Remote Sensing of Environment最新文献

Global urban mapping is vital for understanding various environmental challenges and supporting Sustainable Development Goal 11. Although deep learning models present a potential unified solution, their effectiveness is intrinsically tied to the quality and diversity of the training data, which often present limitations in existing research. To overcome these limitations, this paper introduced a data engine tailored to generate high-quality and diverse training samples at the global scale. This semi-automatic procedure operated in two stages. The initial stage focused on the generation of globally-distributed accurate samples by harmonizing existing open-source datasets. The subsequent stage broadened the sample coverage to the global scale by leveraging published global data products and OpenStreetMap data, ensuring the sample's diversity. Using the dataset generated by the data engine, we trained a Global Urban Mapper (GUM), achieving superior global testing results, outperforming the second-best product (i.e., GISA-10) by 2.89% in Overall Accuracy (OA) and 5.92% in mean Intersection over Union (mIoU). The advancements can primarily be ascribed to the superior quality and heterogeneity of the data generated by the proposed data engine, providing a precise and diverse set of samples for the deep learning model to assimilate. The proposed data engine, built exclusively on open-source data, offers promising prospects for global mapping tasks beyond urban land cover. We will release GUM and the associated preprocessing code in https://github.com/LauraChow77/GlobalUrbanMapper, which will empower users to map specific areas of interest worldwide, thereby facilitating timely urban assessment and monitoring.

ICESat-2's Advanced Topographic Laser Altimeter System (ATLAS) has emerged as a useful tool for calculating attenuation signals in natural surface waters, thus improving our understanding of particulates from open-ocean plankton to nearshore suspended terrigenous sediments. While several studies have employed methods based on Beer's Law to derive attenuation coefficients (including through a machine-learning approach), a rigorous test of specific tuning parameters and processing choices has not yet been performed. Here we present comprehensive sensitivity tests of noise removal, choice of bin sizes, surface-peak exclusion, and beam pairing across four contrasting marine environments as well as solar background removal at an additional site to quantify the impacts of these processing choices on the derived photon-based attenuation coefficient K_dph. Ultimately, calculated K_dph values were not statistically sensitive to choices of horizontal bin sizes, vertical bin sizes, and surface exclusion depths with ranges of 500–2000 m, 0.25–1.0 m, and 0.5–1.0 m, respectively. Use of strong-beam data is recommended because weak-beam data introduce additional noise, though in open-ocean waters where photon counts are sparse, it may be desirable to include weak-beam data. In a daytime/nighttime data comparison, daytime data were found to be usable, though removal of the solar background increased the K_dph estimates by ∼27–64%. A robust solution for removing afterpulses remains elusive, though a gaussian decomposition scheme was attempted. It did not, however, yield statistically different K_dph values relative to the uncorrected dataset. Detailed information about processing choices and a suggested workflow for ocean applications are provided. Together the results pave the way for expanded K_dph analyses of global datasets (including turbid coastal waters).

Mapping land cover in highly heterogeneous landscapes is challenging, and classifications have inherent limitations where the spatial resolution of remotely sensed data exceeds the size of small objects. For example, classifications based on 30-m Landsat data do not capture urban or other heterogeneous environments well. This limitation may be overcome by quantifying the subpixel fractions of different land cover types. However, the selection process and transferability of models designed for subpixel land cover mapping across biomes is yet challenging. We asked to what extent (a) locally trained models can be used for sub-pixel land cover fraction estimates in other biomes, and (b) training data from different regions can be combined into spatially generalized models to quantify fractions across global biomes. We applied machine learning regression-based fraction mapping to quantify land cover fractions of 18 regions in five biomes using Landsat data from 2022. We used spectral-temporal metrics to incorporate intra-annual temporal information and compared the performance of local, spatially transferred, and spatially generalized models. Local models performed best when applied to their respective sites (average mean absolute error, MAE, 9–18%), and also well when transferred to other sites within the same biome, but not consistently so for out-of-biome sites. However, spatially generalized models that combined input data from many sites worked very well when analyzing sites in many different biomes, and their MAE values were only slightly higher than those of the respective local models. A weighted training data selection approach, preferring training data with a lower spectral distance to the image data to be predicted, further enhanced the performance of generalized models. Our results suggest that spatially generalized regression-based fraction models can support multi-class sub-pixel fraction estimates based on medium-resolution satellite images globally. Such products would have great value for environmental monitoring in heterogeneous environments and where land cover varies along spatial or temporal gradients.

As an emerging subject of the implication on revolutionizing many fields from botany to life science, plant intelligence (PI) has been actively studied but also trapped in debate. Inspired by those earlier botanists such as Darwin conceiving this concept when observing plants outdoors, we propose to track Darwin's footprints – go again to the wild where plants show higher-fold adapting performance than in labs for arousing a re-cognition of PI. However, this plan must face a basic challenge on in-situ plant phenotyping, especially in structure, which serves as the three-dimensional (3D) phenomenological display of varying PI behaviors. Aiming at this core bottleneck, we suggest to go but with 3D remote/proximal sensing (R/PS) devices such as Light Detection and Ranging (LiDAR) – a state-of-the-art technology of fully but fine mapping plants, for starting a 3D cognition of PI. Further, to decode the mechanism of PI occurring, we preview the next-generation (e.g., hyperspectral, fluorescence, and polarization) LiDAR with the latent capacity on all-round phenotyping of plants. Their derived 3D biochemical, physiological, and biophysical functional traits can arouse a beyond-3D cognition of PI. Overall, this theoretical prospect, with the available R/PS technology traced for upgrading PI from conceptual debating to mechanistic understanding, can advance the PI field into its 3D and even beyond-3D times and bring the PI and PI-relevant sciences such as sustainability cognition to breathe new life.

Building height is an important indicator for assessing the level of urban development along the vertical dimension. Existing large-scale building height estimation studies focus on coarse spatial resolution (e.g., 10, 500, and 1000 m), which cannot reveal height variations across buildings in urban areas. High-resolution images (e.g., < 5 m resolution) can support building-scale height estimation, but they usually have small spatial coverage and are not openly accessible. In this context, we proposed a deep learning-based super-resolution method to generate building height maps at a spatial resolution of 2.5 m using Sentinel-1/2 images. The proposed method consisted of two parts: 1) a super-resolution module (SR) for learning high-resolution features; and 2) a height stratification estimation module (HS) for guiding the network to learn different height levels to mitigate the imbalanced distribution of height values. We created an open building height dataset with 45,000 samples covering multiple urban areas in the Northern Hemisphere, including China, the conterminous United States (CONUS), and Europe. Experimental results showed that for height estimation at the pixel level, the proposed method obtained a root mean square error of 10.318 m in China, 5.654 m in CONUS, and 4.113 m in Europe, respectively. Predicted results provided rich spatial details, due to the inclusion of the super-resolution module, which was heavily missed by existing large-scale studies. Moreover, we calculated the mean and standard deviation of building height in 301 urban centers, each having at least a population of 500,000, and found that the buildings in China were the highest (11.353 m ± 9.543 m), followed by CONUS (8.487 m ± 6.202 m) and Europe (8.136 m ± 5.020 m). Ablation studies indicated that the joint use of Sentinel-1/2 images and the proposed modules (SR and HS) can effectively improve the performance of building height estimation. The building dataset we generated provides great potential in high-resolution database updating, urban planning, and natural disaster assessment, and indeed, a new perspective of how we can utilize cutting-edge satellite imaging technology in urban observation, measurement, monitoring, and management. The dataset and code of this study will be available at: https://github.com/lauraset/Super-resolution-building-height-estimation.

The new-generation satellite imaging spectrometers provide an unprecedented data stream to be processed into quantifiable vegetation traits. Hybrid models have gained widespread acceptance in recent years due to their versatility in converting spectral data into traits. In hybrid models, the retrieval is obtained through a machine learning regression algorithm (MLRA) trained on a wide range of simulated data. For instance, they are currently under development for trait retrieval in preparation for the upcoming Copernicus Hyperspectral Imaging Mission for the Environment (CHIME), among others targeting routine estimation of canopy nitrogen content (CNC). However, like any retrieval algorithm, the process is not error-free, and most MLRAs inherently lack an uncertainty estimation related to the retrieved traits, which implies a risk of misinterpretation when applying the model to real-world data. Therefore, this study aimed to assess epistemic uncertainty estimation strategies (Bayesian method, drop-out, quantile regression, and bootstrapping) alongside the estimation of CNC using competitive MLRAs. Each of the regression models was evaluated using three data sets: (1) simulated scenes with varying noise using the SCOPE 2.1 radiative transfer model, (2) hyperspectral images from the PRISMA sensor, and (3) field-measured data. Analysis of generated uncertainty intervals led to the following findings: First, Gaussian processes regression (GPR) offers meaningful uncertainties, primarily attributable to spectral data degradation, which provide supplementary insights into the quality of trait mapping. Second, bootstrapping uncertainties can be used as quality indicators of the reliability of the estimates retrieved by hybrid models. Yet, its variability depends on the used MLRA, which impedes trusting its variance as a confidence interval. Third, quantile regression forest (QRF), despite not being top-performing algorithm, exhibit outstanding robustness estimations and uncertainty when the spectral data is degraded, either by Gaussian noise or by striping, often occurring in satellite imagery. Fourth, bootstrapped kernel ridge regression (KRR) demonstrated comparable performance to the benchmark algorithm GPR; the retrievals and uncertainties of these two MLRAs were highly correlated. Fifth, bootstrapped partial least squares regression (PLSR) estimations and uncertainties exhibit poor robustness to noise degradation, with normalized root mean square error (NRMSE) increasing from 19% to 112%. Additionally, a GUI tool was integrated into the ARTMO software package for assessing epistemic uncertainties from the embedded regression algorithms, providing a trait mapping quality indicator for mapping applications, and improving decision-making.