Science of Remote Sensing最新文献

Land Surface Temperature (LST) is important for diagnosing surface energy balance in land surface models (LSMs). However, LST simulation in current LSMs tends to show large cold biases, partially due to the reason that the model's prescribed vegetation parameters (e.g., Leaf Area Index (LAI) and Fraction of Vegetation Cover (FVC)) are misrepresented, especially in regions with complex topography and climate such as Tibetan Plateau. Recent advancements in remote sensing technologies provide a unique opportunity to improve the model's vegetation parameters at large scales. In this study, we practice two experiments to improve LST simulations in Noah-MP LSM by (1) incorporating LAI and FVC from the Global Land Surface Satellite (GLASS) remote sensing product (exp_RS); and (2) incorporating an empirical LAI and FVC parameterization scheme based on the soil temperature stress factor (exp_RL02). Results show that the effect of vegetation on simulated LST is the most significant in summer season when the model-satellite LAI and FVC differences are the largest. Compared to the default experiment that uses static LAI and FVC values from the model's look-up table (exp_CTL), the results in exp_RS and exp_RL02 show domain-wide improvement of the simulated LST. The LAI and FVC effect on LST are also well reflected in model's energy budget components (i.e., longwave emissivity, sensible and latent heat fluxes, etc). Validation of the model simulated soil temperature with in-situ observations further demonstrate the model improvements. Our study underscores the important role of vegetation in regulating surface energy transfer processes. Our study also highlights the feasibility and benefit of incorporating remote sensing data in improving land surface model simulations.

Detailed maps of vegetation composition are vital for restoration planning, implementation, and monitoring, particularly at early stages of succession. This is usually accomplished through ground surveys, which can be costly and impractical depending on extent and accessibility, or conducted at too broad a spatial scale. In this study, we propose a methodology for mapping regenerating vegetation composition at 2 × 2 m² spatial resolution, using very high spatial resolution (<1 m) remote sensing imagery obtained from remotely piloted aerial systems (RPAS) in conjunction with digital aerial photogrammetry (DAP) techniques for reconstructing vegetation structure. We applied logistic regression on multispectral orthomosaics, clusters of vegetation structure, and local illumination estimates to develop presence-absence models for eight key plant groups at various taxonomic levels as well as six plant functional types (conifer tree seedlings, grasses, tall- and low-growing forbs, shrubs, and mosses). Our results show higher accuracies for plant functional types (mean F-score = 0.67) compared to lower taxonomic levels (0.57). Notably, shrubs (F-score = 0.79), low-growing forbs (0.70), and mosses (0.69) exhibited the highest accuracies, while grasses (0.46), the aster family (Asteraceae spp; 0.48), and spruce seedlings (Picea spp; 0.54) demonstrated lower accuracies. Vegetation structure variables were identified as the most influential in the models, with mean NIRv ranking highest among spectral variables. High average ranks of spectral variation metrics (e.g., standard deviation of NIRv) implied the influence of environmental determinants such as plant co-occurrences and micro-habitat conditions, which drive spectral variation. Discrete composition maps were produced for three restoration sites and analogous wildfire-disturbed sites. Plant compositions found at one site pair exhibited similarity (Bray-Curtis = 0.28), however, certain key plant groups covered larger extents of the restoration site than anticipated. Willows (Salix spp; 25.4% vs. 9.3%), which are typically planted for soil stabilization and obstruction, and clovers (Trifolium spp; 11.1% vs. 3.6%), which represent non-native agronomic vegetation, were prominent. The developed methodology facilitates the generation of detailed plant composition maps, aiding evaluations of vegetation patterns that are difficult to discern visually or through conventional field sampling. This approach can effectively help assess restoration goals and guide adaptive management strategies, especially when incorporating the expertise of restoration ecologists in understanding how different vegetation types affect habitat quality.

Following harvest, monitoring reforestation success is a crucial component of sustainable management. In Alberta, Canada, like other jurisdictions, the efficiency of the current plot-based forest regeneration monitoring regime is challenged by the cost of accessibility and the declining availability of qualified field crews. Fine spatial resolution imagery and deep learning have been proposed as alternative monitoring tools and have proven successful under experimental conditions, yet how successfully models can be applied and transferred between a range of untrained sites and conditions remains unclear.

In this research, we repurposed a mask region-based convolutional neural network (Mask R–CNN) model that was previously trained to delineate coniferous tree crowns to instead segment instances of two species of regenerating conifers. We transferred learned parameters by replacing original single-class labels with photo-interpreted species information and retraining a selection of the network's parameters. We assessed the transferability of the new model by testing on five untrained sites, representing a range of forest types and densities typical of reforestation in the region. Results yielded a mean average precision (mAP) of 72% and average class F1 scores of 69% and 78% for lodgepole pine (Pinus contorta) and white spruce (Picea glauca), respectively, demonstrating successful transferability. We then investigated an additional transfer learning scenario by iteratively adding data from four of the five sites to the training set while reserving data from the remaining site for testing. On average, this improved mAP by 5%, lodgepole pine F1 by 7%, and white spruce F1 by 3%, and demonstrated that trained models can be continuously improved as sufficiently representative data becomes available.

Multi-temporal Synthetic Aperture Radar Interferometry (MT-InSAR) is the only geodetic technique allowing to measure ground deformation down to mm/yr over continuous areas. Vegetation cover in equatorial regions favors the use of L-band SAR data to improve interferometric coherence. However, the electron content of ionosphere, affecting the propagation of the SAR signal, shows particularly strong spatio-temporal variations near the equator, while the dispersive nature of the ionosphere makes its effect stronger on low-frequencies, such as L-band signals. To tackle this problem, range split-spectrum method can be implemented to compensate the ionospheric phase contribution. Here, we apply this technique for time-series of ALOS-PALSAR data, and propose optimizations for low-coherence areas. To evaluate the efficiency of this method to retrieve subtle deformation rates in equatorial regions, we compute time-series using four ALOS-PALSAR datasets in contexts of low to medium coherence, showing slow deformation rates (mm/yr to cm/yr). The processed tracks are located in Ecuador, Trinidad and Sumatra, and feature 15 to 19 acquisitions including very high, dominating ionospheric noise, corresponding to equivalent displacements of up to 2 m. The correction method performs well and allows to reduce drastically the noise level due to ionosphere, with significant improvement compared with a simple plane fitting method. This is due to frequent highly non-linear patterns of perturbation, characterizing equatorial TEC distribution. We use semivariograms to quantify the uncertainty of the corrected time-series, highlighting its dependence on spatial distance. Thus, using ALOS-PALSAR-like archive, one can expect a detection threshold on the Line-of-Sight velocity ranging between 3 and 6 mm/yr, depending on the spatial wavelength of the signal to be observed. These values are consistent with the accuracy derived from the comparison of velocities between two tracks in their overlapping area. In the case studies that we processed, the time-series corrected from ionosphere allows to retrieve accurately fault creep and volcanic signal but it is still too noisy for retrieving tiny long-wavelength signals such as slow (mm/yr) interseismic strain accumulation.

As one of the most critical features on the earth's surface, coastal zone mandates high-quality extraction of its representative feature, the coastline. Prior methodologies primarily emphasize on edge and small-scale information. However, during large-scale image processing, misclassification might occur due to the difficulty in determining whether a local area belongs to the land or sea. To address this, we propose a deep learning-based multiscale coastline extraction algorithm in this study. It comprises a multiscale coastal zone dataset built upon a tile map service structure and a scene classification-based multiscale coastal zone classifier, employing quadtree decomposition to identify coastal zones from low to high levels. Contrasting with conventional semantic segmentation, the scene classification network, owing to its larger receptive field, can accurately discern land and sea. This accuracy is further enhanced by using quadtree decomposition to process images with lower resolution and larger coverage. The results suggest that our proposed method effectively eliminates confusing features, with the overall experimental classification accuracy attesting to the effectiveness of our approach, yielding a 6% improvement. Moreover, the screening process in this study significantly reduces the number of input samples for the segmentation network, thus boosting computational speed.

Mapping and monitoring thermokarst lakes are crucial to understanding the impact of climate change on permafrost regions and quantifying permafrost-related carbon emissions. Several automatic methods based on remote sensing images have been developed for thermokarst lake mapping. However, mixed pixels containing both land and water characteristics in the lakeshore zones pose a significant challenge to the accuracy of these methods. Furthermore, few approaches were able to fully automate the identification of thermokarst lakes without the manual training sample selection or parameter tuning. In this study, we present a fully automatic framework for thermokarst lake mapping using moderate-resolution Sentinel-2 images. The proposed method combines multidimensional hierarchical clustering and sub-pixel mapping (SPM) based on the radial basis function (RBF) interpolation and Markov random field (MRF) (referred to as RBF-then-MRF SPM), so as to achieve thermokarst lake mapping at a spatial resolution of 3.3 m. We apply the proposed method to two representative thermokarst lake distribution regions in the Northern Hemisphere and achieve a mean Kappa coefficient of 0.89 and 0.99, and a mean $Quality$ of 89.86% and 96.60% on the central Tibetan Plateau and the northern Seward Peninsula, respectively. The results demonstrate that the proposed method significantly improves the accuracy of mixed pixel extraction, and the automatic thermokarst lake mapping is applicable to diverse permafrost regions.

Airborne laser scanning (ALS) data has been widely used for total aboveground tree biomass (AGB) modelling, however, there is less research focusing on estimating specific tree biomass components (wood, branches, bark, and foliage). Knowledge about these biomass components is essential for carbon accounting, understanding forest nutrient cycling, and other applications. In this study, we compare additive AGB estimation (sum of estimated components) with direct AGB estimation using deep neural network (DNN) and random forest (RF) models. We utilise two point cloud DNNs: point-based Dynamic Graph Convolutional Neural Network (DGCNN) and Octree-based Convolutional Neural Network (OCNN). DNN and RF models were trained using a dataset comprised of 2336 sample plots from a mixed temperate forest in New Brunswick, Canada. Results indicate that additive AGB models perform similarly to direct models in terms of coefficient of determination (R²) and root-mean square error (RMSE), and reduced the mean absolute percentage error (MAPE) by 22% on average. Compared to RF, the DNNs provided a small improvement in performance, with OCNN explaining 5% more variation in the data (R² = 0.76) and reducing MAPE by 20% on average. Overall, this study showcases the effectiveness of additive tree AGB models and highlights the potential of DNNs for enhanced AGB estimation. To further improve DNN performance, we recommend using larger training datasets, implementing hyperparameter optimization, and incorporating additional data such as multispectral imagery.

Sunlight is the primary source of energy in forest ecosystems and subcanopy light regimes largely determine the establishment, growth and dispersal of plants and thus forest floor plant communities. Subcanopy light regimes are highly variable in both space and time, which makes monitoring them challenging. In this study, we assess the potential of Sentinel-1 and Sentinel-2 time series for predicting subcanopy light regimes in temperate mountain forests. We trained different random forest regression models predicting field-measured total site factor (TSF, proportion of potential direct and diffuse solar radiation reaching the forest floor, here defined as the transition zone between belowground and aboveground biomass) from a set of metrics derived from Sentinel-1 and Sentinel-2 time series. Model performance was benchmarked against a model based on structural metrics derived from Airborne Laser Scanning (ALS) data, serving as an empirical gold-standard in modelling subcanopy light regimes. We found that Sentinel-1 and Sentinel-2 time series performed nearly as good as the model based on high-resolution ALS data (R²/RMSE of 0.80/0.11 for Sentinel-1/2 compared to R²/RMSE of 0.90/0.08 for ALS). We furthermore tested the generalizability of the trained models to two new sites not used for training for which field data was available for validation. Prediction accuracy for the ALS model decreased substantially for the two independent test sites due to variable ALS data quality and acquisition date (ΔR²/ΔRMSE of 0.29/0.05 and 0.11/0.03 for both independent test sites). The prediction accuracy of the Sentinel-1/2 model, however, remained more stable (ΔR²/ΔRMSE of 0.13/0.02 and 0.13/0.04). We therefore conclude that a combination of Sentinel-1 and Sentinel-2 time series has the potential to map subcanopy light conditions spatially and temporally independent of the availability of high-resolution ALS data. This has important implications for the operational monitoring of forest ecosystems across large scales, which is often limited by the challenges related to acquiring airborne datasets.

In recent years, armed conflicts are globally on the rise, causing drastic human and environmental harm. The Tigray war in Ethiopia is one of the recent violent conflicts that has abruptly reversed decades of ecosystem restoration efforts. This paper analyzes changes in woody vegetation cover during the period of armed conflict (2020–2022) using remote sensing techniques, supplemented by field testimony and secondary data. Extent of woody vegetation cover was analyzed using Normalized Difference Vegetation Index (NDVI) thresholding method from Sentinel 2 images in Google Earth Engine, and scale of de-electrification was qualitatively analyzed from Black Marble HD nighttime lights dataset, acquired from NASA's Black Marble team. The magnitude, direction as well as the mechanisms of change in woody vegetation cover varied across the region and over time. Tigray's woody vegetation cover fluctuated within 20% of the landmass. Mainly scattered to mountainous areas, the dry Afromontane forest cover declined from about 17% in 2020 to 15% in 2021, and 12% in 2022. About 17% of the overall decline was observed between 500 m and 2000 m elevation, where there is higher anthropogenic pressure. Land restoration practices meant to avert land degradation and desertification were interrupted and the area turned warfare ground. In many areas, forests were burned, the trees cut and the area became barren. The suspension of public services such as electricity for household or industrial use created heavy reliance on firewood and charcoal, further threatening to compound weather and climate. The magnitude of disturbance in a region that is already at a very high risk of desertification requires urgent national and international attention. Continued ecosystem disturbance could eventually make the domain part of a wider desert connecting the Sahel to the Afar Triangle, a scenario which may render the area uninhabitable.

Tropical secondary forests are ecosystems of critical importance for protecting biodiversity, buffering primary forest loss, and sequestering atmospheric carbon. Monitoring growth and carbon sequestration in secondary forests is difficult, with inventory plots sampling $<0.001\%$ of secondary forests. The Global Ecosystem Dynamics Investigation (GEDI), a space-borne LiDAR sampler, provides billions of aboveground carbon density (ACD) estimates across the tropics. We fuse these carbon density estimates with a time series of forest change maps to identify their age since last deforestation and thus estimate the average rate of carbon sequestration in secondary forests across the Amazon biome. To our knowledge, this is the first estimate of these rates made using the new GEDI dataset. Moreover, this paper addresses key statistical and computational challenges of GEDI data fusion and analysis. We propagate both GEDI ACD and geolocation uncertainty to the regrowth rate estimate through a Monte Carlo approach, and we handle heteroskedasticity, outliers, and spatial autocorrelation using robust statistical methods. The large size of the GEDI dataset combined with the proposed Monte Carlo bootstrap can be highly computationally intensive, with a naive implementation taking over a month to complete. Nevertheless, we demonstrate the feasibility of our method by developing optimized open-source code that performs this computation on the 151 million quality-filtered GEDI shots available for the Amazon biome from April 2019–August 2021 in under 25 min in benchmark tests. By resolving these statistical and computational challenges with an efficient open-source pipeline, we create a standard approach that can be used more broadly in any work seeking to combine the GEDI dataset with high-resolution classification maps. Using this approach, we identify approximately 23, 000 GEDI samples of regrowing forest at least 60 m × 60 m wide across the Amazon biome and estimate a carbon sequestration rate of 1.86 MgC/ha/yr with a 95% empirical confidence interval of 1.75–1.97 MgC/ha/yr, with rates varying from 1.27 to 1.99 MgC/ha/yr across smaller subregions.