Coarse Spatial Resolution Research Articles

Optimizing WorldView-2, -3 cloud masking using machine learning approaches

The detection of clouds is one of the first steps in the pre-processing of remotely sensed data. At coarse spatial resolution (> 100 m), clouds are bright and generally distinguishable from other landscape surfaces. At very high-resolution (< 3 m), detecting clouds becomes a significant challenge due to the presence of smaller features, with spectral characteristics similar to other land cover types, and thin (partially transparent) cloud forms. Furthermore, at this resolution, clouds can cover many thousands of pixels, making both the center and boundaries of the clouds prone to pixel contamination and variations in the spectral intensity. Techniques that rely solely on the spectral information of clouds underperform in these situations. In this study, we propose a multi-regional and multi-sensor deep learning approach for the detection of clouds in very high-resolution WorldView satellite imagery. A modified UNet-like convolutional neural network (CNN) was used for the task of semantic segmentation in the regions of Vietnam, Senegal, and Ethiopia strictly using RGB + NIR spectral bands. In addition, we demonstrate the superiority of CNNs cloud predicted mapping accuracy of 81–91%, over traditional methods such as Random Forest algorithms of 57–88%. The best performing UNet model has an overall accuracy of 95% in all regions, while the Random Forest has an overall accuracy of 89%. We conclude with promising future research directions of the proposed methods for a global cloud cover implementation.

Future Changes of East Asian Extratropical Cyclones in the CMIP5 Models

Abstract Future changes of extratropical cyclones (ETCs) over East Asia are investigated using the models participating in phase 5 of the Coupled Model Intercomparison Project (CMIP5). To quantify ETC frequency, intensity, and genesis changes in a warming climate, the objective tracking algorithm is applied to nine CMIP5 models. These models provide 6-hourly wind data with no missing values in the high-terrain regions. The historical simulations reasonably well capture the spatial distribution of ETC properties, except for nonnegligible biases in and downstream of the high-terrain regions. Such biases are particularly pronounced in the models with a coarse spatial resolution and smooth topography, which weaken lee cyclogenesis. The best five models, which show better performance for historical simulations than other models, are used to evaluate the possible changes of East Asian ETCs under the RCP8.5 scenario. These models project a reduced cyclogenesis on the leeward side of the Tibetan Plateau, and over the East China Sea and western North Pacific in the late twenty-first century, resulting in a reduced ETC frequency from the east coast of China to the western North Pacific. The ETC intensity also shows a hint of weakening over the North Pacific. These ETC property changes are largely consistent with an enhanced static stability and a reduced vertical wind shear in a warming climate. This result indicates that the local baroclinicity, instead of increased moisture content, may play a critical role in determining the future changes of East Asian ETCs.

Data analytics for improved closest hospital suggestion for EMS operations in New York City

In addition to fires and various other emergencies, the Fire Department of New York City (FDNY) responds to medical emergencies throughout New York City. When a patient requires hospital care, the FDNY uses a hospital suggestion pattern to recommend the closest hospital that can appropriately treat the patient, given the geographic location of the incident. In a city like New York with a high density of hospitals, determining the closest hospital (in terms of travel time) is not trivial. This paper presents the results of a collaboration between the FDNY and Columbia University which integrates data analytics in the derivation, implementation and monitoring of this hospital suggestion pattern. Both telematics data of city-owned vehicles and historical ambulance travel times to hospitals are utilized to estimate the order of closest hospitals for a given incident location. Implementation constraints (coarse spatial and temporal resolutions of the pattern) and sparsity of the historical ambulance dataset generate large aleatory and epistemic uncertainties that are accounted for using a probabilistic data fusion procedure. The result of this analysis, hospital suggestion pattern N, was implemented in December 2020. A statistical analysis shows an overall decrease in travel times to hospitals as an effect of this pattern. • A data-driven methodology is used to optimize hospital transports in New York City. • The methodology leverages a road network and origin–destination ambulance data. • Uncertainties in travel times are accounted for in the data fusion procedure. • The proposed solution was successfully deployed in December 2020 in New York City.

From wetlands to wetlandscapes: Remote sensing calibration of process‐based hydrological models in heterogeneous landscapes

AbstractWetlands are vital components of landscapes that sustain a range of important ecosystem services. Understanding how wetland‐rich landscapes—or wetlandscapes—will evolve under a changing climate and increasing anthropogenic encroachment is urgent. Wetlandscapes are highly heterogeneous, and scaling local modelling insights from individual instrumented wetlands to characterize landscape‐scale dynamics has been a pervasive challenge. We investigate the potential to use water extent information from satellite imagery to calibrate landscape‐scale process‐based hydrological models. Applications to wetlandscapes in the Prairie Pothole Region (PPR) in North Dakota and the Texas Playa Lakes (TPL) shed light on two important trade‐offs. First, in‐situ monitoring provides accurate water extent information on an arbitrary subset of wetlands, whereas satellite imagery captures landscape‐scale hydrological dynamics but suffers persistent water‐detection challenges. Satellite imagery is a superior source of data for model calibration in sparsely monitored and spatially heterogeneous landscapes like the PPR, where the sampling uncertainty of monitored wetlands exceeds the water detection uncertainty of remote sensing. The two data sources are equivalent for more homogeneous landscapes like the TPL. The second tradeoff concerns the spatial resolution and temporal coverage of satellite imagery. In that regard, the 20 years of bi‐weekly images captured by Landsat 7 provides unprecedented insights into the dynamic nature of the ecohydrological characteristics of wetlandscapes, such as seasonal and inter‐annual changes of their metapopulation capacity. In the PPR, the amplitude of these dynamics far exceeds the bias introduced by Landsat's inability to capture ecologically important connectivity details due to its coarse spatial resolution compared to more recent imagery.

Interoperability of -Band Sentinel-1 SAR and GRACE Satellite Sensors on PSInSAR-Based Urban Surface Subsidence Mapping of Varanasi, India

The Sentinel-1 is an active synthetic aperture radar (SAR) satellite with a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${C}$ </tex-math></inline-formula> -band SAR sensor operating at a center frequency of 5.405 GHz and a wavelength of 5.55 cm. With the availability of freely available SAR datasets from Sentinel-1, the persistent scatterer SAR interferometry (PSInSAR)-based urban surface subsidence mapping has become easy. However, apart from geological causes, the major cause of urban surface subsidence is the over-exploitation of groundwater that results in piezometric pressure loss in the aquifers resulting in net subsidence. With the Gravity Recovery and Climate Experiment (GRACE) satellite sensor, the groundwater level fluctuations can be very easily studied temporally. But the coarse spatial resolution of GRACE data makes the study of groundwater fluctuations difficult to study for smaller watersheds. This study aims to correlate the PSInSAR average surface line of sight (LOS) displacement from Sentinel-1, with the average groundwater fluctuations from the GRACE sensor temporally from May 2017 to February 2022. The study also compared the correlation between PSInSAR displacement in both VV and VH polarizations and observed the <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> values to be 0.63 and 0.65 for VV and VH polarizations, respectively, with GRACE data. After that, using the displacement and groundwater level fluctuation data, an estimation of a gravimetric anomaly due to a decrease in groundwater level was carried out for Varanasi city. The <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> , mean absolute error (MAE), and root-mean-square error (RMSE) were observed to be 0.84, 1.23, and 2.4, respectively, in gravimetric anomaly estimation, thus giving sufficient acceptance for interoperability of the two sensors.

Classification from Sky: A Robust Remote Sensing Time Series Image Classification Using Spatial Encoder and Multi-Fast Channel Attention

The unparalleled availability of Satellite Image Time Series (SITS) for crop phenology classification unravels agricultural parcel observation and monitoring with applications of both economic and ecological importance. Moreover, the need for distinct classification of agricultural parcels into individual crop types falls on state-of-the-art deep learning models for this extrinsic task. However, most existing approaches implemented are complex and ineffective attention incorporated models, which in turn lack the resilience to recognize useful bands in achieving greater accuracy. We propose a Multi-Fast Channel Attention module for deep CNNs based on a Spatial Encoder (SE-MFCA) that requires a few parameters while enhancing the performance-complexity trade-off dilemma. Hence, we leverage on spatial encoder module to extract the images as disorderly sets of pixels to enhance the coarse spatial resolution features. We empirically show that appropriate parameter sharing in the cross channel interaction can preserve performance while significantly reducing model complexity. The proposed multi-channel attention module can efficiently be implemented via an encoder-decoder network to prevent the loss of detailed spatial information. Again, we parallelly distributed the input channel into multiple heads in our network to recover the specialized input features, which will concatenate with the residual to form a rich single feature representation. The extensive experiment has shown that our model SE-MFCA is efficient and effective compared with the previous state-of-the-art time series classification algorithm on a publicly available dataset of Sentinel-2 images for agricultural parcels. Performance-wise SE-MFCA achieves the highest overall accuracy of 94.50% and the highest mean intersection over union score of 51.92%, besides the least trainable params of 131 K and fewer floating point operations of 0.16 M.

Retrieval of High-Resolution Vegetation Optical Depth from Sentinel-1 Data over a Grassland Region in the Heihe River Basin

Vegetation optical depth (VOD), as a microwave-based estimate of vegetation water and biomass content, is increasingly used to study the impact of global climate and environmental changes on vegetation. However, current global operational VOD products have a coarse spatial resolution (~25 km), which limits their use for agriculture management and vegetation dynamics monitoring at regional scales (1–5 km). This study aims to retrieve high-resolution VOD from the C-band Sentinel-1 backscatter data over a grassland of the Heihe River Basin in northwestern China. The proposed approach used an analytical solution of a simplified Water Cloud Model (WCM), constrained by given soil moisture estimates, to invert VOD over grassland with 1 km spatial resolution during the 2018–2020 period. Our results showed that the VOD estimates exhibited large spatial variability and strong seasonal variations. Furthermore, the dynamics of VOD estimates agreed well with optical vegetation indices, i.e., the mean temporal correlations with normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and leaf area index (LAI) were 0.76, 0.75, and 0.75, respectively, suggesting that the VOD retrievals could precisely capture the dynamics of grassland.

An optimized hydrological drought index integrating GNSS displacement and satellite gravimetry data

Space geodetic techniques are rapidly emerging as powerful tools for drought monitoring, including the Global Navigation Satellite System (GNSS) positioning data, and the gravimetric data from Gravity Recovery And Climate Experiment (GRACE) and GRACE Follow-On (GRACE-FO) satellite missions. However, while GNSS-based 3D displacements have high temporal resolution, they are impacted by diverse local effects especially for tectonically active regions. Satellite gravimetry derived terrestrial water storage changes (ΔTWS) are continuous in temporal sampling, however, at much coarser spatial resolutions, and there are systematic noise and a large 11-month data gap between GRACE and GRACE-FO observations. Here for the first time, we developed an Optimized GNSS/GRACE/GRACE-FO Hydrological Drought Index (OGHDI), by optimally combining two drought indices derived from GNSS vertical displacements (VDs) and GRACE/GRACE-FO ΔTWS, to better characterize decadal evolution of droughts over Southwest China, 2011–2020. The optimal relative weights of the two drought indices are determined through maximizing the correlations with the Composite Index (CI) commonly used in the study region. Then OGHDI is compared to the self-calibrating Palmer Drought Severity Index (scPDSI). The results show that VDs agree well with ΔTWS at most of the 30 analyzed GNSS sites, with a mean correlation of −0.44. Compared to the drought index developed using only GNSS data, the optimized technique improved the OGHDI-scPDSI correlations at 90% (27 sites) of the sites, with the highest correlation reaches 0.73. Further, more significant improvement is found at the regional scale than at individual sites, and the regional mean OGHDI-CI and OGHDI-scPDSI correlations increase from 0.37 to 0.57 and from 0.42 to 0.68, respectively. All drought indices show that Southwest China experienced two sustained and excessive drought periods from 2011 to mid-2013, and after 2019. We conclude that the novel integration of two distinct contemporary and complemenatry geodetic datasets, GNSS and GRACE/GRACE-FO, improved the ability of drought monitoring in Southwest China.

Mapping Forest Aboveground Biomass with MODIS and Fengyun-3C VIRR Imageries in Yunnan Province, Southwest China Using Linear Regression, K-Nearest Neighbor and Random Forest

The aboveground biomass (AGB) of a forest is an important indicator of the forest’s terrestrial carbon storage and its relation to climate change. Due to the advantage of extensive spatial coverage and low cost, coarse-resolution remote sensing data is the main data source for wall-to-wall mapping of forest AGB at the regional scale. Despite this, improving the accuracy and efficiency of forest AGB estimation is a major challenge. In this study, two optical imageries, Moderate Resolution Imaging Spectroradiometer (MODIS) 500 m imagery and Fengyun-3C Visible and Infrared Radiometer (FY-3C VIRR) 1000 m imagery, were used and compared for forest AGB estimation in Yunnan Province, southwest China. One parametric approach, multiple linear regression (MLR), and two nonparametric approaches, k-nearest neighbor (KNN) and random forest (RF), were applied for the two imagery datasets, respectively. We evaluated the performance of the combination of remote sensing data and modeling approaches by comparing the accuracies and also explored the potential of FY-3C imagery data in forest AGB estimation at the regional scale as it was used for this purpose for the first time. We found that the machine learning models KNN and RF provided better results than MLR. From the three approaches for both MODIS and FY-3C imagery, RF performed best with R2 values of 0.84 and 0.81 and RMSE of 23.18 and 23.43, respectively. Estimation of forest AGB based on MODIS was marginally better than the estimation based on FY-3C. FY-3C imagery could therefore be an additional optical remote sensing data source of coarse spatial resolution, comparable to MODIS data which has been widely used for regional forest AGB estimation. Indices related to forest canopy moisture levels from both types of imagery were sensitive to forest AGB. The RF model and MODIS imagery were then applied to map the spatial variation of forest AGB of Yunnan Province. As a result of our study, we determined that Yunnan Province has a total forest AGB of 2123.22 Mt, with a mean value of 58.05 t/ha for forestland in 2016.

High-resolution mapping and driving factors of soil erodibility in southeastern Tibet

Quantifying the spatial distribution of soil erodibility (K factor) in the Qinghai-Tibet Plateau is essential for global soil erosion management. However, many K factor maps have a coarse spatial resolution at the regional scale, and high-resolution mapping is still a challenge. Quantitative analysis of the influence of environmental factors such as soil, topography, climate, and vegetation, and human activities on the K factor is also lacking. Therefore, we mapped the high-resolution (90 m) spatial distribution of the K factor values in southeastern Tibet using a random forest model with multiple environmental variables, based on remote sensing, ground observations (114 sampling points), and gridded datasets. The K factor estimates based on soil particle size composition and organic carbon content ranged from 0.09 to 0.35, showing moderate variation. The random forest model yielded a coefficient of determination > 0.9 and provided detailed information on the spatial distribution of K factor values, especially in large unsampled areas. The predicted K factor values tended to be high in the eastern area and low in the western area. Partial least squares path modeling showed that soil physical properties such as fractal dimension and mean weight diameter of aggregates had the largest influence on the K factor (path coefficient 0.695). Climate and topography also had a considerable influence on the K factor (path coefficient −0.489 and −0.469, respectively), while the influence of vegetation and human activities was minimal. Accordingly, the random forest model is an effective tool for high-resolution spatial distribution mapping of the K factor with limited sampling data.

Characterization and attribution of vegetation dynamics in the ecologically fragile South China Karst: Evidence from three decadal Landsat observations.

Plant growth and its changes over space and time are effective indicators for signifying ecosystem health. However, large uncertainties remain in characterizing and attributing vegetation changes in the ecologically fragile South China Karst region, since most existing studies were conducted at a coarse spatial resolution or covered limited time spans. Considering the highly fragmented landscapes in the region, this hinders their capability in detecting fine information of vegetation dynamics taking place at local scales and comprehending the influence of climate change usually over relatively long temporal ranges. Here, we explored the spatiotemporal variations in vegetation greenness for the entire South China Karst region (1.9 million km2) at a resolution of 30m for the notably increased time span (1987-2018) using three decadal Landsat images and the cloud-based Google Earth Engine. Moreover, we spatially attributed the vegetation changes and quantified the relative contribution of driving factors. Our results revealed a widespread vegetation recovery in the South China Karst (74.80%) during the past three decades. Notably, the area of vegetation recovery tripled following the implementation of ecological engineering compared with the reference period (1987-1999). Meanwhile, the vegetation restoration trend was strongly sustainable beyond 2018 as demonstrated by the Hurst exponent. Furthermore, climate change contributed only one-fifth to vegetation restoration, whereas major vegetation recovery was highly attributable to afforestation projects, implying that anthropogenic influences accelerated vegetation greenness gains in karst areas since the start of the new millennium during which ecological engineering was continually established. Our study provides additional insights into ecological restoration and conservation in the highly heterogeneous karst landscapes and other similar ecologically fragile areas worldwide.

Inferring global surface HCHO concentrations from multisource hyperspectral satellites and their application to HCHO-related global cancer burden estimation

Formaldehyde (HCHO) is a toxic and hazardous air pollutant that widely exists in atmosphere. Insufficient spatial and temporal coverage of surface HCHO measurements is limiting studies on surface HCHO-related air quality management and health risk assessment. This study develops a method to derive global ground-level HCHO concentrations from satellite-based tropospheric HCHO columns using TM5-simulated surface-to-column conversion factor with coarse spatial resolution. The method improves the factor more representative in finer grids by constraining TM5-simulated vertical profile shapes with satellite HCHO columns. The surface HCHO concentrations derived by the Ozone Mapping and Profiler Suite (OMPS) show good correlation with in situ HCHO measurements (R = 0.59) from the U.S. Environmental Protection Agency surface network. We investigated how surface HCHO relates to urbanization and population aggregation over seven regions with high HCHO pollution. The results show urban HCHO increases as a power function with population size in China, India, and West Asia. HCHO concentrations in rural aeras also present strong log–log relationship with population aggregation in China, India, the United States, and Europe. Moreover, OMPS-derived ground-level HCHO concentrations were used to estimate global cancer burden caused by long-term outdoor HCHO exposure. The results show that up to 418188 more people worldwide will develop this cancer during the human life cycle. The global cancer burden is mainly from the South-East Asia region (33.11 %) and the Western Pacific region (22.95 %). This cancer occurrence in India and China is ranked 1st and 2nd in the world due to the large population size and serious HCHO pollution. Besides, global surface HCHO concentrations and cancer burden derived from the Environmental Trace Gases Monitoring Instrument which is China’s first hyperspectral space-based spectrometer are found similar patterns with that from OMPS. Our results provide new insight into the impact of population urbanization on HCHO pollution and global outdoor HCHO-caused health risks.

Multiscale extrapolative learning algorithm for predictive soil moisture modeling & applications

We present Multiscale Extrapolative Learning Algorithm (MELA) as a novel artificial-intelligence (AI)-based data extrapolator. MELA is capable of extending temporally limited local hydroclimatic measurements at fine spatial resolution to longer periods, using remotely-sensed hydroclimatic data readily available for longer periods but at coarse spatial resolution. We demonstrate the implementation of MELA to extrapolate the monthly local soil moisture measurements at multiple depths from 2015–2021 to 1958–2021 in a semi-arid region. Such data extrapolators are imperative to generate longer historical data needed to adequately train and test AI models while enhancing the chance of capturing the effects of extreme climates on spatially variable soil moisture. The MELA-extrapolated local soil moisture subsequently allowed the construction of monthly time-series of field-scale soil moisture distributions with a normalized accuracy of 72% and prediction of countywide annual winter wheat yields – using MELA-extrapolated soil moisture data and eXplainable AI (XAI) – with a normalized accuracy of 81%. Furthermore, the XAI model ranked the predictors based on their importance in estimating winter wheat yields, in which the soil moisture near the surface and in the root zone and precipitation totals were found to be more influential than temperature on crop yields in the semi-arid region. The XAI model also unveiled the inflection points of the predictors beyond which crop yields would increase or decrease. Moreover, the AI-based analyses in conjunction with climate projections from global climate models suggest potential reductions in rainfed crop yields in the study area by 2050 and 2100 in the absence of climate-resilient mitigation and adaptation plans.

Benchmarking downscaled satellite-based soil moisture products using sparse, point-scale ground observations

While strides have been made in their accuracy and availability, the overall utility of satellite-derived surface soil moisture (SM) datasets derived from passive microwave radiometry is still reduced by their relatively coarse spatial resolution (typically >30 km). In response to this shortcoming, many independent satellite-based SM downscaling approaches have been introduced recently. However, owing to limitations in the spatial sampling characteristics of existing SM ground-monitoring networks, it has proven difficult to obtain reliable reference SM observations at the target downscaling resolution for these approaches (typically 1 to 10 km). As a result, the objective evaluation of SM downscaling approaches is often challenging and/or limited to very localized conditions. Here, we introduce and evaluate a point-scale downscaling (PSD) benchmarking strategy whereby spatially sparse, long-term, point-scale SM observations available from existing ground-based SM networks are utilized for the objective benchmarking of downscaled satellite-based SM products. First, we define criteria that must be met for a given SM downscaling strategy to add either temporal accuracy or spatial skill relative to its coarse-resolution SM baseline. Next, we illustrate, both analytically and numerically, that such criteria can be accurately evaluated using sparse, point-scale SM observations available from existing ground-based SM networks. Finally, we apply our new PSD benchmarking approach to evaluate existing fine-scale SM products. Results demonstrate that the PSD approach, in concert with existing ground-based network data, can be leveraged to robustly evaluate SM downscaling approaches.

Estimating 10-m land surface albedo from Sentinel-2 satellite observations using a direct estimation approach with Google Earth Engine

Land surface albedo plays an important role in controlling the surface energy budget and regulating the biophysical processes of natural dynamics and anthropogenic activities. Satellite remote sensing is the only practical approach to estimate surface albedo at regional and global scales. It nevertheless remains challenging for current satellites to capture fine-scale albedo variations due to their coarse spatial resolutions from tens to hundreds of meters. The emerging Sentinel-2 satellites, with a high spatial resolution of 10 m and an approximate 5-day revisiting cycle, provide a promising solution to address these observational limitations, yet their potentials remain underexplored. In this study, we integrated the Sentinel-2 observations with an updated direct estimation approach to improve the estimation and monitoring of fine-scale surface albedo. To enable the capability of the direct estimation approach at a 10-m scale, we combined the 10-m resolution European Space Agency (ESA) WorldCover land cover data and the 500-m resolution Moderate-Resolution Imaging Spectroradiometer (MODIS) Bidirectional Reflectance Distribution Function (BRDF)/albedo product to build a high-quality and representative BRDF training database. To evaluate our approach, we proposed an integrated evaluation framework leveraging 3-D physical model simulations, ground measurements, and satellite observations. Specifically, we first simulated a comprehensive dataset of Sentinel-2-like surface reflectance and broadband albedo across a variety of geometric configurations using the MODIS BRDF training samples. With this dataset, we built the Look-Up-Tables (LUTs) that connect surface broadband albedo and Sentinel-2 reflectance through a direct angular bin-based linear regression approach, and further coupled these LUTs with the Google Earth Engine (GEE) cloud-computing platform. We next evaluated the proposed algorithm at two spatial levels: (1) 10-m scale for absolute accuracy assessment using the references from the Discrete Anisotropic Radiative Transfer (DART) simulations and flux-site observations, and (2) 500-m scale for large-scale mapping assessment by comparing the estimated albedo with the MODIS albedo product. Lastly, we presented four examples to show the capability of Sentinel-2 albedo in detecting fine-scale characteristics of vegetation and urban covers. Results show that: (1) the proposed algorithm accurately estimates surface albedo from Sentinel-2-like reflectance across different landscape configurations (overall root-mean-square-error (RMSE) = 0.018, bias = 0.005, and coefficient of determination (R2) = 0.88); (2) the Sentinel-2-derived surface albedo agrees well with ground measurements (overall RMSE = 0.030, bias = -0.004, and R2 = 0.94) and MODIS products (overall RMSE = 0.030, bias = 0.021, and R2 = 0.97); and (3) Sentinel-2-derived albedo accurately captures seasonal leaf phenology and rapid snow events, and detects the interspecific (or interclass) variations of tree species and colored urban rooftops. These results demonstrate the capability of the proposed approach to map high-resolution surface albedo from Sentinel-2 satellites over large spatial and temporal contexts, suggesting the potential of using such fine-scale datasets to improve our understanding of albedo-related biophysical processes in the coupled human-environment system.

Logging Pattern Detection by Multispectral Remote Sensing Imagery in North Subtropical Plantation Forests

Forest logging detection is important for sustainable forest management. The traditional optical satellite images with visible and near-infrared bands showed the ability to identify intensive timber logging. However, less intensive logging is still difficult to detect with coarse spatial resolution such as Landsat or high spatial resolution in fewer spectral bands. Although more high-resolution remote sensing images containing richer spectral bands can be easily obtained nowadays, the questions of whether they facilitate the detection of logging patterns and which spectral bands are more effective in detecting logging patterns, especially in selective logging, remain unresolved. Therefore, this paper aims to evaluate the combinations of visible, near-infrared, red-edge, and short-wave infrared bands in detecting three different logging intensity patterns, including unlogged (control check, CK), selective logging (SL), and clear-cutting (CC), in north subtropical plantation forests with the random forest algorithm using Sentinel-2 multispectral imagery. This study aims to explore the recognition performance of different combinations of spectral bands (visual (VIS) and near-infrared bands (NIR), VIS, NIR combined with red-edge, VIS, NIR combined with short-wave infrared bands (SWIR), and full-spectrum bands combined with VIS, NIR, red edge and SWIR) and to determine the best spectral variables to be used for identifying logging patterns, especially in SL. The study was conducted in Taizishan in Hubei province, China. A total of 213 subcompartments of different logging patterns were collected and the random forest algorithm was used to classify logging patterns. The results showed that full-spectrum bands which contain the red-edge and short-wave infrared bands improve the ability of conventional optical satellites to monitor forest logging patterns and can achieve an overall accuracy of 85%, especially for SL which can achieve 79% and 64% for precision and recall accuracy, respectively. The red-edge band (698–713 nm, B5 in Sentinel-2), short-wave infrared band (2100–2280 nm, B12 in Sentinel-2), and associated vegetation indices (NBR, NDre2, and NDre1) enhance the sensitivity of the spectral information to logging patterns, especially for the SL pattern, and the precision and recall accuracy can improve by 10% and 6%, respectively. Meanwhile, both clear-cutting and unlogged patterns could be well-classified whether adding a red-edge or SWIR band or both in VIS and NIR bands; the best precision and recall accuracies for clear-cutting were enhanced to 97%, 95% and 81%, 91% for unlogged, respectively. Our results demonstrate that the optical images have the potential ability to detect logging patterns especially for the clear-cutting and unlogged patterns, and the selective logging detection accuracy can be improved by adding red-edge and short-wave infrared spectral bands.

Using GRACE to Detect Groundwater Variation in North China Plain after South-North Water Diversion.

The gravity recovery and climate experiment (GRACE) and its Follow-On mission provide a versatile tool for monitoring groundwater depletion in North China Plain (NCP). However, intermittent data gaps and inherent coarse spatial resolution have restricted the continuous detection of regional groundwater storage anomaly (GWSA) after 2014, the period of interest during the implementation of the south-to-north water diversion middle route project (SNWDP). Here, we investigated the spatiotemporal changes of GWSA in the NCP during 2004 to 2020 based on continuous downscaled GRACE data. First, we derived the continuous terrestrial water storage anomaly from six GRACE and Follow-On solutions (i.e., spherical harmonics (SH) and mass concentration [mascon] solutions). Second, we employed a long short-term memory (LSTM) model and water balance equation to downscale GWSA (i.e., 0.25° × 0.25°). Lastly, we investigated its spatiotemporal characteristics before (2004 to 2014) and after (2015 to 2020) the SNWDP operation. We show the applicability of the continuous downscaled GWSA to capture the characteristics of in situ measurements. The GWSA detects groundwater depletion at a significant (p < 0.05) rate of -17.09 ± 1.80 (SH) and -17.87 ± 1.65 (mascon) mm/a during 2004 to 2014, but a recovering trend of 7.18 ± 3.98 (SH) and 8.23 ± 4.99 (mascon) during 2015 to 2018. The subsequent groundwater extraction and precipitation reduction from 2019 to 2020, resulted in the decreasing trend of GWSA from 2015 to 2020, which is -19.11 ± 8.75 (SH) and -19.72 ± 9.08 mm/a (mascon), respectively. Spatially, the overall depletion trends become nonsignificant along the canals of SNWDP compared to the period 2004 to 2014, and groundwater recovering with trends <6 mm/a near Beijing and Tianjin are detected by the mascon solution during 2015 to 2020.

Using airborne and DESIS imaging spectroscopy to map plant diversity across the largest contiguous tract of tallgrass prairie on earth

Grassland ecosystems are under threat globally, primarily due to land-use and land-cover changes that have adversely affected their biodiversity. Given the negative ecological impacts of biodiversity loss in grasslands, there is an urgent need for developing an operational biodiversity monitoring system that functions in these ecosystems. In this paper, we assessed the capability of airborne and spaceborne imaging spectroscopy (also known as hyperspectral imaging) to capture plant α-diversity in a large naturally-assembled grassland while considering the impact of common management practices, specifically prescribed fire. We collected a robust in-situ plant diversity data set, including species composition and percent cover from 2500 sampling points with different burn ages, from recently-burned to transitional and pre-prescribed fire at the Joseph H. Williams Tallgrass Prairie Preserve in Oklahoma, USA. We expressed in-situ plant α-diversity using the first three Hill numbers, including species richness (number of observed species in a plant community), exponential Shannon entropy index (hereafter Shannon diversity; effective number of common species, where species are weighed proportional to their percent cover), and inverse Simpson concentration index (hereafter Simpson diversity; effective number of dominant species, where more weight is given to dominant species) at four different plot sizes, including 60 m × 60 m, 120 m × 120 m, 180 m × 180 m, and 240 m × 240 m. We collected full-range airborne hyperspectral data with fine spatial resolution (1 m) and visible and near-infrared spaceborne hyperspectral data from DESIS sensor with coarse spatial resolution (30 m), and used the spectral diversity hypothesis—i.e., that the variability in spectral data is largely driven by plant diversity—to estimate α-diversity remotely. In recently-burned plots and those at the transitional stage, both airborne and spaceborne data were capable of capturing Simpson diversity—a metric that calculates the effective number of dominant species by emphasizing abundant species and discounting rare species—but not species richness or Shannon diversity. Further, neither airborne nor spaceborne hyperspectral data sets were capable of capturing plant α-diversity of 60 m × 60 m or 120 m × 120 m plots. Based on these results, three main findings emerged: (1) management practices influence grassland biodiversity patterns that can be remotely detected, (2) both fine- and coarse-resolution remotely-sensed data can detect the effective number of dominant species (e.g., Simpson diversity), and (3) attention should be given to site-specific plant diversity field data collection to appropriately interpret remote sensing results. Findings of this study indicate the feasibility of estimating Simpson diversity in naturally-assembled grasslands using forthcoming spaceborne imagers such as National Aeronautics and Space Administration's Surface Biology and Geology mission.

Development of spatiotemporal high-resolution temperature data for São Paulo, Brazil; a valuable resource for epidemiologists

Background and aim Most studies investigating the temperature-mortality association in cities rely on temperature recordings from few meteorological stations or models at coarse spatial resolution, implicitly assuming an homogenous spatial distribution across. Evidence from urban heat island studies, remote sensing data, and atmospheric and physical sciences suggest this assumption does not hold. Overlooking the spatial variability of temperature exposure in cities may lead to biased epidemiological findings and limit our ability to identify areas of high risk. Herein, we aim to develop an open-access dataset of daily ambient temperature at 500 meters resolution for the municipality of São Paulo (2015-2020). Methods We obtained daily mean temperature data from 60 ground stations within the city and used spatiotemporal regression kriging to predict temperature at unmeasured locations. This technique uses multiple linear regression to model the spatiotemporal trend in temperature, and spatiotemporal kriging to model the regression residuals spatiotemporal autocorrelation. For the multilinear regression, we used several earth observation products including data relating to topography, the built environment and atmospheric processes, as explanatory covariates. The regression residuals were modelled by fitting a sum-metric spatiotemporal variogram model. We validated the model using a leave-one-out and 5-fold cross-validation. Results Of the 60 monitoring stations, 36 had data &amp;#x3e;75% valid data. Temperature records showed some spatial heterogeneity (interstation standard deviation: 1.4 ̊ C), supporting the need for spatially resolved temperature data. Of all covariates used in the multilinear regression model, remotely sensed land surface temperature was the best predictor. Any residual variability was modelled through spatiotemporal kriging. Validation using R2 and root-mean-square-error indicators is ongoing. Conclusions This dataset provides epidemiologists with a unique opportunity to investigate exposure to daily mean temperature in São Paulo at high spatio-temporal resolution, and to identify areas of high risk for certain health outcomes, e.g., mortality, over time and space.

The impact of COVID-19 on NO2 and PM2.5 levels and their associations with human mobility patterns in Singapore

ABSTRACT The decline in NO2 and PM2.5 pollutant levels were observed during COVID-19 around the world, especially during lockdowns. Previous studies explained such observed decline with the decrease in human mobility, overlooking the meteorological changes that could simultaneously mediate air pollution levels. This pitfall could potentially lead to over- or under-estimation of the effect of COVID-19 on air pollution. This study, thus, aims to re-evaluate the impact of COVID-19 on NO2 and PM2.5 pollutant levels in Singapore, by incorporating the effect of meteorological parameters in predicting NO2 and PM2.5 baseline in 2020 using machine learning methods. The results show that the mean NO2 and PM2.5 declined by 12% and 19%, which were less than the observed drops (i.e. 54% and 29%, respectively) without considering the effect of meteorological parameters. As two proxies for change in human mobility, taxi availability and carpark availability were found to increase and decrease by a maximum of 12.6% and 9.8%, respectively, in 2020 from 2019. Two correlation analyses were conducted to investigate how human mobility influenced air pollutant levels: one between daily PM2.5 and mobility changes at a regional scale and the other between weekly NO2 and mobility changes at a spatial resolution of 0.01°. The NO2 variation was found to be more associated with the change in human mobility and a cluster of stronger correlations was found in the South and East Coast of Singapore. Contrarily, PM2.5 and mobility had a weak correlation, which could be due to the limit of a coarse spatial resolution.