429 glacial lakes larger than 0.0054 km² with a total area of ∼1779.9 km² (including non-glacier-fed lakes). The model trained in this study can be fine-tuned for large-scale mapping of glacial lakes in other mountain regions worldwide.

Peatlands store enormous amounts of carbon in a peat layer, the formation and preservation of which can only occur under waterlogged conditions. Monitoring peatland moisture conditions is critically important because a decrease in moisture leads to peat oxidation and the release of accumulated carbon back into the atmosphere as a greenhouse gas. Optical remote sensing enables the indirect monitoring of peatland moisture conditions by identifying moisture-driven changes in vegetation spectral signatures. The vegetation of northern peatlands is dominated by Sphagnum mosses, whose spectral signatures are known to be highly sensitive to changes in moisture content. In this study, we tested methods to estimate Sphagnum moisture content from spectral data using seven spectral moisture indices, the OPtical TRApezoid Model (OPTRAM) and the Continuous Wavelet Transform (CWT). This study was based on data representing nine Sphagnum species sampled from two habitats in southern boreal peatlands in Finland. Our results showed that both multi- and hyperspectral data can be used to estimate the moisture content of Sphagnum mosses. Nevertheless, the optimal retrieval method depended on habitat characteristics. Using hyperspectral data, we found that: (i) the CWT exhibited superior performance for all studied moss species ($R_{\text{Marg}}^2$= 0.72, ICC = 0.40), (ii) the exponential OPTRAM performed best for the mesotrophic species ($R_{\text{Marg}}^2$= 0.70, ICC = 0.08), and (iii) the Modified Moisture Stress Index (MMSI) yielded the best results ($R_{\text{Marg}}^2$= 0.68, ICC = 0.55) for the ombrotrophic species. Furthermore, we demonstrated that using multispectral data instead of hyperspectral data provides comparable results in moisture estimation when used as input with OPTRAM or Moisture Stress Index (MSI). This approach could lead to new insights into the moisture dynamics in Sphagnum-dominated peatlands over the span of the multispectral satellite era.

Launched by China in 2021, the Sustainable Development Goals Science Satellite 1 (SDGSAT-1) is the world's first science satellite dedicated to serving the United Nations 2030 Agenda for Sustainable Development Goals. In keeping with international aims of this 2030 agenda, the SDGSAT-1 data will be made available for open accee without any restrictions. The Thermal Infrared Spectrometer (TIS) onboard SDGSAT-1 has three thermal infrared channels with a high spatial resolution of 30 m. This enables precise monitoring of land surface temperature (LST), which is one of the most important variables measured by satellite remote sensing. This paper presents the development and validation of three split-window (SW) algorithms and the temperature and emissivity separation (TES) algorithm for SDGSAT-1 TIS data. These algorithms were rigorously tested through simulation, application, and validation to assess their retrieval accuracy and sensitivity. Simulation results indicate that the theoretical accuracy of the SW algorithms exceeds 1.0 K in most cases, and the TES algorithm shows higher retrieval accuracy with an average LST Root Mean Square Error (RMSE) of 0.60 K. With consideration of the comprehensive effects of instrument noise, land surface emissivity, and atmospheric parameter error, the LST retrieval accuracy of SW algorithms remains better than 1.7 K, and that of the TES algorithm is better than 1.5 K under most conditions. The ground validation utilized site data from the Heihe Integrated Observatory Network and the Surface Radiation budget network. The SW and TES algorithms achieved an accuracy of approximately 1.75 and 1.9 K, respectively. Additionally, a cross-validation based on Moderate Resolution Imaging Spectroradiometer (MODIS) data indicated average RMSDs of approximately 2.25 K for SW algorithms and 3.84 K for TES algorithm. Among the algorithms, the three-channel SW algorithm SW-2 has the best overall performance and is recommended as the LST retrieval method for SDGSAT-1 data. The TES algorithm is also suitable for SDGSAT-1 images because of its ability to retrieve LST and emissivity during both daytime and nighttime.

Ecosystem dynamics and ecological disturbances manifest as breakpoints in long-term multispectral remote sensing time series. Typically, these breakpoints are captured using univariate methods applied individually to each band, with subsequent integration of the results. However, multivariate analysis provides a promising way to fully incorporate the multispectral bands into breakpoints detection methods, but it has been rarely applied in monitoring ecosystem dynamics and detecting ecological disturbances. In this research, we developed a multivariate algorithm, named breakpoints-Detection algoRithm using MultivAriate Time series (DRMAT). DRMAT can fully use multispectral bands simultaneously with the consideration of the inter-correlation among bands. It decomposes a multivariate time series into trend, seasonality, and noise, iteratively segmenting the detrended/de-seasonalized signals. We quantitatively evaluated DRMAT using both simulated multivariate data and randomly sampled real-world data, including subtle land cover changes caused by forest disturbances (depletions) and recovery (return of vegetation), as well as subtle changes over a broad range of land cover types. We also qualitatively assessed DRMAT in mapping real-world disturbances. For simulated data with prescribed breakpoints in both trend and seasonality, DRMAT detected breakpoints in trend with an F1 score of 85.5 % and in seasonality with an F1 score of 91.7 %. For real-world data in forested land cover, DRMAT unveiled both disturbances and subsequent recovery with an F1 score of 95.1 % for disturbances and 77.1 % for recovery. It detected disturbances in broader land cover types with an F1 score of 84.0 %. We demonstrated that using all-band data was more accurate than using selected bands in breakpoint detection. The inclusion of vegetation indices as model inputs did not improve accuracy unless the original input bands lacked the specific band information in the vegetation indices. As a multivariate approach, DRMAT leverages the full information in the multispectral data and avoids the necessity of integrating results derived from individual bands.

The acquisition of high-quality reference data is essential for effectively modelling forest attributes. Incorporating close-range Light Detection and Ranging (LiDAR) systems into the reference data collection stage of remote sensing-based forest inventories can not only increase data collection efficiency but also increase the number of attributes measured with high quality. Therefore, we propose a model-based forest inventory method that uses reference data collected by a car-mounted mobile laser scanning (MLS) system along boreal forest roads. This approach is used for the estimation of diameter at breast height (DBH) and stem volume at the individual tree-level from airborne laser scanning (ALS) data. In addition, we compare the estimates obtained using the proposed method with the ones derived from reference data collected by traditional field inventory of 265 field plots systematically distributed over the study area. The accuracy of the estimates remained comparable regardless of the reference dataset used for estimation of DBH and stem volume. When using the field inventory dataset for model training, the root mean square error (RMSE) of DBH estimates were 4.06 cm (18.8 %) for Norway spruce trees, 6.3 cm (29.6 %) for Scots pine and 8.61 cm (55.9 %) for deciduous trees. Similarly, when evaluating predictions based on the MLS dataset as reference, RMSEs were equal to 3.97 cm (18.4 %) for Norway spruce, 6.12 cm (28.8 %) for Scots pine, and 8.98 cm (58.3 %) for deciduous trees. In general, biases were below 1 cm for most species classes, with the exception of deciduous trees. The accuracy of stem volume also had RMSEs varying across different tree species. For the estimates based on traditional field inventory, the RMSEs were 0.176 m³ (38.8 %) for Norway spruce, 0.228 m³ (52.4 %) for Scots pine and 0.246 m³ (158 %) for deciduous trees. When using the MLS dataset as a reference, the RMSEs were equal to 0.176 m³ (38.8 %), 0.228 m³ (52.4 %), and 0.246 m³ (158 %) for Norway spruce, Scots pine, and deciduous trees, respectively. Car-mounted MLS demonstrated its potential as an efficient alternative for collecting reference data in remote sensing-based forest inventories, which could complement traditional methods.

Sea surface currents are of significant importance in various scientific and maritime applications. There are several measurement techniques available to study surface currents, however, they have limitations in spatial coverage and resolution. This study presents a proof-of-concept for a new measurement principle that relies on the difference between a ship’s speed relative to water and land. The approach involves estimating the ship speed vector relative to water from optical satellite imagery of Kelvin wakes. This ship speed vector is subtracted from the ship speed over ground, which is determined from Automatic Identification System (AIS) data, to estimate the surface current. A case study in the Strait of Gibraltar was performed using two months of Sentinel-2 imagery, which yielded 81 visible Kelvin wakes over 25 images. Surface currents were estimated in directions parallel and perpendicular to the ship’s sailing line for each Kelvin wake. The estimated currents were validated with respect to surface currents derived from High-Frequency Radars (HFRs) and modelled currents from the Copernicus Marine Environmental Monitoring Service (CMEMS). The uncertainty in the two surface current components was estimated using triple collocation. After removing 12 data points with large ship course variability, standard deviations of 0.14 and 0.16 m s⁻¹ were estimated for the surface currents along and across the sailing line, respectively. Despite limitations in measurement frequency due to satellite revisit times, cloud cover and Kelvin wake visibility, this new method can provide accurate estimates of sea surface currents in regions with high vessel density.

Land surface temperature (LST) derived from satellite observations and weather modeling has been widely used for investigating Earth surface-atmosphere energy exchange and radiation budget. However, satellite-derived LST has a trade-off between spatial and temporal resolutions and missing observations caused by clouds, while there are limitations such as potential bias and expensive computation in model calibration and simulation for weather modeling. To mitigate those limitations, we proposed a WRFM framework to estimate LST at a spatial resolution of 1 km and temporal resolution of an hour by integrating the Weather Research and Forecasting (WRF) model and MODIS satellite data using the morphing technique. We tested the framework in eight counties, Iowa, USA, including urban and rural areas, to generate hourly LSTs from June 1st to August 31st, 2019, at a 1 km resolution. Upon evaluation with in-situ LST measurements, our WRFM framework has demonstrated its ability to capture hourly LSTs under both clear and cloudy conditions, with a root mean square error (RMSE) of 2.63 K and 3.75 K, respectively. Additionally, the assessment with satellite LST observations has shown that the WRFM framework can effectively reduce the bias magnitude in LST from the WRF simulation, resulting in a reduction of the average RMSE over the study area from 4.34 K (daytime) and 4.12 K (nighttime) to 2.89 K (daytime) and 2.75 K (nighttime), respectively, while still capturing the hourly patterns of LST. Overall, the WRFM is effective in integrating the complementary advantages of satellite observations and weather modeling and can generate LSTs with high spatiotemporal resolutions in areas with complex landscapes (e.g., urban).

The vegetation dynamics in highly heterogeneous landscapes (e.g., riparian vegetation surrounding waterholes and oases) are difficult to detect from large (e.g., MODIS) and moderate (e.g., Landsat) spatial resolution remote sensing products. Within a “classify-to-monitor” approach, a method to monitor spatially heterogeneous riparian vegetation dynamics is developed by integrating high spatial resolution discrete return airborne LiDAR data (1 m pixels) with moderate resolution Landsat fraction of Photosynthetically Active Radiation absorbed by vegetation (fPAR) data (30 m). LiDAR was used to identify and classify vegetation surrounding permanent waterholes within the Cooper Creek floodplain, in dryland Australia. These waterholes are important areas for ecological conservation given their highly spatially heterogeneous vegetation structure. Landsat fPAR was temporally decomposed into persistent and recurrent components and then integrated with the LiDAR-derived vegetation classes. The LiDAR data were used as a mask to separate the fPAR signal of each vegetation class, capturing their specific dynamics and which fPAR component they are associated with. The newly developed method provides the means to improve the interpretation of Landsat fPAR by monitoring distinct vegetation functional groups within each Landsat pixel. Results showed that LiDAR data provided good estimates of vegetation cover compared to field measurements (R²=0.952). LiDAR data identified different vegetation structural classes within the riparian zone. The integration of LiDAR and Landsat data permitted the distinction of temporal patterns of each vegetation structural class, uncovering the specific temporal and spatial variability of fPAR that would otherwise be undetected. Landsat fPAR provided information on which vegetation component contributed to the fPAR variability in each class, thus providing the means for enhanced ecological interpretation of the temporally decomposed fPAR components. The method can be applied to other similar highly spatially heterogeneous ecosystems to monitor structurally specific vegetation dynamics more accurately than if only using moderate spatial resolution time-series optical satellite imagery.

Accurate nearshore bathymetry estimation remains a critical challenge, impacting coastal forecasting evolution assessments through the inaccuracies in both in-situ and remote sensing surveys. This article introduces the Satellite Derived Bathymetry (SDB) temporal correlation method, showcasing its ability in deriving accurate nearshore bathymetry from one minute spaceborne videos. The approach utilises correlation of pixel intensity time series, shifted in time and space, extracted from a frame stack within a defined correlation window. The resulting correlation is then projected using the Radon Transform to infer wave characteristics (celerity and wavelength) for the estimation of depth through wave linear dispersion. Moreover, the adaptation of the correlation window based on a first wavelength estimation provided a more focused assessment of the wavefield that reveals morphological features such as sandbars in the bathymetric estimation. The method’s capabilities using adapted correlation window is illustrated through its application to a metric resolution Jilin satellite video (57 s at 5 Hz) along the Saint-Louis coast in Senegal. Through this demonstration, the temporal correlation method is among the first SDB methods to successfully capture the submerged sandbar along a beach. Comparison against in-situ measurements conducted three years prior to the video acquisition shows a good agreement with a bias of 0.97 m within the initial 2 km of the cross-shore profile. Furthermore, the application of previously developed sky-glint surface elevation analysis on video pixel intensity, prior to the bathymetry estimation, significantly reduces the bias to 0.44 m in the Saint-Louis estimation. This article highlights the potential applications of future Earth observation satellite missions that will capture image sequences (or videos) such as CO3D (CNES/Airbus).