查看更多>>摘要：The unprecedented availability of optical satellite data in cloud-based computing platforms, such as Google Earth Engine (GEE), opens new possibilities to develop crop trait retrieval models from the local to the planetary scale. Hybrid retrieval models are of interest to run in these platforms as they combine the advantages of physically-based radiative transfer models (RTM) with the flexibility of machine learning regression algorithms. Previous research with GEE primarily relied on processing bottom-of-atmosphere (BOA) reflectance data, which requires atmospheric correction. In the present study, we implemented hybrid models directly into GEE for processing Sentinel-2 (S2) Level-1C (L1C) top-of-atmosphere (TOA) reflectance data into crop traits. To achieve this, a training dataset was generated using the leaf-canopy RTM PROSAIL in combination with the atmospheric model 6SV. Gaussian process regression (GPR) retrieval models were then established for eight essential crop traits namely leaf chlorophyll content, leaf water content, leaf dry matter content, fractional vegetation cover, leaf area index (LAI), and upscaled leaf variables (i.e., canopy chlorophyll content, canopy water content and canopy dry matter content). An important pre-requisite for implementation into GEE is that the models are sufficiently light in order to facilitate efficient and fast processing. Successful reduction of the training dataset by 78% was achieved using the active learning technique Euclidean distance-based diversity (EBD). With the EBD-GPR models, highly accurate validation results of LAI and upscaled leaf variables were obtained against in situ field data from the validation study site Munich-North-Isar (MNI), with normalized root mean square errors (NRMSE) from 6% to 13%. Using an independent validation dataset of similar crop types (Italian Grosseto test site), the retrieval models showed moderate to good performances for canopy-level variables, with NRMSE ranging from 14% to 50%, but failed for the leaf-level estimates. Obtained maps over the MNI site were further compared against Sentinel-2 Level 2 Prototype Processor (SL2P) vegetation estimates generated from the ESA Sentinels' Application Platform (SNAP) Biophysical Processor, proving high consistency of both retrievals (R2 from 0.80 to 0.94). Finally, thanks to the seamless GEE processing capability, the TOA-based mapping was applied over the entirety of Germany at 20 m spatial resolution including information about prediction uncer-tainty. The obtained maps provided confidence of the developed EBD-GPR retrieval models for integration in the GEE framework and national scale mapping from S2-L1C imagery. In summary, the proposed retrieval workflow demonstrates the possibility of routine processing of S2 TOA data into crop traits maps at any place on Earth as required for operational agricultural applications.
查看更多>>摘要：Solar-induced chlorophyll fluorescence (SIF) can be used as an indicator of crop photosynthetic activity and a proxy for vegetation stress in plant phenotyping and precision agriculture applications. SIF quantification is sensitive to the spectral resolution (SR), and its accurate retrieval requires sensors with sub-nanometer resolutions. However, for accurate SIF quantification from imaging sensors onboard airborne platforms, sub-nanometer imagers are costly and more difficult to operate than the commonly available narrow-band imagers (i.e., 4-to 6nm bandwidths), which can also be installed on drones and lightweight aircraft. Although a few theoretical and experimental studies have evaluated narrow-band spectra for SIF quantification, there is a lack of research focused on comparing the effects of the SR on SIF from airborne hyperspectral imagers in practical applications. This study investigates the effects of SR and sensor altitude on SIF accuracy, comparing SIF quantified at the 760nm O-2-A band (SIF760) from two hyperspectral imagers with different spectral configurations (full width at half maximum resolutions of 0.1-0.2 nm and 5.8 nm) flown in tandem on board an aircraft. SIF760 retrievals were compared from two different wheat and maize phenotyping trials grown under different nitrogen fertilizer application rates over the 2019-2021 growing seasons. SIF760 from the two sensors were correlated (R-2 = 0.77-0.9, p < 0.01), with the narrow-band imager producing larger SIF760 estimates than the sub-nanometer imager (root mean square error (RMSE) 3.28-4.69 mW/m(2)/nm/sr). Ground-level SIF760 showed strong relationships with both sub-nanometer (R-2 = 0.90, p < 0.001, RMSE = 0.07 mW/m(2)/nm/sr) and narrow-band (R-2 = 0.88, p < 0.001, RMSE = 3.26 mW/m(2)/nm/sr) airborne retrievals. Simulation-based assessments of SIF760 for SRs ranging from 1 to 5.8 nm using the SCOPE model were consistent with experimental results showing significant relationships among SIF760 quantified at different SRs. Predictive algorithms of leaf nitrogen concentration using SIF760 from either the narrow-band or sub-nanometer sensor yielded similar performance, supporting the use of narrow-band resolution imagery for assessing the spatial variability of SIF in plant phenotyping, vegetation stress detection and precision agriculture contexts.
查看更多>>摘要：We retrieved and examined the partial-column densities of carbon dioxide (CO2) in the lower (LT, typically 0-4 km) and upper (UT, typically 4-12 km) troposphere (XCO2LT and XCO2UT) collected over six global megacities: Beijing, New Delhi, New York City, Riyadh, Shanghai, and Tokyo. The radiance spectra were collected using the Thermal And Near-infrared Sensor for carbon Observation Fourier-Transform Spectrometer (TANSO-FTS) onboard the Greenhouse gases Observing SATellite (GOSAT). Our retrieval method uniquely utilizes reflected sunlight with two orthogonal components of polarization and thermal emissions. We defined megacity concentration enhancement due to surface CO2 emissions as XCO2LT minus XCO2UT, allowing us to overcome some of the challenges in the enhancement analysis using existing column density data. We examined the relationship between the XCO2LT enhancements from the time series of intensive target observations over megacities and the inverse of simulated wind speed, which could be potentially used to estimate surface emissions. Next, we attempted to estimate the average emission intensity for each city from the linear regression slope. We also compared our obtained emission estimates with the Open-Data Inventory for Anthropogenic CO2 (ODIAC) inventory for evaluation. Our results demonstrate the potential utility of the new partial-column density retrievals for estimating megacity CO2 emissions. More frequent and comprehensive coverage characterizing the spatial distribution of emissions is necessary to reduce random error and bias associated with the obtained estimate.
查看更多>>摘要：Leaf area index (LAI) is a terrestrial essential climate variable that is required in a variety of ecosystem and climate models. The Global LAnd Surface Satellite (GLASS) LAI product has been widely used, but its current version (V5) from Moderate Resolution Imaging Spectroradiometer (MODIS) data has several limitations, such as frequent temporal fluctuation, large data gaps, high dependence on the quality of surface reflectance, and low computational efficiency. To address these issues, this paper presents a deep learning model to generate a new version of the LAI product (V6) at 250-m resolution from MODIS data from 2000 onward. Unlike most existing algorithms that estimate one LAI value at one time for each pixel, this model estimates LAI for 2 years simultaneously. Three widely used LAI products (MODIS C6, GLASS V5, and PROBA-V V1) are used to generate global representative time-series LAI training samples using K-means clustering analysis and least difference criteria. We explore four machine learning models, the general regression neural network (GRNN), long short-term memory (LSTM), gated recurrent unit (GRU), and Bidirectional LSTM (Bi-LSTM), and identify Bi-LSTM as the best model for product generation. This new product is directly validated using 79 high-resolution LAI reference maps from three in situ observation networks. The results show that GLASS V6 LAI achieves higher accuracy, with a root mean square (RMSE) of 0.92 at 250 m and 0.86 at 500 m, while the RMSE is 0.98 for PROBA-V at 300 m, 1.08 for GLASS V5, and 0.95 for MODIS C6 both at 500 m. Spatial and temporal consistency analyses also demonstrate that the GLASS V6 LAI product is more spatiotemporally continuous and has higher quality in terms of presenting more realistic temporal LAI dynamics when the surface reflectance is absent for a long period owing to persistent cloud/aerosol contaminations. The results indicate that the new Bi-LSTM deep learning model runs significantly faster than the GLASS V5 algorithm, avoids the reconstruction of surface reflectance data, and is resistant to the noises (cloud and snow contamination) or missing values contained in surface reflectance than other methods, as the Bi-LSTM can effectively extract information across the entire time series of surface reflectance rather than a single time point. To our knowledge, this is the first global time-series LAI product at the 250-m spatial resolution that is freely available to the public (www.geodata.cn and www.glass.umd.edu).
查看更多>>摘要：Synthetic aperture radar (SAR) is a powerful tool for monitoring sea states in terms of the significant wave height (SWH). Regarding the specific wave mode, to date, the previous empirical models for estimating SWH from SAR data rely on single polarization. In the emerging deep learning era, few published quad-polarized SAR SWH retrieval algorithms have been based on machine learning technique, and whether quad-polarimetry improves the skill of wave height estimation remains a question. Here we propose a deep residual convolutional neural network-based SAR SWH retrieval algorithm in quad-polarization. By collocating WaveWatch III sea state hindcasts and all available archives of quad-polarized Chinese Gaofen-3 SAR imagettes in wave mode, a database with approximately 30,000 matchups was employed to establish our deeply-learned network. The GaoFen-3 significant wave height retrievals were validated against the hindcast dataset independent of training along with altimeter observations. The result of good consistency in terms of a root mean square error of 0.32 m (under sea state conditions of approximately 0.5-7.0 m) outperforms the existing Gaofen-3 wave height retrieval algorithms. Additionally, this paper introduces a discussion about the contribution of polarizations by comparing SWH derived from single-, dual-and quad-polarized deep convolutional neural networks. Single-polarized Gaofen-3 SAR data are found to be sufficient to provide accurate estimates compared to quad-polarization via a deep learning model under moderate sea conditions. Exploitation of SAR quad-polarimetry information will improve SAR wave height retrievals under high sea conditions.
Leight, C. J.McCanta, Molly C.Glotch, Timothy D.Thomson, Bradley J....
10页
查看更多>>摘要：Laboratory spectral libraries of well-characterized natural samples are necessary for accurate interpretation of remote sensing spectral data. Tephra deposits, the result of explosive volcanic eruptions, are potentially found on all differentiated terrestrial bodies and are important chronologic and compositional marker beds on Earth. Here we present a visible/near-infrared (VNIR, 0.35-2.5 mu m) and mid-infrared (MIR, 3.5-25 mu m) spectral library composed of nineteen natural tephra samples from ten volcanic sources that span a range of compositions and components. The bulk, glass, and mineral phase compositions of each sample are measured and spectra from multiple size fractions of each sample were collected. The library can be found via the Terrestrial Analog portal (DOI: https://doi.org/10.5066/P9O54M4Q).
查看更多>>摘要：Under the influence of climate change, permafrost landforms are sensitive to seasonal heave and contraction, thus exacerbating surface instability and fostering landslides as a consequence. In the pastureland of Zhimei on the Qinghai-Tibet Plateau (QTP), a typical earthflow has drawn significant attention through social media. However, detailed knowledge of the deformation characteristics, internal hydrothermal regime, and structure is still scarce. In this study, we aim to enhance traditional satellite synthetic aperture radar interferometry to divide ground deformation into the seasonal oscillation and slope deformation components and identify the magnitude and spatial distribution of unstable slopes in frozen regions. Then, the use of unmanned aerial vehicles (UAVs) was combined with geophysical monitoring techniques to recognise the deformation dynamics from the pre- to post-failure stages. Sentinel-1 images, covering almost five years, highlighted that obvious creep behaviour dominated at the pre-failure stage, while a seasonal deformation pattern characterised by a piecewise distribution associated with the hydrothermal regime was observed at the post-failure stage. Fast retrogressive erosion on the head scarps at the post-failure stage was clearly identified by multidifferential digital surface models from the UAV observations. To better understand the internal structure, both electrical resistivity tomography and ground-penetrating radar were combined to determine the seasonal frozen thickness, underlying thawing materials, and vertical cracks, which controlled the kinematic evolution from the initial creep to the narrow and long oversaturated flow that represented the terminal portion of the landslide. Finally, by comparing in situ monitoring data with field investigations, the main driving factors controlling the movement mechanism are discussed. Our results highlight the specific kinematic behaviour of an earthflow and can provide a reference for slope destabilisation on the QTP under the influence of climate change.
查看更多>>摘要：Remotely sensed evapotranspiration (ET) with high spatial and temporal resolution is frequently required to understand the regional hydrological processes, particularly in agricultural areas with complex planting structures. Most of the existing spatio-temporal fusion models lacked the fusion of ET because they ignored the physiological characteristics of the vegetation moisture condition. Therefore, we propose a classification-based spatiotemporal adaptive fusion model (CSAFM) for the evaluation of remotely sensed ET in an irrigated agricultural area with a complex planting structure. This model combines the unmixing-based and weight-based fusion approaches to produce ET maps with high spatiotemporal resolution. It uses the mainstream weight based fusion algorism-the spatial and temporal adaptive reflectance fusion model (STARFM)-in the fusion step. However, in contrast to the existing reflectance-based fusion algorithms, the CSAFM considers the effects of soil moisture and crop category on evapotranspiration rates. It replaces the unmixing window with an irregular hydrological response unit (HRU) containing homogeneous meteorological and irrigation conditions, and then unmixing the mixed pixels using a planting structure map. Moreover, an ET correction method was proposed in CSAFM to restore the spatial heterogeneity. The performance of CSAFM was compared to that of two mainstream fusion models using Landsat-ET and MODIS-ET based on the surface energy balance algorithm for land (SEBAL): the enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM) and the flexible spatiotemporal data fusion method (FSDAF). The models were validated with ground-based ET monitored by eddy covariance observed ET and Landsat inverted ET. It was found that the CSAFM model (mean MAE: 0.40 mm/day) beat the ESTARFM model (mean MAE: 0.49 mm/day) and the FSDAF (mean MAE: 0.53 mm/day) in accurately fusion ET and reproduce the details of complex surface landscapes. Additionally, CSAFM (RMSE: 0.66-1.10 mm/day) is less sensitive to the update frequency of input data in the crop growing season than ESTARFM (RMSE: 0.79-1.36 mm/day) and FSDAF (RMSE: 0.79-1.17 mm/day), indicating its suitability in areas with limited input dataset. Overall, the proposed CSAFM model can greatly improve the ET fusion accuracy in irrigated agricultural areas.
Sahaar, Shukran A.Elhaddad, AymnNiemann, Jeffrey D.
18页
查看更多>>摘要：Remote-sensing methods based on optical and thermal satellite imagery have been proposed to estimate fine-resolution (30 m) rootzone soil moisture (theta) over bar. In these methods (theta) over bar. is most commonly estimated using a single empirical relationship with evaporative fraction Lambda(SEB) or evaporative index Lambda(PET). Methods have been proposed recently to estimate these relationships based on regional climate, soil, and vegetation characteristics, but those methods have not yet been applied to estimate (theta) over bar from remote sensing data. The objective of this study is to evaluate the. estimates from remote sensing when the Lambda(SEB) vs. (theta) over bar and Lambda(PET) vs. (theta) over bar relationships are inferred from regional characteristics using the previously proposed methods. Four study regions are considered including the Walnut Gulch Experimental Watershed in Arizona, the Pinon Canyon Maneuver Site and Lower Arkansas River Valley in Colorado, the Little Washita and Fort Cobb Experimental Watersheds in Oklahoma, and the Mississippi Delta region in Mississippi. The (theta) over bar estimates from the regionally adapted relationships are compared to (theta) over bar estimates from the single empirical relationship and to in situ (theta) over bar measurements. The estimates from the regionally adapted relationships consistently outperform the estimates form the single empirical relationship. The performance typically improves as more regional information is used in the relationships and reduces the root mean squared error of (theta) over bar by an average of 45% among the four regions. The regional method typically performs better for the arid and semiarid regions with root mean squared errors of 0.05 cm(3) cm(-3) and 0.04 cm(3) cm(-3), respectively. The regionally adapted relationships better capture both spatial and temporal variations of soil moisture than the single empirical relationship.
查看更多>>摘要：Clouds play a significant role in the climate system, which affects the radiation balance and modulates the global hydrological cycle. However, the existing cloud property products have poor spatiotemporal continuity with only daytime cloud property retrieval results, which makes it challenging for us to carry out researches related to clouds at night. In this study, the effect of parallax error between the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) and Himawari-8 data is corrected based on parallax correction algorithm by referring to cloud parameters extracted from the CALIOP (Cloud-Aerosol Lidar with Orthogonal Polarization) cloud property product. Due to the different sensitivity of different channels to clouds, this study treats the CALIOP data with cloud optical depth greater than 0.2 as clouds, otherwise as clear sky. Cloud property retrieval model based on the XGBoost machine learning (ML) algorithm is developed for the advanced Himawari imager (AHI) onboard the Himawari-8 satellite. The ML model can achieve unified cloud mask, cloud top temperature (CTT), and cloud top height (CTH) retrieval for both daytime and nighttime with high spatial (0.02) and temporal (10 min) resolutions using the AHI thermal data. The retrieval results are extensively evaluated over study regions (80E ~ 135E, 18N ~ 55N) by comparisons with cloud property products of JAXA (Japan Aerospace Exploration Agency) AHI, MODIS, and CALIOP data. The ML algorithm has higher cloud detection accuracy with cloudy sky (clear sky) hit rate (CHR) and false alarm rate (FAR) of 90.79% (88.35%) and 5.74% (4.67%) compared with the JAXA AHI (CHR = 92.81% (60.64%) and FAR = 19.66% (3.61%)) and MODIS (CHR = 88.50% (72.32%) and FAR = 13.10% (6.06%)) cloud products, and the ML algorithm also performs well for thin clouds over bright surfaces. The CTH and CTT retrieved by the ML algorithm (RMSE = 1.83 km and 12.29 K) are in good agreement with the CALIOP cloud property product, and the root mean squared error (RMSE) of CTH and CTT is reduced by 33.70% (18.67%) and 35.28% (16.45%) on average, respectively, compared with the JAXA AHI (MODIS) cloud product with RMSE of 2.76 km (2.25 km) and 18.99 K (14.71 K). In addition, the ML algorithm can obtain higher cloud property retrieval accuracy at night compared with the results during the day.
NSTL
Elsevier