Atmospheric Correction Assessment and Normalization Procedure for Coupling Sentinel-2 and Worldview-3 Imagery

Sentinel-2 and WorldView-3 atmospheric correction and signal normalization based on ground-truth spectroradiometric measurements

Estimating mangrove leaf area index based on red-edge vegetation indices: A comparison among UAV, WorldView-2 and Sentinel-2 imagery

The main objective of this research is to apply unmanned aerial system (UAS) data in synergy with field spectroradiometry for the accurate radiometric correction of Landsat-8 (L8) and Sentinel-2 (S2) imagery. The central hypothesis is that imagery acquired with multispectral UAS sensors that are well calibrated with highly accurate field measurements can fill in the scale gap between satellite imagery and conventional in situ measurements; this can be possible by sampling a larger area, including difficult-to-access land covers, in less time while simultaneously providing good radiometric quality. With this aim and by using near-coincident L8 and S2 imagery, we applied an upscaling workflow, whereby: (a) UAS-acquired multispectral data was empirically fitted to the reflectance of field measurements, with an extensive set of radiometric references distributed across the spectral domain; (b) drone data was resampled to satellite grids for comparison with the radiometrically corrected L8 and S2 official products (6S-LaSRC and Sen2Cor-SNAP, respectively) and the CorRad-MiraMon algorithm using pseudo-invariant areas, such as reflectance references (PIA-MiraMon), to examine their overall accuracy; (c) then, a subset of UAS data was used as reflectance references, in combination with the CorRad-MiraMon algorithm (UAS-MiraMon), to radiometrically correct the matching bands of UAS, L8, and S2; and (d) radiometrically corrected L8 and S2 scenes obtained with UAS-MiraMon were intercompared (intersensor coherence). In the first upscaling step, the results showed a good correlation between the field spectroradiometric measurements and the drone data in all evaluated bands (R2 > 0.946). In the second upscaling step, drone data indicated good agreement (estimated from root mean square error, RMSE) with the satellite official products in visible (VIS) bands (RMSEVIS < 2.484%), but yielded poor results in the near-infrared (NIR) band (RMSENIR > 6.688% was not very good due to spectral sensor response differences). In the third step, UAS-MiraMon indicated better agreement (RMSEVIS < 2.018%) than the other satellite radiometric correction methods in visible bands (6S-LaSRC (RMSE < 2.680%), Sen2Cor-SNAP (RMSE < 2.192%), and PIA-MiraMon (RMSE < 3.130%), but did not achieve sufficient results in the NIR band (RMSENIR < 7.530%); this also occurred with all other methods. In the intercomparison step, the UAS-MiraMon method achieved an excellent intersensor (L8-S2) coherence (RMSEVIS < 1%). The UAS-sampled area involved 51 L8 (30 m) pixels, 143 S2 (20 m) pixels, and 517 S2 (10 m) pixels. The drone time needed to cover this area was only 10 min, including areas that were difficult to access. The systematic sampling of the study area was achieved with a pixel size of 6 cm, and the raster nature of the sampling allowed for an easy but rigorous resampling of UAS data to the different satellite grids. These advances improve human capacities for conventional field spectroradiometry samplings. However, our study also shows that field spectroradiometry is the backbone that supports the full upscaling workflow. In conclusion, the synergy between field spectroradiometry, UAS sensors, and Landsat-like satellite data can be a useful tool for accurate radiometric corrections used in local environmental studies or the monitoring of protected areas around the world.

While access to discharge data is key to hydrologic studies, it is a serious obstacle in ungauged basins. Currently, Sentinel-2 imagery at high spatiotemporal resolution offers a unique opportunity to infer the relation between pixel-based discharge rate and surface reflectance. One promising approach in this respect has been to find the complex relationship between river discharge and the spectral ratio between two benchmark pixels, namely the wet and dry pixels, whose dynamics resembles river discharge variation. However, this has been challenging due to the adverse impact of soil moisture and mixed land cover on the spectral behavior of the dry pixel. The selection of the wet pixel must also guarantee sufficient sensitivity of its spectral response to water depth fluctuations. To tackle the above issues, in this study, we developed a novel framework that automatizes the selection of the wet and dry pixels by using Sentinel-2 imagery. We also introduced the Normalized Difference Discharge Index (NDDI), as the best band combination, to predict river discharge. We used linear regression with leave-one-out cross-validation as the prediction model, which leverages limited satellite data due to the cloud cover. By implementing the developed framework at multiple gauged points across the continental United States, the best location of the dry pixel was consistently found in urban pixels whose longwave reflectance fall within a certain range. By analyzing the pixel-wised correlation coefficient between surface reflectance at NIR band and river discharge across the studied river widths, we found that the best wet pixels are located along river banks with shallow water depth. These pixels were characterized by the average reflectance higher than the 98th percentile in the green band. Finally, by testing over 4000 band combinations as input to the river discharge prediction model, we found that the normalized difference between B11 and B4 for the wet pixel, as well as the B11 ratioing between the dry and wet pixels yielded the most accurate predictions with R2 = 0.88 and R2 = 0.73, respectively.

The fraction of absorbed photosynthetic active radiation (FAPAR) is an essential climate variable for assessing the productivity of ecosystems. Satellite remote sensing provides spatially distributed FAPAR products, but their accurate and efficient validation is challenging in forest environments. As the FAPAR is linked to the canopy structure, it may be approximated by the fractional vegetation cover (FCOVER) under the assumption that incoming radiation is either absorbed or passed through gaps in the canopy. With FCOVER being easier to retrieve, FAPAR validation activities could benefit from a priori information on FCOVER. Spatially distributed FCOVER is available from satellite remote sensing or can be retrieved from imagery of Unmanned Aerial Vehicles (UAVs) at a centimetric resolution. We investigated remote sensing-derived FCOVER as a proxy for in situ FAPAR in a dense mixed-coniferous forest, considering both absolute values and spatiotemporal variability. Therefore, direct FAPAR measurements, acquired with a Wireless Sensor Network, were related to FCOVER derived from UAV and Sentinel-2 (S2) imagery at different seasons. The results indicated that spatially aggregated UAV-derived FCOVER was close (RMSE = 0.02) to in situ FAPAR during the peak vegetation period when the canopy was almost closed. The S2 FCOVER product underestimated both the in situ FAPAR and UAV-derived FCOVER (RMSE > 0.3), which we attributed to the generic nature of the retrieval algorithm and the coarser resolution of the product. We concluded that UAV-derived FCOVER may be used as a proxy for direct FAPAR measurements in dense canopies. As another key finding, the spatial variability of the FCOVER consistently surpassed that of the in situ FAPAR, which was also well-reflected in the S2 FAPAR and FCOVER products. We recommend integrating this experimental finding as consistency criteria in the context of ECV quality assessments. To facilitate the FAPAR sampling activities, we further suggest assessing the spatial variability of UAV-derived FCOVER to benchmark sampling sizes for in situ FAPAR measurements. Finally, our study contributes to refining the FAPAR sampling protocols needed for the validation and improvement of FAPAR estimates in forest environments.

In inland waters, Atmospheric Correction (AC), including Adjacency Effect (AE) correction, is a major challenge for water quality retrieval using optical satellite data. This study evaluated three image pre-processing options for turbidity retrieval in the St. Lawrence River using Sentinel-2 (S2) and Landsat 8/9 (L8/9) imagery with the Light Gradient Boosting Machine (LightGBM) model: (1) No pre-processing, i.e., use of Top-of-Atmosphere (TOA) reflectance, (2) AC pre-processing, obtaining water-leaving reflectance (Rw) from AC for the Operational Land Imager lite (ACOLITE)’s Dark Spectrum Fitting (DSF) technique, and (3) AE pre-processing, correcting for the AE using T-Mart before obtaining Rw from DSF. Results demonstrated that AE pre-processing outperformed the other two options. For L8/9, AE pre-processing reduced the Root Mean Square Error (RMSE) and improved the median symmetric accuracy (ε) by 48.8% and 19.0%, respectively, compared with AC pre-processing, and by 48.5% and 50.7%, respectively, compared with No pre-processing. For S2, AE pre-processing performed better than AC pre-processing and also outperformed No pre-processing, reducing RMSE by 28.4% and ε by 50.8%. However, No pre-processing yielded the lowest absolute symmetric signed percentage bias (|β|) among all pre-processing options. Analysis indicated that AE pre-processing yielded superior performance within 0–300 m from shore than other options, where the AE influence is strongest. Turbidity maps generated using AE pre-processing were smoother and less noisy compared to the other pre-processing options, particularly in cloud-adjacent regions. Overall, our findings suggest that incorporating AE correction through T-Mart improves the performance of the LightGBM model for turbidity retrieval from both L8/9 and S2 imagery in the St. Lawrence River, compared to the alternative pre-processing options.

Timely and accurate estimates of sugarcane yield provide valuable information for food management, bio-energy production, (inter)national trade, industry planning and government policy. Remote sensing and machine learning approaches can improve sugarcane yield estimation. Previous attempts have however often suffered from too few training samples due to the fact that field data collection is expensive and time-consuming. Our study demonstrates that unmanned aerial vehicle (UAV) data can be used to generate field-level yield data using only a limited number of field measurements. Plant height obtained from RGB UAV-images was used to train a model to derive intra-field yield maps based on 41 field sample plots spread over 20 sugarcane fields in the Udon Thani Province, Thailand. The yield maps were subsequently used as reference data to train another model to estimate yield from multi-spectral Sentinel-2 (S2) imagery. The integrated UAV yield and S2 data was found efficient with RMSE of 6.88 t/ha (per 10 m × 10 m pixel), for average yields of about 58 t/ha. The expansion of the sugarcane yield mapping across the entire region of 11,730 km2 was in line with the official statistical yield data and highlighted the high spatial variability of yields, both between and within fields. The presented method is a cost-effective and high-quality yield mapping approach which provides useful information for sustainable sugarcane yield management and decision-making.

High temporal resolution of synthetic aperture radar (SAR) imagery (e.g., Sentinel-1 (S1) imagery) creates new possibilities for monitoring green vegetation in urban areas and generating land-cover classification (LCC) maps. This research evaluates how different pre-processing steps of SAR imagery affect classification accuracy. Machine learning (ML) methods were applied in three different study areas: random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGB). Since the presence of the speckle noise in radar imagery is inevitable, different adaptive filters were examined. Using the backscattering values of the S1 imagery, the SVM classifier achieved a mean overall accuracy (OA) of 63.14%, and a Kappa coefficient (Kappa) of 0.50. Using the SVM classifier with a Lee filter with a window size of 5×5 (Lee5) for speckle reduction, mean values of 73.86% and 0.64 for OA and Kappa were achieved, respectively. An additional increase in the LCC was obtained with texture features calculated from a grey-level co-occurrence matrix (GLCM). The highest classification accuracy obtained for the extracted GLCM texture features using the SVM classifier, and Lee5 filter was 78.32% and 0.69 for the mean OA and Kappa values, respectively. This study improved LCC with an evaluation of various radiometric and texture features and confirmed the ability to apply an SVM classifier. For the supervised classification, the SVM method outperformed the RF and XGB methods, although the highest computational time was needed for the SVM, whereas XGB performed the fastest. These results suggest pre-processing steps of the SAR imagery for green infrastructure mapping in urban areas. Future research should address the use of multitemporal SAR data along with the pre-processing steps and ML algorithms described in this research.

Regular monitoring of inland water bodies is essential to identify the areas of deteriorating water quality and to take measures to curb the impairment. The present study aims to develop a semi-analytical model for estimating Trophic State Index (TSI) from Sentinel 2 (S2) imagery. A semi-analytical algorithm, QAA-v6, is parameterized for S2 data (referred to as QAA-S2) based on the correlation between the retrieved inherent optical properties (IOPs) from S2 imagery and the measured TSI data of Milford Lake in the USA. The accuracy of estimation of TSI by this modified model from three different lakes located in USA and from one lake in India has increased by almost 50%, when compared with that of QAA-v6. A correlation analysis of the retrieved IOPs from QAA-S2 using the outputs of three atmospheric correction processors (ACPs) (namely C2RCC, Acolite and Sen2Cor) was carried out and C2RCC gave the least mean absolute percentage error (MAPE<8%) for TSI estimation. TSI estimation using single-scattering albedo (u) at B5 and B6 bands of S2 was reasonable (MAPE<12%) to mark them as computationally efficient estimators of TSI. These results indicate the usefulness and transferability of the QAA-S2 to the various parts of the globe for estimating TSI.

The paper compares the potential of WorldView-3 (WV-3) and Sentinel-2 (S-2) satellite data for mapping naturally occurring asbestos (NOA) outcrops to be used by geologists in the planning phase of environmental monitoring. The wide distribution as well as the variety and extent of asbestos-bearing rocks make the selected area a significant case study for the evaluation of the feasibility of multispectral VNIR-SWIR (0.425-2.330 µm) remote sensing observations for NOA outcrops mapping, in those areas where the density of vegetation allows their spectral identification. Different classification procedures were used to produce NOA outcrops maps for the study area. In our study, we found in general a good agreement (k > 0.8) between the produced NOA outcrops maps and the extensive available in situ data for the accessible locations.

Soil properties variability is a factor that greatly influences cereals crops production and interacts with a proper assessment of crop nutritional status, which is fundamental to support site-specific management able to guarantee a sustainable crop production. Several management strategies of precision agriculture are now available to adjust the nitrogen (N) input to the actual crop needs. Many of the methods have been developed for proximal sensors, but increasing attention is being given to satellite-based N management systems, many of which rely on the assessment of the N status of crops. In this study, the reliability of the crop nutritional status assessment through the estimation of the nitrogen nutrition index (NNI) from Sentinel-2 (S2) satellite images was examined, focusing of the impact of soil properties variability for crop nitrogen deficiency monitoring. Vegetation indices (VIs) and biophysical variables (BVs), such as the green area index (GAI_S2), leaf chlorophyll content (Cab_S2), and canopy chlorophyll content (CCC_S2), derived from S2 imagery, were used to investigate plant N status and NNI retrieval, in the perspective of its use for guiding site-specific N fertilization. Field experiments were conducted on maize and on durum wheat, manipulating 4 groups of plots, according to soil characteristics identified by a soil map and quantified by soil samples analysis, with different N treatments. Field data collection highlighted different responses of the crops to N rate and soil type in terms of NNI, biomass (W), and nitrogen concentration (Na%). For both crops, plots in one soil class (FOR1) evidenced considerably lower values of BVs and stress conditions with respect to others soil classes even for high N rates. Soil samples analyses showed for FOR1 soil class statistically significant differences for pH, compared to the other soil classes, indicating that this property could be a limiting factor for nutrient absorption, hence crop growth, regardless of the amount of N distributed to the crop. The correlation analysis between measured crop related BVs and satellite-based products (VIs and S2_BVs) shows that it is possible to: (i) directly derive NNI from CCC_S2 (R2 = 0.76) and either normalized difference red edge index (NDRE) for maize (R2 = 0.79) or transformed chlorophyll absorption ratio index (TCARI) for durum wheat (R2 = 0.61); (ii) indirectly estimate NNI as the ratio of plant nitrogen uptake (PNUa) and critical plant nitrogen uptake (PNUc) derived using CCC_S2 (R2 = 0.77) and GAI_S2 (R2 = 0.68), respectively. Results of this study confirm that NNI is a good indicator to monitor plants N status, but also highlights the importance of linking this information to soil properties to support N site-specific fertilization in the precision agriculture framework. These findings contribute to rational agro-practices devoted to avoid N fertilization excesses and consequent environmental losses, bringing out the real limiting factors for optimal crop growth.

Electromagnetic radiance acquired by sensors is distorted mainly by atmospheric absorbing and scattering. Atmospheric correction is required for quantitatively analysis of remote sensing information. Radiation transfer model based atmospheric correction usually needs some atmospheric parameters to be chosen and estimated reasonably in advance when atmospheric observation data is lacked. In our work, a radiometric calibration was applied on the satellite data using revised coefficients at first. Then several parameters were determined for the correction process, taking into account the earth's surface and atmospheric properties of the study area. Moreover, the atmospheric correction was implemented using 6S code and the surface reflectance was retrieved. Lastly, the influence of atmospheric correction on spectral response characteristics of different land covers was discussed in respects of the spectral response curve, NDVI and the classification process, respectively. The results showed that the reflectance of all land covers decreases evidently in three visible bands, but increases in the near-infrared and shortwave infrared bands after atmospheric correction. NDVI of land covers also increases obviously after atmospheric influence was removed, and NDVI derived from the surface reflectance is the highest comparing to that from the original digital number and the top of atmosphere reflectance. The accuracy of the supervised classification is improved greatly, which is up to 87.23%, after the atmospheric effect is corrected. Methods of the parameter determination can be used for reference in similar studies.

The availability of the coarse to moderate resolution no-cost remote sensing data and advances in image processing algorithms have exponentially increased the usage of geo-spatial technology in the last few decades. The latest Sentinel-2 Multispectral Imager (MSI) provides the surface reflectance data in VNIR and SWIR ranges since 2015 at higher spatial, temporal, and spectral resolution compared to the Landsat multispectral sensors, which have been providing such data since 1970s. The symmetry in spectral bands, sensor’s spectral response, spatial and radiometric resolution of Sentinel-2 MSI with the Landsat-8 OLI (Operational Land Imager) enables their integrated and complimentary use. In this study, we have compared the surface reflectance and vegetation indices (such as NDVI and EVI) values obtained from MSI and OLI sensors in four homogeneous land use land cover (LULC) features as cropland, agriculture fallow land, dense forest and open forest. The assessment is carried out for pre- and post-monsoon seasons over the Banki sub-division region of Cuttack district, Odisha, India. For all the LULC classes, high similarity is observed in the surface reflectance values in each band except NIR, green and red band. Similarly, for both the vegetation indices derived from Landsat 8 and Sentinel-2 data, high correlation with lower RMSE is observed for all the LULC classes. The correlation (R2) for cropland varied between 0.87 and 0.96, which varied between 0.56 and 0.97 for agriculture fallow, between 0.58 and 0.9 for dense forest, and between 0.68 and 0.87 for open forest. The surface reflectance pattern obtained for different vegetated features are similar for Landsat 8 and Sentinel-2. However, significantly a higher surface reflectance is observed for Landsat 8 in NIR band followed by red and green bands, where the differences are low for blue, SWIR1 and SWIR2 bands. The comparative assessment of indices suggests a higher correlation in values between Landsat 8 and Sentinel-2 for homogeneous features compared to heterogeneous class as forest. Thus, the integrated or complementary use of Landsat 8 and Sentinel-2 for heterogeneous features may induce some biases with a limited accuracy.

The reflectance-based method of absolute radiometric calibration has been used successfully in the solar reflective portion of the spectrum. This method corrects for atmospheric effects by using a plane-parallel radiative transfer code. This code is intended to be modified for use in the atmospheric correction of data from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) due to be launched on the Earth Observing System's AM1 platform. The radiative transfer code numerically integrates the radiative transfer equation for a plane-parallel, horizontally-homogeneous atmosphere using Gauss-Seidel iteration. This code will be used to generate a look-up table of top-of-the-atmosphere (TOA) and surface radiance for several values of surface reflectance using given atmospheric optical depths. The atmospheric correction relies on the fact that surface reflectance and TOA radiance are linearly related for limited ranges of surface reflectance. Then surface reflectance and radiance are determined by linearly interpolating the TOA radiances measured by ASTER from the look-up-table results. Planned tests of this method are presented as well as anticipated accuracies. >

Description of the subject. Understanding the current situation and evolution of forests is essential for a sustainable management plan that maintains forests’ ecological and socio-economic functions. Remote sensing is a helpful tool in developing this knowledge. Objectives. This paper investigates the new opportunities offered by using Sentinel-2 (S2) imagery for forest mapping in Belgian Ardenne ecoregion. The first classification objective was to create a forest map at the regional scale. The second objective was the discrimination of 11 forest classes (Fagus sylvatica L., Betula sp., Quercus sp., other broad-leaved stands, Pseudotsuga menziesii (Mirb.) Franco, Larix sp., Pinus sylvestris L., Picea abies (L.) H.Karst., young needle-leaved stands, other needle-leaved stands, and recent clear-cuts). Method. Two S2 scenes were used and a series of spectral indices were computed for each. We applied supervised pixel-based classifications with a Random Forest classifier. The classification models were processed with a pure S2 dataset and with additional 3D data to compare obtained precisions. Results. 3D data slightly improved the precision of each objective, but the overall improvement in accuracy was only significant for objective 1. The produced forest map had an overall accuracy of 93.3%. However, the model testing tree species discrimination was also encouraging, with an overall accuracy of 88.9%. Conclusions. Because of the simple analyses done in this study, results need to be interpreted with caution. However, this paper confirms the great potential of S2 imagery, particularly SWIR and red-edge bands, which are the most important S2 bands in our study.