Multispectral Instrument Research Articles

Modeling and Mapping of Aboveground Biomass and Carbon Stock Using Sentinel-2 Imagery in Chure Region, Nepal

The concerns about climate change in recent decades have heightened the need for effective methods for assessing and reporting forest biomass and Carbon Stocks (CS) at local, national, continental, and global scales. Accurate assessment of Aboveground Biomass (AGB) is critical for the sustainable management of forests, especially in the Chure region, a fragile and young mountainous in the lesser Himalaya of Nepal. This paper presents the modeling and mapping approach and shows how medium-resolution Sentinel-2 multispectral instrument (MSI) data can be used instead of hyperspectral data in inaccessible areas of the Chure region. The data were collected and analyzed from 72 circular sample plots. 60% (43 random sample plots) were used to create the model, while the remaining 40% (29 plots) were used for model validation. This study involved calculating 12 different vegetation indices and correlating them with plot-level AGB. Five models, including linear, logarithmic, quadratic, power, and exponential, were created, but the best model was found to be the quadratic model using normalized difference vegetation indices (NDVIs) with an R2 value of 0.777 and a correlation coefficient of 0.881. The model’s AIC and BIC values were 313.60 and 320.65, respectively. The validity of the model was performed using observed and predicted AGB values, resulting in an r value of 0.9128, an R2 value of 0.8332, and an RMSE value of 10.7657 t·h−1. Finally, the developed regression equation was used to map AGB in the study area. The AGB per pixel ranges from 0 to 129.18 t·h−1, whereas the amount of CS ranges from 0 to 61.01 t·h−1. Among the different vegetation indices used in the study, NDVI was found to be more precise in estimating and mapping biomass and carbon stocks in this study. Therefore, the study recommends using the quadratic model of NDVI for accurate estimation of AGB and CS in the Chure region of Sainamaina municipality.

Local Climate Zone Classification by Seasonal and Diurnal Satellite Observations: An Integration of Daytime Thermal Infrared Multispectral Imageries and High-Resolution Night-Time Light Data

Accurate, rapid, and automatic local climate zone (LCZ) mapping is essential for urban climatology and studies in terms of urban heat islands. Remotely sensed imageries incorporated with machine learning algorithms are widely utilized in LCZ labeling. Nevertheless, large-scale LCZ mapping is still challenging due to the complex vertical structure of underlying urban surfaces. This study proposed a new method of LCZ labeling that uses a random forest classifier and multi-source remotely sensed data, including Sentinel 1A Synthetic Aperture Radar (SAR), Sentinel 2 Multispectral Instrument, and Luojia1-01 night-time light data. In particular, leaf-on and -off imageries and surface thermal dynamics were utilized to enhance LCZ labeling. Additionally, we systematically evaluated how daytime and night-time features influence the performance of the classification procedure. Upon examination, the results for Beijing, China, were confirmed to be robust and refined; the Overall Accuracy (OA) value of the proposed method was 88.86%. The accuracy of LCZs 1–9 was considerably increased when using the land surface temperature feature. Among these, the Producer Accuracy (PA) value of LCZ 3 (compact low-rise) significantly increased by 16.10%. Notably, it was found that NTL largely contributed to the classification concerning LCZ 3 (compact low-rise) and LCZ A/B (dense trees). The performance of integrating leaf-on and -off imageries for LCZ labeling was better than merely uses of leaf-on or -off imageries (the OA value increased by 4.75% compared with the single use of leaf-on imagery and by 3.62% with that of leaf-off imagery). Future studies that use social media big data and Very-High-Resolution imageries are required for LCZ mapping. This study shows that combining multispectral, SAR, and night-time light data can improve the performance of the random forest classifier in general, as these data sources capture significant information about surface roughness, surface thermal feature, and night-time features. Moreover, it is found that incorporating both leaf-on and leaf-off remotely sensed imageries can improve LCZ mapping.

Vicarious Radiometric Calibration of the Multispectral Imager Onboard SDGSAT-1 over the Dunhuang Calibration Site, China

The multispectral imager (MII), onboard the Sustainable Development Science Satellite 1 (SDGSAT-1), performs detailed terrestrial change detection and coastal monitoring. SDGSAT-1 was launched at 2:19 UTC on 5 November 2021, as the world’s first Earth science satellite to serve the United Nations 2030 Sustainable Development Agenda. A vicarious radiometric calibration experiment was conducted at the Dunhuang calibration site (Gobi Desert, China) on 14 December 2021. In-situ measurements of ground reflectance, aerosol optical depth (AOD), total columnar water vapor, radiosonde data, and diffuse-to-global irradiance (DG) ratio were performed to predict the top-of-atmosphere radiance by the reflectance-, irradiance-, and improved irradiance-based methods using the moderate resolution atmospheric transmission model. The MII calibration coefficients were calculated by dividing the top-of-atmosphere radiance by the average digital number value of the image. The radiometric calibration coefficients calculated by the three calibration methods were reliable (average relative differences: 2.20% (reflectance-based vs. irradiance-based method) and 1.43% (reflectance-based vs. improved irradiance-based method)). The total calibration uncertainties of the reflectance-, irradiance-, and improved irradiance-based methods were 2.77–5.23%, 3.62–5.79%, and 3.50–5.23%, respectively. The extra DG ratio measurements in the latter two methods did not improve the calibration accuracy for AODs ≤ 0.1. The calibrated MII images were verified using Landsat-8 Operational Land Imager (OLI) and Sentinel-2A MultiSpectral Instrument (MSI) images. The retrieved ground reflectances of the MII over different surface types were cross-compared with those of OLI and MSI using the FAST Line-of-sight Atmospheric Analysis of Hypercubes software. The MII retrievals differed by <0.0075 (7.13%) from OLI retrievals and <0.0084 (7.47%) from MSI retrievals for calibration coefficients from the reflectance-based method; <0.0089 (7.57%) from OLI retrievals and <0.0111 (8.65%) from MSI retrievals for the irradiance-based method; and <0.0082 (7.33%) from OLI retrievals and <0.0101 (8.59%) from MSI retrievals for the improved irradiance-based method. Thus, our findings support the application of SDGSAT-1 data.

Mapping the forage nitrogen, phosphorus, and potassium contents of alpine grasslands by integrating Sentinel-2 and Tiangong-2 data

Nitrogen (N), phosphorus (P), and potassium (K) contents are crucial quality indicators for forage in alpine natural grasslands and are closely related to plant growth and reproduction. One of the greatest challenges for the sustainable utilization of grassland resources and the development of high-quality animal husbandry is to efficiently and accurately obtain information about the distribution and dynamic changes in N, P, and K contents in alpine grasslands. A new generation of multispectral sensors, the Sentinel-2 multispectral instrument (MSI) and Tiangong-2 moderate-resolution wide-wavelength imager (MWI), is equipped with several spectral bands suitable for specific applications, showing great potential for mapping forage nutrients at the regional scale. This study aims to achieve high-accuracy spatial mapping of the N, P, and K contents in alpine grasslands at the regional scale on the eastern Qinghai-Tibet Plateau. The Sentinel-2 MSI and Tiangong-2 MWI data, coupled with multiple feature selection algorithms and machine learning models, are applied to develop forage N, P, and K estimation models from data collected at 92 sample sites ranging from the vigorous growth stage to the senescent stage. The results show that the spectral bands of both the Sentinel-2 MSI and Tiangong-2 MWI have an excellent performance in estimating the forage N, P, and K contents (the R2 values are 0.68–0.76, 0.54–0.73, and 0.74–0.82 for forage N, P, and K estimations, respectively). Moreover, the model integrating the spectral bands of these two sensors explains 78%, 74%, and 84% of the variations in the forage N, P, and K contents, respectively. These results indicate that the estimation ability of forage nutrients can be further improved by integrating Tiangong-2 MWI and Sentinel-2 MSI data. In conclusion, integration of the spectral bands of multiple sensors is a promising approach to map the forage N, P, and K contents in alpine grasslands with high accuracy at the regional scale. This study offers valuable information for growth monitoring and real-time determination of forage quality in alpine grasslands.

Orthorectification of Data from the AHI Aboard the Himawari-8 Geostationary Satellite

The use of geostationary meteorological satellites for land remote sensing has attracted much attention after the launch of the Himawari-8 satellite equipped with a sensor with enhanced land observation capabilities. In the context of land remote sensing, geolocation errors are often a critical issue, especially in mountainous regions, where a precise orthorectification process is required to maintain high geometric accuracy. The present work addresses the issues related to orthorectification of the new-generation geostationary Earth orbit (GEO) satellites by applying an algorithm known as the ray-tracing indirect method to the data acquired by the Advanced Himawari Imager (AHI) aboard the Himawari-8 satellite. The orthorectified images of the AHI were compared with data from the Sentinel-2 Multispectral Instrument (MSI). The comparison shows a clear improvement of the geometric accuracy, especially in high-elevation regions located far from the subsatellite point. The results indicate that approximately 7.3% of the land pixels are shifted more than 3 pixels during the orthorectification process. Furthermore, the maximum displacement after the orthorectification is up to 7.2 pixels relative to the location in the original image, which is of the Tibetan Plateau. Moreover, serious problems caused by occlusions in the images of GEO sensors are clearly indicated. It is concluded that special caution is needed when using data from GEO satellites for land remote sensing in cases where the target is in a mountainous region and the pixels are located far from the subsatellite point.

Comparison of Simulated Multispectral Reflectance among Four Sensors in Land Cover Classification

Multispectral images accessible free of charge have increased significantly from the acquisitions by the wide-field-of-view (WFV) sensors onboard Gaofen-1/-6 (GF-1/-6), the Operational Land Imager (OLI) onboard Landsat 8 (L8), and the Multi-Spectral Instrument (MSI) onboard Sentinel-2 (S2). These images with medium spatial resolutions are beneficial for land-cover mapping to monitor local to global surface dynamics. Comparative analyses of the four sensors in classification were made under different scenarios with five classifiers, mainly based on the simulated multispectral reflectance from well-processed hyperspectral data. With channel reflectance, differences in classification between the L8 OLI and the S2 MSI were generally dependent on the classifier considered, although the two sensors performed similarly. Meanwhile, without channels over the shortwave infrared region, the GF-1/-6 WFVs showed inferior performances. With channel reflectance, the support vector machine (SVM) with Gaussian kernel generally outperformed other classifiers. With the SVM, on average, the GF-1/-6 WFVs and the L8 OLI had great increases (more than 15%) in overall accuracy relative to using the maximum likelihood classifier (MLC), whereas the overall accuracy improvement was about 13% for the S2 MSI. Both SVM and random forest (RF) had greater overall accuracy, which partially solved the problems of imperfect channel settings. However, under the scenario with a small number of training samples, for the GF-1/-6 WFVs, the MLC showed approximate or even better performance compared to RF. Since several factors possibly influence a classifier’s performance, attention should be paid to a comparison and selection of methods. These findings were based on the simulated multispectral reflectance with focusing on spectral channel (i.e., number of channels, spectral range of the channel, and spectral response function), whereas spatial resolution and radiometric quantization were not considered. Furthermore, a limitation of this paper was largely associated with the limited spatial coverage. More case studies should be carried out with real images over areas with different geographical and environmental backgrounds. To improve the comparability in classification among different sensors, further investigations are definitely required.

Area Estimation of Mango and Coconut Crops using Machine Learning in Hesaraghatta Hobli of Bengaluru Urban District, Karnataka

Timely and accurate estimation of acreage and production of horticulture crops is necessary for deciding how much, where and when to export these commodities in the national and global markets. Remote sensing has been one of the methods adopted, in addition to conventional sampling methods, for improving the estimates. Parametric image classification algorithms have been used by many researchers for identification and area estimation of horticulture crops. But these algorithms result in several unclassified pixels leading to over/underestimates. This study has been undertaken to estimate the area of two horticulture crops (i.e., mango and coconut) of Hesaraghatta hobli of Bengaluru urban district using Convolutional Neural Network (CNN) on Google Colaband Random Forest (RF) algorithms on Google Earth Engine (GEE). Remotely sensed data acquired by the Multi-Spectral Instrument (MSI) onboard Sentiel-2A satellite was used. Spectral signatures of horticulture crops and other associated cover types have been generated to identify the cover types and for selecting appropriate band combinations. Two different band combinations were used for area estimation of selected horticulture crops: i) Near-InfraRed (NIR), Red, and Green all three having a spatial resolution of 10 m, ii) Red edge-3, Short-Wave InfraRed1 (SWIR1) and Short-Wave InfraRed2 (SWIR2) having 20 m spatial resolution. Area estimates of horticulture crops and associated cover types were validated with respect to ground truth and statistical reports from Karnataka State Directorate of Horticulture (KSDH). It was found that the CNN model performed better than RF using NIR, Red, and Green band combination with an overall accuracy of 84%, but it failed to give similar accuracies with Red edge 3, SWIR1, and SWIR2 band combination. We attempted transfer learning using the trained CNN model at two different study areas far away from the study area and found encouraging results.

Validation of Simplified Level 2 Prototype Processor Sentinel-2 fraction of canopy cover, fraction of absorbed photosynthetically active radiation and leaf area index products over North American forests

Canopy biophysical variables such as the fraction of canopy cover (fCOVER), fraction of absorbed photosynthetically active radiation (fAPAR), and leaf area index (LAI) are widely used for ecosystem modelling and monitoring. The Sentinel-2 mission was designed for systematic global mapping of these variables at 20 m resolution using imagery from the MultiSpectral Instrument. The Simplified Level 2 Prototype Processor (SL2P) is available as a baseline mapping solution. Previous validation over limited sites indicates that SL2P generally satisfies user requirements for all three variables over crops, but underestimates LAI over forests. In this study, Sentinel-2 fAPAR, fCOVER, and LAI products, from SL2P, were validated over 281 sites representative of most North American forest ecozones and also compared to Moderate Resolution Imaging Spectrometer (MODIS) and Copernicus Global Land Service (CGLS) products. In addition to meeting the Committee on Earth Observation Satellites Stage 3 validation requirements for these areas, our study also explores the relationship between bias in SL2P products and canopy clumping and provides empirical bias correction functions for each variable.SL2P was implemented within the Landscape Evolution and Forecasting Toolbox in Google Earth Engine both for efficiency and due to bugs in the Sentinel Application Platform implementation. SL2P was found to underestimate LAI by 20% to 50% over forests with LAI > 2; in agreement with other studies and with comparisons to MODIS and CGLS products. SL2P bias for fCOVER and fAPAR transitions from ∼0.1 at low values to ∼ − 0.1 at high values. Precision error, at one standard deviation, was ∼0.5 for LAI and slightly less than ∼0.1 for fCOVER and fAPAR. Total uncertainty was dominated by bias for LAI and was slightly greater than precision error for fCOVER and fAPAR. Target user requirements were satisfied for 51% of LAI, 37% of fCOVER and 31% of fAPAR comparisons to in-situ measurements. For all variables, accuracy exhibited weak to moderate linear relationships to clumping (r2 ≤0.52), but scatter plots indicated larger negative LAI biases over northern latitude sites where canopies exhibited greater clumping. With the exception of evergreen broadleaf forests, empirical bias correction using in-situ data reduced accuracy error by 40% for fCOVER, 57% for fAPAR and, 92% for LAI and increased the agreement rate with uncertainty requirements by up to 8%. Users of SL2P LAI over forests are recommended to apply bias correction or consider recalibrating SL2P with spatially heterogenous radiative transfer model simulations.

Window-Based Filtering Aerosol Retrieval Algorithm of Fine-Scale Remote Sensing Images: A Case Using Sentinel-2 Data in Beijing Region

The satellite-based Aerosol Optical Depth (AOD) retrieval algorithms are generally needed to construct Land Surface Reflectance (LSR) database. However, errors are unavoidable due to the surface complexity, especially for the short observation period and high-resolution images, such as Sentinel-2 Multi-Spectral Instrument (MSI) data. To address this, reference day images are used instead of the LSR database. The surface is assumed to be Lambertian; however, the fact is that not all pixels meet it well. Therefore, we proposed a window-based AOD retrieval algorithm, which can ignore the unreliable/non-Lambertian pixels in a retrieval window based on two main filtering processes. Finally, using Sentinel-2 Band 1 (60 m), the AODs (120 m) of 134 reference images to 43 reference images were retrieved by this algorithm from 2017 to 2021 in Beijing region, China. The results show that the retrieved AOD with the proposed algorithm exhibits good agreement with the ground-based measured AOD (R > 0.97). The high-resolution AOD presents comparable spatial distributions to the Multiangle Implementation of Atmospheric Correction (MAIAC) algorithm AOD (1 km) products. Moreover, the very little noise and very high spatial continuity of retrieval AOD imply that this algorithm could be ported to other algorithms as part of improving AOD quality.

The Development of Dark Hyperspectral Absolute Calibration Model Using Extended Pseudo Invariant Calibration Sites at a Global Scale: Dark EPICS-Global

This research aimed to develop a novel dark hyperspectral absolute calibration (DAHAC) model using stable dark targets of “Global Cluster-36” (GC-36), one of the clusters from the “300 Class Global Classification”. The stable dark sites were identified from GC-36 called “Dark EPICS-Global” covering the surface types viz. dark rock, volcanic area, and dark sand. The Dark EPICS-Global shows a temporal variation of 0.02 unit reflectance. This work used the Landsat-8 (L8) Operational Land Imager (OLI), Sentinel-2A (S2A) Multispectral Instrument (MSI), and Earth Observing One (EO-1) Hyperion data for the DAHAC model development, where well-calibrated L8 and S2A were used as the reference sensors, while EO-1 Hyperion with a 10 nm spectral resolution was used as a hyperspectral library. The dark hyperspectral dataset (DaHD) was generated by combining the normalized hyperspectral profile of L8 and S2A for the DAHAC model development. The DAHAC model developed in this study takes into account the solar zenith and azimuth angles, as well as the view zenith and azimuth angles in Cartesian coordinates form. This model is capable of predicting TOA reflectance in all existing spectral bands of any sensor. The DAHAC model was then validated with the Landsat-7 (L7), Landsat-9 (L9), and Sentinel-2B (S2B) satellites from their launch dates to March 2022. These satellite sensors vary in terms of their spectral resolution, equatorial crossing time, spatial resolution, etc. The comparison between the DAHAC model and satellite measurements showed an accuracy within 0.01 unit reflectance across the overall spectral band. The proposed DAHAC model uncertainty level was determined using Monte Carlo simulation and found to be 0.04 and 0.05 unit reflectance for the VNIR and SWIR channels, respectively. The DAHAC model double ratio was used as a tool to perform the inter-comparison between two satellites. The sensor inter-comparison results for L8 and L9 showed a 2% difference and 1% for S2A and S2B across all spectral bands.

Uni-temporal Sentinel-2 imagery for wildfire detection using deep learning semantic segmentation models

Wildfires are common disasters that have long-lasting climate effects and serious ecological, social, and economic effects due to climate change. Since Earth observation (EO) satellites were launched into space, remote sensing (RS) has become a more efficient technique that can be used in agriculture, environmental protection, geological exploration, and wildfires. The increasing number of EO satellites orbiting the earth provides huge amounts of data, such as Sentinel-2 with its Multi Spectral Instrument (MSI) sensor. Using uni-temporal Sentinel-2 imagery, we proposed a workflow based on deep learning (DL) semantic segmentation models to detect wildfires. In particular, we created a new big wildfire dataset suitable for semantic segmentation models. We tested our dataset using DL models such as U-Net, LinkNet, DeepLabV3+, U-Net++, and Attention ResU-Net. The results are analysed and compared in terms of the F1 score, the intersection over union (IoU) score, the precision and recall metrics, and the amount of training time each model possesses. The best results were achieved using U-Net with the ResNet50 encoder, with F1-score of 98.78% and IoU of 97.38%, and we developed it into a pre-trained DL Package (DLPK) model that is able to detect and monitor the wildfire from Sentinel-2 images automatically.

Overview of the Application of Remote Sensing in Effective Monitoring of Water Quality Parameters

This study provides an overview of the techniques, shortcomings, and strengths of remote sensing (RS) applications in the effective retrieval and monitoring of water quality parameters (WQPs) such as chlorophyll-a concentration, turbidity, total suspended solids, colored dissolved organic matter, total dissolved solids among others. To be effectively retrieved by RS, these WQPs are categorized as optically active or inactive based on their influence on the optical characteristics measured by RS sensors. RS applications offer the opportunity for decisionmakers to quantify and monitor WQPs on a spatiotemporal scale effectively. The use of RS for water quality monitoring has been explored in many studies using empirical, analytical, semi-empirical, and machine-learning algorithms. RS spectral signatures have been applied for the estimation of WQPs using two categories of RS, namely, microwave and optical sensors. Optical RS, which has been heavily applied in the estimation of WQPs, is further grouped as spaceborne and airborne sensors based on the platform they are on board. The choice of a particular sensor to be used in any RS application depends on various factors including cost, and spatial, spectral, and temporal resolutions of the images. Some of the known satellite sensors used in the literature and reviewed in this paper include the Multispectral Instrument aboard Sentinel-2A/B, Moderate Resolution Imaging Spectroradiometer, Landsat Thematic Mapper, Enhanced Thematic Mapper, and Operational Land Imager.

Mapping of Land Use in Cijengkol Village, Subang Regency Using Sentinel-2 MSI (MultiSpectral Instrument)

Every year, land use in Indonesia has increased, both for settlements, agriculture, and other uses that are used to meet the needs of human life for certain purposes. Cijengkol Village is one of the agricultural development villages in Subang Regency and is affected by topography, resulting in different types of land use. This mapping aimed to provide information related to the classification of land use for settlements, agriculture, plantations, fields, and others in Cijengkol Village. Land use mapping was carried out in this village to reveal the distribution of land use so that it could be taken into consideration, as well as directions for determining spatial planning by the local government. Therefore, this mapping was carried out by involving the Sentinel-2 MultiSpectral Instrument (MSI) image data source and processed using a cloud computing-based Google Earth Engine (GEE) platform. Six spectral scoring index algorithms exist the Enhanced Vegetation Index (EVI), Soil Adjusted Vegetation Index (SAVI), Specific Leaf Area Vegetation Index (SLAVI), Index-Based Built-Up Index (IBI), Normalized Difference Built-up Index (NDBI), and the Normalized Difference Water Index (NDWI). The results of the random forest (RF) classification algorithm resulted in six types of land use with percentages, namely mixed gardens (39.69%), agriculture (34.08%), homogeneous gardens (13.57%), residential (10.58%), open land (2.09%), and water bodies (0.001%). Image classification in this mapping also produces an accuracy rate of 82.43% (Overall Accuracy) and 0.78 (Kappa Statistics). The results of this research are of a good level of accuracy, so it is hoped that this research will become a database for the local village government and become a reference for further research.

Modelling Above-ground Biomass Using Machine Learning Algorithm: Case Study Miombo Woodlands of Malawi

Malawi’s Miombo Woodlands provide critical wood products and ecosystem services, including carbon storage through above ground-biomass (AGB). However, accurate and efficient estimation of AGB and C remains a huge challenge due to the complexity of heterogeneous natural and anthropogenic environmental factors exacerbated by choice of data collection and analysis techniques. A study was conducted to explore the potential of Machine Learning (ML) algorithmic on modelling AGB in Malawi’s Miombo Woodlands using remotely sensed data. A combination of AGB field measurements and Sentinel 2 Multi-Spectral Instrument (S2 MSI) imagery, and environmental data were employed to train and evaluate the performance of Random Forest (RF) regression ML model, thus against the backdrop of traditional simple linear and multiple linear regression models used in the past. The randomForest package in Rstudio R-3.6.1 software was used to model the S2 median seasonal composite band imagery datasets acquired in 2019 over Ntchisi Forest. Results demonstrate an outstanding performance of the RF with Mean Decrease in Accuracy (MDA) importance of >80 acquired from the Green, Blue and Near-Infra-Red (NIR) bands of the post-rainy season imagery while the dry season scene registered a MDA importance of >52 obtained from the Red, Blue and NIR bands. These findings underscore the superiority of the RF, revealing its ability to model AGB using different spectral bands, and more importantly, in different seasons. The outcomes have also revealed the better accuracy that the RF render in generating AGB maps (providing a minimum range of 25-33 tCha-1, which is in line with the ground reference estimates). Overall, the study has shown that RF Regression is a better technique to estimate AGB in Miombo Woodlands using the S2 MSI imagery under the given conditions. In conclusion, the research has further shown that the optical band combination of NIR and the Red regions of the Electro-Magnetic Spectrum are indispensable variables for successful modelling of AGB as well as delineating forest cover from non-forest attributes in the Miombo Woodlands. However, the study recommends exploring modelling the AGB using averaged vegetation indices i.e., NDVI, SAVI, and MSAVI, for robust results.

Machine-learning-estimation of high-spatiotemporal-resolution chlorophyll-a concentration using multi-satellite imagery

Chlorophyll-a concentration for quantifying phytoplankton biomass is commonly used as an indicator for evaluating the trophic level of lakes and water quality. This research aimed to develop a high spatiotemporal-resolution model for the retrieval of chlorophyll-a in inland water. Firstly, the machine learning based models considering Sentinel-2 Multispectral Instrument and Sentinel-3 Ocean and Land Color Instrument (OLCI) images were applied to estimate chlorophyll-a concentrations (R2 = 0.873 and 0.822, respectively). The spatiotemporal fusion was performed to fuse the OLCI and MSI chlorophyll-a images with low temporal resolution but fine spatial-resolution, and with high temporal resolution but coarse spatial-resolution. The random forest was applied to fuse images from two distinct sensors, and to refine the spatial resolution of OLCI estimations to be the same as those of Sentinel-2 MSI. Results showed that the spatiotemporal fusion can estimate dense-temporal 10 m spatial resolution chlorophyll-a concentration in the Tsengwen Reservoir (Root-Mean-Square Error, RMSE = 1.25–1.47 μg L−1). The spatiotemporal fusion model was effectively applied to determine high spatiotemporal-resolution chlorophyll-a measurements in the aquatic system.

Understanding the Susceptibility of the Tropical Proglacial Environment in Peru Using Optical Imagery and Radon Measurements

The tropical glaciers of the Cordillera Blanca have played host to some of the most significant mass movements ever recorded in the world and Peru; many proglacial lakes formed in this mountain range have natural dikes made of moraine material, which, if they collapse, would present a risk for the cities located downstream of a proglacial lake, where the proglacial lake Palcacocha has a remarkable background regarding floods. The Sentinel-2 MSI (Multi-Spectral Instrument, Level-2A) has a specific band for snow probability mapping that indicates glaciers and snow cover; this is effective for recognizing proglacial lakes by calculating the NDWIice. It is also helpful for lithology with SWIR for granite moraine deposits and slate moraines in the proglacial environment Palcacocha; these deposits surround the proglacial lake, with NDWIice determining the perimeter where sediment interacts with the rocks and meltwater. In addition, there are high radon concentrations made by ice avalanche impacts on the proglacial lake. Unstable glacier blocks cause ice avalanches into this proglacial lake, and the radon responds to flow variations from these high-impact avalanches. We used the device RadonEye PLus2, which allows real-time detection of radon flux changes in the proglacial environment. Our results indicated that ice avalanches making a high impact in the proglacial lake cause turbulent flow and generate radon concentration marks with a rising magnitude, while the absence of ice avalanches in the lake will cause the values to go down. The relationships of radon concentrations in the atmosphere for a tropical proglacial environment are radon and temperature (R2 = 0.364), radon and humidity (R2 = 0.469). In a passive proglacial environment with prolonged rainfall, radon concentrations tend to decrease, with an inversely proportional relationship between humidity and radon in the tropical proglacial environment. Proglacial lakes in the tropical zone often have large volumes of freshwater with high slopes from tropical glaciers, and climate change effects are an imminent danger for nearby cities.

Variations in chlorophyll-a concentration in response to hydrodynamics in a flow-through lake: Remote sensing and modeling studies

A convolution neural network (CNN) machine learning algorithm is established for Chlorophyll-a retrieval in Poyang Lake using in-situ Chlorophyll-a concentrations and concurrent satellite image data. For comparison, other machine learning methods, e.g., deep neural network, XGBoost, random forest, gradient boost, decision tree, and linear regression, are also employed, together with two empirical models. Accuracy analysis validates the robustness of the CNN algorithm, with the R2 of 0.97 and the mean absolute errors of 1.74 μgL−1, and its performance is better than other machine learning and empirical algorithms. The Chl-a in Poyang Lake is derived from Sentinel-2 Multi-Spectral Instrument (MSI) images based on the robust CNN algorithm. As a follow-through type, Poyang Lake has abundant hydrodynamics affecting the Chl-a in the lake water. Thus, the Delft3D Flexible Mesh (FM) model is employed in Poyang Lake to simulate the active hydrodynamics with a high spatial resolution (∼80 m) under the realistic forcing in the period from 1 October 2020 to 15 March 2022. The simulation results reveal that the flow speed is a significant factor modulating the lake Chl-a with a high concentration in summer and a low concentration in winter; the flow speed and water storage exceed the temperature and solar radiance in modulating the Chl-a. The jacking effect of the Yangtze River can slow down the flow speed in the lake’s north channel, therefore, augmenting the Chl-a by ∼20%, favoring phytoplankton blooms in Poyang Lake. The active eddies may significantly affect the lake Chl-a; cyclonic eddies enhance the local Chl-a, and anticyclonic eddies weaken the Chl-a. A detailed analysis indicates that the Coriolis force plays a vital role in the momentum balance, which may cause the lift (restriction) of bottom nutrients in the cyclonic (anticyclonic) eddy area, increasing (decreasing) the surface nutrient supply. This study uses Poyang Lake as an example to show that such eddies conventionally appearing in a flow-through lake may dramatically affect the ecological environment in the lake water.

Sentinel-2 MSI Observations of Water Clarity in Inland Waters across Hainan Island and Implications for SDG 6.3.2 Evaluation

Freshwater on islands represents a precious resource and highly vulnerable ecosystem. For monitoring freshwater, satellite remote sensing is efficient and has large-scale application. This study proposed a modified model of the quasi-analytical algorithm (ZSD-QAAv6m) to retrieve the water clarity of inland waters (>1 km2) across Hainan Island, China using Sentinel-2 multispectral instrument data. By adjusting the threshold of Rrs(665), the proposed model could accurately estimate water clarity with diverse optical properties on the island and avoid underestimation in moderately clear waters. Based upon this, the first spatiotemporal analysis of recent water clarity in Hainan Island was conducted. The results show that lake water clarity in the central region was generally higher (with average value of 1.4 m) than that of coastal regions (with average value of 1.2 m). Seasonally, the water clarity during the wet season was usually lower than that in the dry season, with average values of 1.1 m and 1.3 m across the island respectively. From 2017 to 2021, the proportion of water bodies with water clarity > 0.5 m increased from 60% to 100%. The overall spatial pattern of water clarity was correlated to the regional vegetation cover in Hainan Island, with higher clarity associated with higher vegetation cover in the central regions. The seasonal variation of water clarity may be attributed to heavy rainfall and runoff during the wet season; while the distinct annual variation may be benefited from the strengthened surface water protections in Hainan Province in recent years. This study provides a practical approach for evaluating the SDG 6.3.2 indicator in Hainan Island using remote sensed water clarity as a comprehensive water quality indicator and the findings could facilitate the island’s water resource management and conservation.

Application of machine learning techniques to derive sea water turbidity from Sentinel-2 imagery

Earth Observation (EO) from satellites has the potential to provide comprehensive, rapid and inexpensive information about water bodies, integrating in situ measurements. Traditional methods to retrieve optically active water quality parameters from satellite data are based on semi-empirical models relying on few bands, which often revealed to be site and season specific. The use of machine learning (ML) for remotely sensed water quality estimation has spread in recent years thanks to the advances in algorithm development and computing power. These models allow to exploit the wealth of spectral information through more flexible relationships and are less affected by atmospheric and other background factors. The present study explores the use of Sentinel-2 MultiSpectral Instrument (MSI) Level-1C Top of Atmosphere spectral radiance to derive water turbidity, through application of machine learning techniques. A dataset of 222 combination of turbidity measurements, collected in the North Tyrrhenian Sea – Italy from 2015 to 2021, and values of the 13 spectral bands in the pixel corresponding to the sample location was used. Two regression techniques were tested and compared: a Stepwise Linear Regression (SLR) and a Polynomial Kernel Regression. The two models show accurate and similar performance (R2 = 0.736, RMSE = 2.03 NTU, MAE = 1.39 NTU for the SLR and R2 = 0.725, RMSE = 2.07 NTU, MAE = 1.40 NTU for the Kernel). A band importance analysis revealed the contribution of the different spectral bands and the main role of the red-edge range. The work shows that it is possible to reach a good accuracy in turbidity estimation from MSI TOA reflectance using ML models, fed by the whole spectrum of available bands, although the possible generation of errors related to atmospheric effect in turbidity estimates was not evaluated. Comparison between turbidity estimates obtained from the models with turbidity data from Copernicus CMEMS dataset named ‘Mediterranean Sea, Bio-Geo-Chemical, L3, daily observation’ produced consistent results. Finally, turbidity maps from satellite imagery were produced for the study area, showing the ability of the models to catch extreme events.

High Heterogeneity LULC Classification in Ujung Kulon National Park, Indonesia: A Study Testing 11 Indices, Random Forest, Sentinel-2 MSI, and GEE-based Cloud Computing

The Ujung Kulon National Park (UKNT) is one of the national parks on the island of Java and has an essential role in saving endemic species in Indonesia. As a form of national park conservation effort, the completeness of LULC spatial data is a primary database that is indispensable in determining national park management policies. Therefore, this research was conducted to map the LULC (Land Use - Land Cover) in the forest landscape with high heterogeneity in UKNT. Sentinel-2 MSI (Multispectral Instrument) image data were classified using the Random Forest (RF) classification algorithm and tested using 11 index algorithms. The classification process takes place on a cloud computing-based geospatial platform, Google Earth Engine (GEE). This test resulted in 10 LULC classes; water had the broadest percentage of 45.44%. Meanwhile, the primary forest has an area of 21,868.41 or about 19.53% of the total area of the national park. However, there are some discrepancies in the spatial information generated by this classification process, so it is considered necessary to evaluate the combination of indexes, training data, and classification algorithms to limit the classification area. Therefore, this study is expected to be considered for further research related to LULC in high-heterogeneity landscapes.