Surface Reflectance Products Research Articles

Evaluation of Surface Reflectance Products Based on Optimized 6S Model Using Synchronous In Situ Measurements

Surface reflectance (SR) estimation is the most essential preprocessing step for multi-sensor remote sensing inversion of geophysical parameters. Therefore, accurate and stable atmospheric correction is particularly important, which is the premise and basis of the quantitative application of remote sensing. It can also be used to directly compare different images and sensors. The Landsat-8 Operational Land Imager (OLI) and Sentinel-2 Multi-Spectral Instrument (MSI) surface reflectance products are publicly available and demonstrate high accuracy. However, there is not enough validation using synchronous spectral measurements over China’s land surface. In this study, we utilized Moderate Resolution Imaging Spectroradiometer (MODIS) atmospheric products reconstructed by Categorical Boosting (CatBoost) and 30 m ASTER Global Digital Elevation Model (ASTER GDEM) data to adjust the relevant parameters to optimize the Second Simulation of Satellite Signal in the Solar Spectrum (6S) model. The accuracy of surface reflectance products obtained from the optimized 6S model was compared with that of the original 6S model and the most commonly used Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) model. Surface reflectance products were validated and evaluated with synchronous in situ measurements from 16 sites located in five provinces of China: Fujian, Gansu, Jiangxi, Hunan, and Guangdong. Through the indirect and direct validation across two sensors and three methods, it provides evidence that the synchronous measurements have the higher and more reliable validation accuracy. The results of the validation indicated that, for Landsat-8 OLI and Sentinel-2 MSI SR products, the overall root mean square error (RMSE) calculated results of optimized 6S, original 6S and FLAASH across all spectral bands were 0.0295, 0.0378, 0.0345, and 0.0313, 0.0450, 0.0380, respectively. R2 values reached 0.9513, 0.9254, 0.9316 and 0.9377, 0.8822, 0.9122 respectively. Compared with the original 6S model and FLAASH model, the mean percent absolute error (MPAE) of the optimized 6S model was reduced by 32.20% and 15.86% for Landsat-8 OLI, respectively. On the other, for the Sentinel-2 MSI SR product, the MPAE value was reduced by 33.56% and 33.32%. For the two kinds of data, the accuracy of each band was improved to varying extents by the optimized 6S model with the auxiliary data. These findings support the hypothesis that reliable auxiliary data are helpful in reducing the influence of the atmosphere on images and restoring reality as much as is feasible.

A Comparison of Multi-Angle Implementation of Atmospheric Correction and MOD09 Daily Surface Reflectance Products From MODIS

This study presents the first systematic comparison of MAIAC Collection 6 MCD19A1 daily surface reflectance (SR) product with standard MODIS SR (MOD/MYD09). The study was limited to four tiles located in mid-Atlantic United States (H11V05), Canada (H12V03), central Amazon (H11V09), and North-Eastern China (H27V05) and used over 5000 MODIS granules in 2018. Overall, there is a remarkable agreement between the best quality pixels of the two products, in particular in the Red and NIR bands. Over selected tiles, the evaluation found that MAIAC provides from 4 to 25% more high-quality retrievals than MOD09 annually, with the largest difference in tropical regions, confirming results of the previous studies. The comparison of spectral characteristics showed a systematic MAIAC-MOD09 difference increasing from NIR to Blue, typical of biases of a Lambertian assumption in MOD09 algorithm. Over the North-Eastern China, MCD19A1 SR is found more stable at wide range of aerosol optical depth (AOD) variations, whereas MOD09 SR shows a consistent positive bias increasing with AOD and at shorter wavelengths. The observed SR differences can be attributed to differences in cloud detection, aerosol retrieval and in atmospheric correction which is performed using an accurate BRDF-coupled radiative transfer model in MAIAC and a Lambertian surface model in MOD09. While this study is not representative of the global performance because of its limited geographical coverage, it should help the land community to better understand the differences between the two products.

Optical Remote Sensing Indexes of Soil Moisture: Evaluation and Improvement Based on Aircraft Experiment Observations

Optical remote sensing (about 0.4~2.0 μm) indexes of soil moisture (SM) are valuable for some specific applications such as monitoring agricultural drought and downscaling microwave SM, due to their abundant data sources, higher spatial resolution, and easy-to-use features, etc. In this study, we evaluated thirteen typical optical SM indexes with aircraft and in situ observed SM from two field campaigns, the Soil Moisture Active Passive Validation Experiment 2012 (SMAPVEX12) and 2016 (SMAPVEX16) conducted in Manitoba, Canada. MODIS surface reflectance products (MOD09A1) and Sentinel-2 multispectral imager Level-1C data were utilized to calculate the optical SM indexes. The evaluation results demonstrated that (1) the Visible and Shortwave Infrared Drought Index (VSDI) and Optical TRApezoid Model (OPTRAM) outperform the other eleven optical SM indexes as compared with aircraft and in situ observed SM. They also presented well consistence in temporal variation with the in situ observed SM. (2) The VSDI achieved comparable performance with the OPTRAM while the former has very simple calculation expression and the latter requires complex process to determine the dry and wet boundaries. (3) Both the VSDI and OPTRAM utilize two sensitive bands of soil and vegetation moisture, i.e., Red and SWIR bands, whereas the other eleven SM indexes only employ one sensitive band. This may be the main reason of the evaluation results. (4) Based on this recognition, improvements of the VSDI and OPTRAM were created and validated in this study through adding more sensitive band to VSDI and combining NDVI and modified VSDI into a new feature space for calculating the optical SM index as with OPTRAM. The results are conducive to selecting and utilizing the current numerous optical SM indexes for SM and drought monitoring.

Contribution of Biophysical Factors to Regional Variations of Evapotranspiration and Seasonal Cooling Effects in Paddy Rice in South Korea

Previous studies have observed seasonal cooling effects in paddy rice as compared to temperate forest through enhanced evapotranspiration (ET) in Northeast Asia, while rare studies have revealed biophysical factors responsible for spatial variations of ET and its cooling effects. In this study, we adopted a data fusion method that integrated MODIS 8-day surface reflectance products, gridded daily climate data of ground surface, and a remote sensing pixel-based Penman-Monteith ET model (i.e., the RS–PM model) to quantify ET patterns of paddy rice in South Korea from 2011 to 2014. Results indicated that the regional variations of the rice-growing season ET (RGS-ET, the sum of daily ET from the season onset of rapid canopy expansion (SoS) to the end of the rice-growing season (EGS)) were primarily influenced by phenological factors (i.e., the length of growing period-LGP), followed by growing season mean climatic factors (i.e., vapor pressure deficit-VPD, and air temperature). For regional variations of the paddy field ET (PF-ET, the sum of daily ET from the field flooding and transplanting date detected by satellite observations (FFTDsat) to SoS, and to EGS), the extents were substantially reduced, only accounting for 54% of the RGS-ET variations. The FFTDsat and SoS were considered critical for the reduced PF-ET variations. In comparison to the temperate forest, changes in monthly ground surface air temperature (Ts) in paddy fields showed the V-shaped seasonal pattern with significant cooling effects found in late spring and early summer, primarily due to a large decline in daytime Ts that exceeded the nighttime warming. Bringing FFTDsat towards late spring and early summer was identified as vital field management practices, causing significant declines in daytime Ts due to enhanced ET. Results highlighted climate-warming mitigation by paddy fields due to early flooding practices.

Augmenting Landsat time series with Harmonized Landsat Sentinel-2 data products: Assessment of spectral correspondence

An increase in the temporal revisit of satellite data is often sought to increase the likelihood of obtaining cloud- and shadow-free observations as well as to improve mapping of rapidly- or seasonally-changing features. Currently, as a tandem, Landsat-7 Enhanced Thematic Mapper Plus (ETM+) and −8 Operational Land Imager (OLI) provide an acquisition opportunity on an 8-day revisit interval. Sentinel-2A and -2B MultiSpectral Instrument (MSI), with a wider swath, have a 5-day revisit interval at the equator. Due to robust pre- and post-launch cross-calibration, it has been possible for NASA to produce the Harmonized Landsat Sentinel-2 (HLS) data product from Landsat-8 OLI and Sentinel-2 MSI: L30 and S30, respectively. Knowledge of the agreement of HLS outputs (especially S30) with historic Landsat surface reflectance products will inform the ability to integrate historic time-series information with new and more frequent measures as delivered by HLS. In this research, we control for acquisition date and data source to cross-compare the HLS data (L30, S30) with established Landsat-8 OLI surface-reflectance measures as delivered by the USGS (hereafter BAP, Best Available Pixel). S30 and L30 were found to have high agreement (R = 0.87–0.96) for spectral channels and an r = 0.99 for Normalized Burn Ratio (NBR) with low relative root-mean-square difference values (1.7%–3.3%). Agreement between L30 and BAP was lower, with R values ranging from 0.85 to 0.92 for spectral channels and R = 0.94 for NBR. S30 and BAP had the lowest agreement, with R values ranging from 0.71 to 0.85 for spectral channels and r = 0.90 for NBR. Comparisons indicated a stronger agreement at latitudes above 55° N. Some dependency between spectral agreement and land cover was found, with stronger correspondence for non-vegetated cover types. The level of agreement between S30 and BAP reported herein would enable integration of HLS outputs with historic Landsat data. The resulting increased temporal frequency of data allows for improvements to current cloud screening practices and increases data density and the likelihood of temporal proximity to target date for pixel compositing approaches. Furthermore, additional within-year observations will enable change products with a higher temporal fidelity and allow for the incorporation of phenological trends into land cover classification algorithms.

High-resolution aerosol retrieval over urban areas using sentinel-2 data

Accurate retrieval of Aerosol Optical Depth (AOD) data from satellite remote sensing images over urban areas is of great importance, e.g. for air-quality monitoring. However, limited by coarse spatial resolution, available aerosol products are unable to fulfil the increasing demand of aerosol retrieval below the city scale. To address this issue, a new high-resolution (60 m) aerosol retrieval algorithm based on Sentinel-2 data over urban areas was developed. To improve the estimation of surface reflectance, two algorithms: (1) the new visible/2.19 μm surface reflectance relationships as a function of normalized difference vegetation index (NDVI) for vegetated areas; and (2) the surface reflectance database preconstructed from the operational high-quality and high-resolution Landsat 8 surface reflectance products for bright areas, were adopted depending on the surface types. For validation, the derived retrievals from Sentinel-2, along with the operational MODerate resolution Imaging Spectrometer Collection 6 (C6) aerosol products (MOD04_L2 and MCD19A2), were compared with measurements from Aerosol Robotic Network (AERONET) stations. The validation results show that the Sentinel-2 AOD retrievals agreed well with the AERONET AOD measurements, with an overall correlation coefficient of 0.907, expected error (EE) of 73.76%, mean absolute error (MAE) of 0.087, and root mean square error (RMSE) of 0.117. Comparison of the results with MOD04_L2 products (10 km) show that the Sentinel-2 AODs were superior in the situations considered, indicating that the new algorithm performed better in AOD retrieval over urban areas. Compared with the fine spatial resolution MCD19A2 products (1 km), retrievals of the proposed algorithm were comparable. The new algorithm is able to retrieve high-resolution AODs reasonably well from Sentinel-2 images over urban areas, and can provide continuous and detailed aerosol spatial distributions. Owing to the ability of the Sentinel-2 satellite to provide observations at 5-day intervals, the proposed algorithm is capable of monitoring aerosol distributions at relatively fine temporal resolution.

Errors associated with atmospheric correction methods for airborne imaging spectroscopy: Implications for vegetation indices and plant traits

Hyperspectral airborne imagery can provide rich information on plant physiological and structural properties at a scale intermediate to that of proximal and satellite remote sensing and has broad applications in assessing ecosystem function and biodiversity. A key processing step of airborne hyperspectral data is the atmospheric correction that compensates for path radiance, aerosol effects and gas absorption to derive an accurate surface reflectance that can be compared across time and space. In practice, routine correction procedures are often customized for various platforms without fully reporting or checking the errors systematically in the atmospheric correction. Such errors can have significant effects on downstream analyses such as vegetation indices or trait retrievals, and not all subsequent analyses are equally affected by the accuracy of reflectance retrievals. In this study, we examined the errors in three types of atmospheric correction methods including a radiative transfer model (RTM), empirical line correction (ELC) and a hybrid method that combines elements of the two via Bayesian inference. Our results revealed that the individual correction methods had different effects on the reflectance retrievals that impacted downstream measurements. Including spectral measurements from ground vegetation targets in addition to painted calibration targets improved the performance of the ELC method. The hybrid method yielded reflectance spectra that most closely matched the spectra of the ground validation data. The errors in vegetation indices differed with the methods, and certain indices (such as PRI) were more affected than indices that rely on stable, broader spectral features (e.g., NDVI). Plant pigment retrievals via partial least squares regression were less sensitive to errors in atmospheric correction. These findings demonstrate that obtaining high-quality, field spectral measurements over well-characterized calibration targets and representative land cover types within the scene is critical for accurate surface reflectance and subsequent downstream products, such as vegetation indices or plant traits.

Fiducial Reference Measurements for Vegetation Bio-Geophysical Variables: An End-to-End Uncertainty Evaluation Framework

With a wide range of satellite-derived vegetation bio-geophysical products now available to users, validation efforts are required to assess their accuracy and fitness for purpose. Substantial progress in the validation of such products has been made over the last two decades, but quantification of the uncertainties associated with in situ reference measurements is rarely performed, and the incorporation of uncertainties within upscaling procedures is cursory at best. Since current validation practices assume that reference data represent the truth, our ability to reliably demonstrate compliance with product uncertainty requirements through conformity testing is limited. The Fiducial Reference Measurements for Vegetation (FRM4VEG) project, initiated by the European Space Agency, is aiming to address this challenge by applying metrological principles to vegetation and surface reflectance product validation. Following FRM principles, and in accordance with the International Standards Organisation’s (ISO) Guide to the Expression of Uncertainty in Measurement (GUM), for the first time, we describe an end-to-end uncertainty evaluation framework for reference data of two key vegetation bio-geophysical variables: the fraction of absorbed photosynthetically active radiation (FAPAR) and canopy chlorophyll content (CCC). The process involves quantifying the uncertainties associated with individual in situ reference measurements and incorporating these uncertainties within the upscaling procedure (as well as those associated with the high-spatial-resolution imagery used for upscaling). The framework was demonstrated in two field campaigns covering agricultural crops (Las Tiesas–Barrax, Spain) and deciduous broadleaf forest (Wytham Woods, UK). Providing high-spatial-resolution reference maps with per-pixel uncertainty estimates, the framework is applicable to a range of other bio-geophysical variables including leaf area index (LAI), the fraction of vegetation cover (FCOVER), and canopy water content (CWC). The proposed procedures will facilitate conformity testing of moderate spatial resolution vegetation bio-geophysical products in future validation exercises.

Mangrove Forest Cover and Phenology with Landsat Dense Time Series in Central Queensland, Australia

Wetlands are one of the most biologically productive ecosystems. Wetland ecosystem services, ranging from provision of food security to climate change mitigation, are enormous, far outweighing those of dryland ecosystems per hectare. However, land use change and water regulation infrastructure have reduced connectivity in many river systems and with floodplain and estuarine wetlands. Mangrove forests are critical communities for carbon uptake and storage, pollution control and detoxification, and regulation of natural hazards. Although the clearing of mangroves in Australia is strictly regulated, Great Barrier Reef catchments have suffered landscape modifications and hydrological alterations that can kill mangroves. We used remote sensing datasets to investigate land cover change and both intra- and inter-annual seasonality in mangrove forests in a large estuarine region of Central Queensland, Australia, which encompasses a national park and Ramsar Wetland, and is adjacent to the Great Barrier Reef World Heritage site. We built a time series using spectral, auxiliary, and phenology variables with Landsat surface reflectance products, accessed in Google Earth Engine. Two land cover classes were generated (mangrove versus non-mangrove) in a Random Forest classification. Mangroves decreased by 1480 hectares (−2.31%) from 2009 to 2019. The overall classification accuracies and Kappa coefficient for 2008–2010 and 2018–2020 land cover maps were 95% and 95%, respectively. Using an NDVI-based time series we examined intra- and inter-annual seasonality with linear and harmonic regression models, and second with TIMESAT metrics of mangrove forests in three sections of our study region. Our findings suggest a relationship between mangrove growth phenology along with precipitation anomalies and severe tropical cyclone occurrence over the time series. The detection of responses to extreme events is important to improve understanding of the connections between climate, extreme weather events, and biodiversity in estuarine and mangrove ecosystems.

The superiority of the normalized difference phenology index (NDPI) for estimating grassland aboveground fresh biomass

Accurate monitoring of grassland aboveground fresh biomass (called AGB in the study) and its spatial-temporal dynamics is indispensable for sustainable grassland management. The most common method used in estimating AGB with remotely sensed data is based on the relationship between field AGB measurements and vegetation indices (VIs); however, the existing VIs do not deliver adequate results due to the soil background and spatial, temporal and sampling size variability. In this study, the AGB estimation model with the normalized difference phenology index (NDPI) was evaluated in terms of model robustness and spatial and temporal scalability based on comparisons with the widely used ratio vegetation index (RVI), difference vegetation index (DVI), normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), soil-adjusted vegetation index (SAVI), modified soil-adjusted vegetation index (MSAVI), and optimized soil-adjusted vegetation index (OSAVI). The field measurements of AGB of the natural grassland in Inner Mongolia, China, collected in 2013, 2016, and 2017 and Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance products were used for analysis. The results based on training and independent validation data showed the following: (1) The R2 value between AGB and the NDPI was the highest (0.73) among all VIs, followed by soil-line-adjusted VIs, while the R2 values of the RVI and DVI were the lowest; (2) The NDPI-based model had the best robustness for different sampling sizes; (3) The NDPI-based model also had superior spatial and temporal scalability. The results from simulation experiments using the PROSAIL model also support the superiority of the NDPI in estimating AGB. The simulation analysis further reveals that the overall superiority of the NDPI originates from the fact that the NDPI overcomes the adverse impacts of the heterogeneity of the soil background and accounts for changes in the leaf water content that contribute substantially to AGB in grassland. These findings suggest that the NDPI-based AGB estimation model is advantageous for monitoring AGB in large grasslands with significant spatial-temporal heterogeneity.

SENTINEL-2 SURFACE REFLECTANCE PRODUCTS GENERATED BY CNES AND DLR: METHODS, VALIDATION AND APPLICATIONS

Abstract. To allow for a robust and automatic exploitation of Sentinel-2 data, Analysis Ready Data (ARD) products are requested by most users. The processors of ARD products take care of the common burdens necessary for most applications, that include precise orthorectification, cloud detection and atmospheric correction steps, as well as the generation of periodic syntheses of cloud free surface reflectances. The French Theia land data center, and the German Earth Observation Center (EOC) started delivering Sentinel-2 surface reflectance products to users in 2016 in France and 2019 in Germany respectively. Both centers produce and distribute these data sets in near real time, over large regions requested by French users such as Western Europe, Maghreb, Sahel, Madagascar… Theia’s and EOC products include an instantaneous surface reflectance product (Level-2A), and a monthly cloud free synthesis of surface reflectance (Level-3A). This article shortly describes the methods used to generate the Level-2A products with the MAJA processor, and the Level-3A products with theWASP processor. The MAJA processor is based on multi-temporal methods, that use the slow variation of surface reflectance to detect clouds and estimate aerosol depth, while WASP, thanks to the quality of MAJA cloud mask, calculates a weighted average of all the cloud free observations over 45 days, every month. The article also provides validation results for Level-2A and Level-3A products, resulting from comparison with in-situ data and with other methods. A last section gives first insights from the monitoring of user uptake of the distributed products.

EVALUATING THE IMPACT OF LASRC AND SEN2COR ATMOSPHERIC CORRECTION ALGORITHMS ON LANDSAT-8/OLI AND SENTINEL-2/MSI DATA OVER AERONET STATIONS IN BRAZILIAN TERRITORY

Abstract. Accurate and consistent Surface Reflectance estimation from optical remote sensor observations is directly dependant on the used atmospheric correction processor and the differences caused by it may have implications on further processes, e.g. classification. Brazil is a continental scale country with different biomes. Recently, new initiatives, as the Brazil Data Cube Project, are emerging and using free and open data policy data, more specifically medium spatial resolution sensor images, to create image data cubes and classify the Brazilian territory crops. For this reason, the purpose of this study is to verify, on Landsat-8 and Sentinel-2 images for the Brazilian territory, the suitability of the atmospheric correction processors maintained by their image providers, LaSRC from USGS and Sen2cor from ESA, respectively. To achieve this, we tested the surface reflectance products from Landsat-8 processed through LaSRC and Sentinel-2 processed through LaSRC and Sen2cor comparing to a reference dataset computed by ARCSI and AERONET. The obtained results point that Landsat-8/OLI images atmospherically corrected using the LaSRC corrector are consistent to the surface reflectance reference and other atmospheric correction processors studies, while for Sentinel-2/MSI images, Sen2cor performed best. Although corrections over Sentinel-2/MSI data weren’t as consistent as in Landsat-8/OLI corrections, in comparison to the surface reflectance references, most of the spectral bands achieved acceptable APU results.

Global‐Scale Consistency of Spaceborne Vegetation Indices, Chlorophyll Fluorescence, and Photosynthesis

AbstractThe new TROPOspheric Monitoring Instrument (TROPOMI) solar‐induced chlorophyll fluorescence (SIF) data provides new opportunities to corroborate and improve global photosynthesis estimates. Here we report the spatiotemporal consistency between TROPOMI SIF and vegetation indices from the bidirectional reflectance distribution function (BRDF) adjusted (MCD43) and standard MODIS (MOD09) surface reflectance products, estimates of absorbed photosynthetically active radiation by chlorophyll (APARchl) derived from National Centers for Environmental Prediction Reanalysis‐2 (NCEP2), MODIS MCD18, and European Reanalysis (ERA5) data, and two GPP products (GPPVPM and GPPMOD17). We find (a) non‐adjusted VIs were more highly correlated with SIF at mid and high latitude than BRDF‐adjusted VIs, but were less correlated in the tropics, (b) negligible differences in the correlation between SIF and non‐adjusted NIRv and EVI, but BRDF‐adjusted NIRv had higher correlations with SIF at mid to high latitude than BRDF‐adjusted EVI but lower correlations in the tropics, (c) choice of PAR data set likely to cause substantial differences in global and regional GPP estimates and the correlation between modeled GPP and SIF, (d) SIF was more highly correlated with APARchl at high to mid latitude than EVI but more highly correlated with EVI at lower latitudes, and (e) GPPVPM is more highly correlated with SIF than GPPMOD17, except in sub‐Sahara Africa. Our results highlight that spaceborne photosynthesis would likely be improved by using a non‐linear response to PAR and that the fundamental differences between the vegetation indices and PAR data sets are likely to yield important differences in global and regional estimates of photosynthesis.

MOSEV: a global burn severity database from MODIS (2000–2020)

Abstract. To make advances in the fire discipline, as well as in the study of CO2 emissions, it is of great interest to develop a global database with estimators of the degree of biomass consumed by fire, which is defined as burn severity. In this work we present the first global burn severity database (MOSEV database), which is based on Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance and burned area (BA) products from November 2000 to near real time. To build the database we combined Terra MOD09A1 and Aqua MYD09A1 surface reflectance products to obtain dense time series of the normalized burn ratio (NBR) spectral index, and we used the MCD64A1 product to identify BA and the date of burning. Then, we calculated for each burned pixel the difference of the NBR (dNBR) and its relativized version (RdNBR), as well as the post-burn NBR, which are the most commonly used burn severity spectral indices. The database also includes the pre-burn NBR used for calculations, the date of the pre- and post-burn NBR, and the date of burning. Moreover, in this work we have compared the burn severity metrics included in MOSEV (dNBR, RdNBR and post-burn NBR) with the same ones obtained from Landsat-8 scenes which have an original resolution of 30 m. We calculated the Pearson's correlation coefficients and the significance of the relationships using 13 pairs of Landsat scenes randomly distributed across the globe, with a total BA of 6904 km2 (n=32 163). Results showed that MOSEV and Landsat-8 burn severity indices are highly correlated, particularly the post-burn NBR (R=0.88; P<0.001), and dNBR (R=0.74; P<0.001) showed stronger relationships than RdNBR (R=0.42; P<0.001). Differences between MOSEV and Landsat-8 indices are attributable to variability in reflectance values and to the different temporal resolution of both satellites (MODIS: 1–2 d; Landsat: 16 d). The database is structured according to the MODIS tiling system and is freely downloadable at https://doi.org/10.5281/zenodo.4265209 (Alonso-González and Fernández-García, 2020).

Estimation of pixel-level seismic vulnerability of the building environment based on mid-resolution optical remote sensing images

Building damage after seismic disasters is one of the most vital factors threatening people's lives. The seismic vulnerability of buildings over large scales at the regional or country level is a key parameter for the mitigation of seismic disaster risk and rapid assessment of casualties after seismic events. To acquire the seismic vulnerability of buildings over large areas, a machine learning method based on mid-resolution satellite optical images is proposed. Taking field-investigated building vulnerability at the satellite pixel scale as a reference, the 15 most correlated features are calculated based on VIIRS nighttime light, MODIS vegetation index and surface reflectance, and texture data from Landsat-8 OLI surface reflectance products. Taking Yancheng, Jiangsu Province, China, as the study area, where 401 pixel-level seismic vulnerabilities (PLSVs) of the building environment are acquired based on field investigations, support vector regression (SVR) and random forest (RF) models are proposed using the 15 features calculated from satellite optical images. The results show that the proposed method can be used to estimate the PLSV with a root mean square error of approximately 0.1, with the PLSV normalized between 0 and 1. The machine learning model proposed in this study has a better accuracy for PLSV estimation than spatial interpolation.

Remote Sensing of Lake Water Clarity: Performance and Transferability of Both Historical Algorithms and Machine Learning

There has been little rigorous investigation of the transferability of existing empirical water clarity models developed at one location or time to other lakes and dates of imagery with differing conditions. Machine learning methods have not been widely adopted for analysis of lake optical properties such as water clarity, despite their successful use in many other applications of environmental remote sensing. This study compares model performance for a random forest (RF) machine learning algorithm and a simple 4-band linear model with 13 previously published empirical non-machine learning algorithms. We use Landsat surface reflectance product data aligned with spatially and temporally co-located in situ Secchi depth observations from northeastern USA lakes over a 34-year period in this analysis. To evaluate the transferability of models across space and time, we compare model fit using the complete dataset (all images and samples) to a single-date approach, in which separate models are developed for each date of Landsat imagery with more than 75 field samples. On average, the single-date models for all algorithms had lower mean absolute errors (MAE) and root mean squared errors (RMSE) than the models fit to the complete dataset. The RF model had the highest pseudo-R2 for the single-date approach as well as the complete dataset, suggesting that an RF approach outperforms traditional linear regression-based algorithms when modeling lake water clarity using satellite imagery.

Using the Negative Soil Adjustment Factor of Soil Adjusted Vegetation Index (SAVI) to Resist Saturation Effects and Estimate Leaf Area Index (LAI) in Dense Vegetation Areas.

Saturation effects limit the application of vegetation indices (VIs) in dense vegetation areas. The possibility to mitigate them by adopting a negative soil adjustment factor X is addressed. Two leaf area index (LAI) data sets are analyzed using the Google Earth Engine (GEE) for validation. The first one is derived from observations of MODerate resolution Imaging Spectroradiometer (MODIS) from 16 April 2013, to 21 October 2020, in the Apiacás area. Its corresponding VIs are calculated from a combination of Sentinel-2 and Landsat-8 surface reflectance products. The second one is a global LAI dataset with VIs calculated from Landsat-5 surface reflectance products. A linear regression model is applied to both datasets to evaluate four VIs that are commonly used to estimate LAI: normalized difference vegetation index (NDVI), soil adjusted vegetation index (SAVI), transformed SAVI (TSAVI), and enhanced vegetation index (EVI). The optimal soil adjustment factor of SAVI for LAI estimation is determined using an exhaustive search. The Dickey-Fuller test indicates that the time series of LAI data are stable with a confidence level of 99%. The linear regression results stress significant saturation effects in all VIs. Finally, the exhaustive searching results show that a negative soil adjustment factor of SAVI can mitigate the SAVIs’ saturation in the Apiacás area (i.e., X = −0.148 for mean LAI = 5.35), and more generally in areas with large LAI values (e.g., X = −0.183 for mean LAI = 6.72). Our study further confirms that the lower boundary of the soil adjustment factor can be negative and that using a negative soil adjustment factor improves the computation of time series of LAI.

A data-driven approach to estimate leaf area index for Landsat images over the contiguous US

Leaf Area Index (LAI) is a fundamental vegetation biophysical variable serving as an essential input to many land surface and atmospheric models. Long-term LAI maps are typically generated with satellite images at moderate spatial resolution (0.25 to 1 km), such as those from the Moderate Resolution Imaging Spectroradiometer (MODIS). While useful for regional-scale land surface modeling, these moderate resolution products often cannot resolve spatial heterogeneity important for many agricultural and hydrological applications. This paper proposes an approach to map LAI at 30-m resolution based on Landsat images for the Contiguous US (CONUS) consistent with the MODIS product, aimed at multi-scale modeling applications. The algorithm was driven by 1.6 million spatially homogeneous samples derived from MODIS LAI and Landsat surface reflectance products from 2006 to 2018. Based on these samples, we trained separate random forest models to estimate LAI from Landsat surface reflectance for eight biomes of the National Land Cover Database (NLCD). A balanced sample design regarding the saturation status of MODIS LAI and a machine-learning-based noise detection technique were introduced to mitigate the trade-off in estimation accuracy between medium LAI (e.g., 3 to 4, unsaturated) and high LAI (e.g., 4–6, saturated).This approach was evaluated using ground measurements from 19 National Ecological Observatory Network (NEON) sites and eight independent sites from other sources. These sites comprise a representative sample of forests, grasslands, shrublands, and croplands across the US. For NEON sites, the LAI estimates show an overall Root Mean Squared Error (RMSE) of 0.8 with r2 of 0.88. For the eight independent sites, the Landsat LAI algorithm achieves RMSE between 0.52 and 0.91. The uncertainty in Landsat estimated LAI varies across biomes and locations. The proposed algorithm was implemented on the Google Earth Engine platform, allowing for the rapid generation of long-term high-resolution LAI records from the 1980s using Landsat images (code is available at https://github.com/yanghuikang/Landsat-LAI). Our findings also highlight the importance of sample balance on regression-based modeling in remote sensing applications.

MODIS-Based AVHRR Cloud and Snow Separation Algorithm

The long-term data record (LTDR) has the goal of developing a quality and consistent Advanced Very High Resolution Radiometer (AVHRR) surface reflectance and albedo products dating back to 1982 at 0.05° spatial resolution. Distinguishing between cloud and snow is of critical importance when analyzing global albedo trends, for they influence the Earth’s energy balance. However, this task is specially challenging when working with AVHRR given its limited spectral bands. Therefore, the current version of the LTDR does not distinguish between snow and clouds. To this end, we propose the Moderate Resolution Imaging Spectroradiometer (MODIS)-based AVHRR Class Separation Algorithm (MACSSA), whose goal is to identify clear land and snow pixels using AVHRR data. We make use of a combination of optical and thermal information from satellite and reanalysis data, along with monthly climatology information. These are used as inputs for two different support vector machine (SVM) models, which are then applied to AVHRR data to retrieve the MACSSA predicted tags. These are compared first against reference tags retrieved from the MYD10C1 product over pixels with less than 2-min overpass time difference between MODIS Aqua and NOAA16–19, distributed all around the world, and second against the Climate Change Initiative Cloud (Cloud_cci AVHRR) project. We found the product to be highly accurate in identifying clear land pixels, with a probability of detection of clear pixels (PODclear) of 97%. The discrimination of snow and clouds shows a PODsnow of 89%, which is encouraging given the spectral limitations of the AVHRR sensor.

Prediction of grape yields from time-series vegetation indices using satellite remote sensing and a machine-learning approach

Grape yield prediction is an important tool used by growers to optimize vineyard management and obtain better income. In this regard, the growth stages of grapes play important roles in the evaluation of vineyard production throughout the season. Predicted yield maps allow growers to view spatial variability across fields and to determine the best harvesting time and marketing strategy. Different methods are used to estimate yield; however, large-scale estimation is difficult due to labor and time requirements. Machine learning and satellite remote sensing have the potential to obtain quick and rapid assessments over large areas with lower costs and in shorter timeframes. In this context, the purpose of this research was to develop yield prediction models based on a machine-learning approach using satellite-based time-series images. In this study, Landsat 8 surface reflectance products from 2017, 2018 and 2019 were used to map the satellite-based normalized difference vegetation index (NDVI), leaf area index (LAI) and normalized difference water index (NDWI). Moreover, different growth stages were observed using moving averages and exponential smoothing methods based on the per-pixel values from the satellite imagery. The vegetation indices had particularly close relationships with each other at the time of maximum canopy expansion. To remove the seasonality of the vegetation indices, a moving average was applied to determine one representative mean for each vineyard. The generated models were validated using regression analysis and an artificial neural network (ANN) approach. The results indicated that of all the vegetation indices, NDVI had the highest accuracy (r2 = 0.79) in 2017 and 2019; however, the LAI accuracy was higher than the accuracies of the other indices (r2 = 0.79) in 2019. Nevertheless, the artificial neural network-based machine learning results indicated that NDVI had the highest accuracy in 2017 (R = 0.94), 2018 (R = 0.95) and 2019 (R = 0.92) among all the vegetation indices. Thus, machine learning achieved reliable grape yield monitoring across the studied years at local and regional levels. Ground reference yield datasets were used for comparison with the predicted yields. The findings suggest that vegetation indices can be used for calculating site-specific management of vineyards and for predicting yields. The machine-learning methods applied with satellite time-series images can achieve reliable table grape yield prediction models. These integrated models could be used for logistics and decision-making regarding table grape production.