Multispectral Satellite Imagery Research Articles

Challenges for monitoring artificial turf expansion with satellite remote sensing.

Urban green spaces are central components of urban ecosystems, providing refuge for wildlife while helping 'future proof' cities against climate change. Conversion of urban green spaces to artificial turf has become increasingly popular in various developed countries, such as the UK, leading to reduced urban ecosystem services delivery. To date, there is no established satellite remote sensing method for reliably detecting and mapping artificial turf expansion at scale. We here assess the combined use of very high-resolution multispectral satellite imagery and classical, open source, supervised classification approaches to map artificial lawns in a typical British city. Both object-based and pixel-based classifications struggled to reliably detect artificial turf, with large patches of artificial turf not being any more reliably identified than small patches of artificial turf. As urban ecosystems are increasingly recognised for their key contributions to human wellbeing and health, the poor performance of these standard methods highlights the urgency of developing and applying new, easily accessible approaches for the monitoring of these important ecosystems.

Automatic detection of methane emissions in multispectral satellite imagery using a vision transformer

Curbing methane emissions is among the most effective actions that can be taken to slow down global warming. However, monitoring emissions remains challenging, as detection methods have a limited quantification completeness due to trade-offs that have to be made between coverage, resolution, and detection accuracy. Here we show that deep learning can overcome the trade-off in terms of spectral resolution that comes with multi-spectral satellite data, resulting in a methane detection tool with global coverage and high temporal and spatial resolution. We compare our detections with airborne methane measurement campaigns, which suggests that our method can detect methane point sources in Sentinel-2 data down to plumes of 0.01 km2, corresponding to 200 to 300 kg CH4 h−1 sources. Our model shows an order of magnitude improvement over the state-of-the-art, providing a significant step towards the automated, high resolution detection of methane emissions at a global scale, every few days.

Cross-sensor transfer learning for fire smoke scene detection using variable-bands multi-spectral satellite imagery aided by spectral patterns

ABSTRACT This paper addresses the challenge of training deep learning models for fire smoke scene detection from multi-sensor, multi-spectral satellite imagery, where spectral bands vary and training data is scarce, especially for new sensors. We introduce a novel cross-sensor transfer learning approach enhanced by spectral pattern extraction, notably aided by a deep learning (DL) module called Input Amplification (IA). Our contribution includes the creation of two datasets, Landsat_smk2770 and Sentinel2_smk, with varying spectral bands. We employ a model named IA_VIB_SD which incorporates IA and is designed specifically for fire smoke scene detection. We investigated the transferability of the Landsat_smk2770-trained IA_VIB_SD model to the Sentinel2_smk dataset using various transfer learning techniques, and compared the performance of the transferred model with the benchmark IA_VIB_SD model exclusively trained on pure Sentinel2_smk data. Our proposed transfer learning approach resulted in a transferred model with an average accuracy 5% higher than the benchmark model. Notably, Our proposed transfer learning approach outperformed existing transfer learning methods by more than 1% in terms of the accuracy, even when trained on a mere 10% of the Sentinel2_smk dataset. This research advances fire smoke scene detection from multi-sensor, multi-spectral satellite imagery.

Urban explosion and hotspots of forest loss in western Himalaya: Mapping land use/cover change trends since 1975

The Western Himalayan Region (WHR) is under constant threat from Land Use and Land Cover Change (LULCC) brought on by various socioeconomic activities such as urbanization, agricultural expansion, and population growth. Detailed research into the scope of these changes is necessary to comprehend the challenges they represent. This study aimed at (1) examining and quantifying spatiotemporal patterns of LULCC from 1975 to 2015, (2) mapping the hotspots of forest loss, and (3) modelling future LULCC for the year 2055 using medium spatial resolution multispectral satellite imagery (MSS, TM, LISS-I, LISS-III, AWiFS), field data, visual interpretation and logistic regression model in the WHR. Logistic regression integrated various drivers of LULCC, including climatic, topographic, and socio-economic factors and past LULCC trends to forecast potential LULC in the WHR by 2055. The results highlight an outrageous 184% increase in regional urbanization. Expansion of urban and cropland (12%) led to deforestation which reduced the area of natural habitats, specifically forests (−11%), water bodies (−8%) and scrubland (−6%). A significant rise in barren land by 30% was noted, while there was a decrease of 20% in glacier/snow melt. The future projection for 2055 anticipates continuing current trends, emphasizing an intensified urban expansion of 63% and an approximate 9% reduction in overall forest coverage. The accuracy of the predicted 2015 LULC map was assessed using ROC/AUC (Receiver Operating Characteristic/Area under Curve) which was found to be 0.79.

Estimation of Intercepted Solar Radiation and Stem Water Potential in a Table Grape Vineyard Covered by Plastic Film Using Sentinel-2 Data: A Comparison of OLS-, MLR-, and ML-Based Methods.

In the framework of precision viticulture, satellite data have been demonstrated to significantly support many tasks. Specifically, they enable the rapid, large-scale estimation of some viticultural parameters like vine stem water potential (Ψstem) and intercepted solar radiation (ISR) that traditionally require time-consuming ground surveys. The practice of covering table grape vineyards with plastic films introduces an additional challenge for estimation, potentially affecting vine spectral responses and, consequently, the accuracy of estimations from satellites. This study aimed to address these challenges with a special focus on the exploitation of Sentinel-2 Level 2A and meteorological data to monitor a plastic-covered vineyard in Southern Italy. Estimates of Ψstem and ISR were obtained using different algorithms, namely, Ordinary Least Square (OLS), Multivariate Linear Regression (MLR), and machine learning (ML) techniques, which rely on Random Forest Regression, Support Vector Regression, and Partial Least Squares. The results proved that, despite the potential spectral interference from the plastic coverings, ISR and Ψstem can be locally estimated with a satisfying accuracy. In particular, (i) the OLS regression-based approach showed a good performance in providing accurate ISR estimates using the near-infrared spectral bands (RMSE < 8%), and (ii) the MLR and ML algorithms could estimate both the ISR and vine water status with a higher accuracy (RMSE < 7 for ISR and RMSE < 0.14 MPa for Ψstem). These results encourage the adoption of medium-high resolution multispectral satellite imagery for deriving satisfying estimates of key crop parameters even in anomalous situations like the ones where plastic films cover the monitored vineyard, thus marking a significant advancement in precision viticulture.

An Introduction to Machine and Deep Learning Methods for Cloud Masking Applications

Cloud cover assessment is crucial for meteorology, Earth observation, and environmental monitoring, providing valuable data for weather forecasting, climate modeling, and remote sensing activities. Depending on the specific purpose, identifying and accounting for pixels affected by clouds is essential in spectral remote sensing imagery. In applications such as land monitoring and various remote sensing activities, detecting/removing cloud-contaminated pixels is crucial to ensuring the accuracy of advanced processing of satellite imagery. Typically, the objective of cloud masking is to produce an image where every pixel in a satellite spectral image is categorized as either clear or cloudy. Nevertheless, there is also a prevalent approach in the literature that yields a multi-class output. With the progress in Machine and Deep Learning, coupled with the accelerated capabilities of GPUs, and the abundance of available remote sensing data, novel opportunities and methods for cloud detection have emerged, improving the accuracy and the efficiency of the algorithms. This paper provides a review of these last methods for cloud masking in multispectral satellite imagery, with emphasis on the Deep Learning approach, highlighting their benefits and challenges.

Satellite-Derived Bathymetry in Support of Maritime Archaeological Research—VENμS Imagery of Caesarea Maritima, Israel, as a Case Study

Deriving bathymetry by means of multispectral satellite imagery proves to be a replicable method, offering high-resolution coverage over large areas while keeping costs low. Maritime archaeologists often require bathymetric mapping at a high resolution and with a large spatial coverage. In this paper, we demonstrate the implementation of SDB in maritime archaeology using high-resolution (5 m/pixel) data from Vegetation and Environment monitoring on a New Micro-Satellite (VENμS) imagery. We focus on the area of the Roman harbour of Sebastos, located at Caesarea Maritima along the Israeli coast of the Eastern Mediterranean. For extracting SDB, we take an empirical approach, which is based on the integration of satellite imagery and sonar depth measurements, resulting in a blue-green band ratio algorithm that provides reliable results up to a water depth of 17 m. Comparison with in situ depth measurements yielded an RMSE of 0.688 m. The SDB mapping is complemented by satellite-based identification of above- and below-water rocks. The presented approach can readily be replicated in other regions using various types of multispectral satellite imagery, particularly when only coarse bathymetric sonar data are available, thus substantially contributing to our ability to perform maritime archaeological research.

Automated classification of valid and invalid satellite derived bathymetry with random forest

Recent decades have seen rapid growth in algorithms and workflows for generating bathymetry from multispectral satellite imagery, with the output typically referred to as satellite derived bathymetry (SDB). An inherent challenge is that, while SDB algorithms generally output a value for every pixel in each input scene, the value of any particular raster cell may not represent a valid depth. Typically, large portions of a satellite scene will correspond to optically deep waters, where SDB depth retrieval is impossible. Boats, boat wakes, breaking waves, clouds, and land can also cause erroneous depth estimates. Compounding this challenge is the fact that SDB algorithms tend to be poor at self-diagnosing invalid retrievals. As a result, SDB grids often require substantial manual editing to generate reliable bathymetric digital elevation models (DEMs). This study investigates the ability to automatically classify valid and erroneous bathymetry from SDB grids using random forest and surface features generated from the SDB grids, spectral bands values, and band indices. The trained models achieved mean intersection over union accuracies of 73–91% at five geographically-diverse test sites of differing sizes, seafloor morphologies, substrates, and wind and wave climates. These methods can be used to rapidly assess satellite image-based bathymetry grids and are currently being adapted for operational use within the National Oceanic and Atmospheric Administration’s National Ocean Service.

Post-drainage vegetation, microtopography and organic matter in Arctic drained lake basins

Wetlands in Arctic drained lake basins (DLBs) have a high potential for carbon storage in vegetation and peat as well as for elevated greenhouse gas emissions. However, the evolution of vegetation and organic matter is rarely studied in DLBs, making these abundant wetlands especially uncertain elements of the permafrost carbon budget. We surveyed multiple DLB generations in northern Alaska with the goal to assess vegetation, microtopography, and organic matter in surface sediment and pond water in DLBs and to provide the first high-resolution land cover classification for a DLB system focussing on moisture-related vegetation classes for the Teshekpuk Lake region. We associated sediment properties and methane concentrations along a post-drainage succession gradient with remote sensing-derived land cover classes. Our study distinguished five eco-hydrological classes using statistical clustering of vegetation data, which corresponded to the land cover classes. We identified surface wetness and time since drainage as predictors of vegetation composition. Microtopographic complexity increased after drainage. Organic carbon and nitrogen contents in sediment, and dissolved organic carbon (DOC) and dissolved nitrogen (DN) in ponds were high throughout, indicating high organic matter availability and decomposition. We confirmed wetness as a predictor of sediment methane concentrations. Our findings suggest moderate to high methane concentrations independent of drainage age, with particularly high concentrations beneath submerged patches (up to 200 μmol l−1) and in pond water (up to 22 μmol l−1). In our DLB system, wet and shallow submerged patches with high methane concentrations occupied 54% of the area, and ponds with high DOC, DN and methane occupied another 11%. In conclusion, we demonstrate that DLB wetlands are highly productive regarding organic matter decomposition and methane production. Machine learning-aided land cover classification using high-resolution multispectral satellite imagery provides a useful tool for future upscaling of sediment properties and methane emission potentials from Arctic DLBs.

Cloud and snow neural network segmentation using the Electro-L No. 2 satellite low-resolution data

This paper is devoted to cloud and snow semantic segmentation using multispectral satellite imagery, received from a multizone scanning instrument (MSU-GS) installed on the Russian satellite Electro-L No. 2. For segmentation, some additional geographic information is used such as latitude, longitude, and altitude. The main difficulty of evaluating snow and cloud masks from MSU-GS data is absence of shortwave infrared (SWIR) spectral channels in the 1.3-1.6μm range, which are necessary for accurate snow and cloud distinction. This fact makes it impossible to apply any of existing algorithms for cloud and snow destinction used for other multispectral satellite imagery. The results of the presented work include a self-collected dataset, consisting of cloud masks from the meteorological satellites GOES-16 and Meteosat-10 L2 products, snow masks from Terra/MODIS products, considered to be the ground truth for multispectral imagery from Electro-L No. 2 with geographical information for each sample, as well as a trained Multi-Scale Attention Network (MANet), able to perform snow and cloud segmentation for multispectral data from Electro-L No. 2. The purpose of this study is to create a stable algorithm for MSU-GS imagery to conduct an operational snow and cloud coverage monitoring of the full Earth disk, that can not be evaluated by using multispectral data from other Russian satellites. The novelty of the work is the combining geographical and multispectral information with weighted sampling approach for hard samples, as well as using other satellite products for creating ground truth masks for needed satellite imagery. The developed algorithm was tested on various scenes during all seasons with F1 and IoU metrics. We reach segmentation quality equal to F1=0.7454,F1=0.8773 and IoU=0.7398,IoU=0.7976 for snow and clouds respectively. The proposed neural network can be used at any time of year during daytime for processing Electro-L No. 2 multispectral data in real time.

A river runs through it: Robust automated mapping of riparian woodlands and land surface phenology across dryland regions

Riparian woodlands in drylands are critically important to human society, global biodiversity, and regional water and energy budgets. These sensitive ecosystems have experienced substantial degradation over the last several decades from climatic change and direct human activity. Nevertheless, quantifying long-term change in dryland riparian woodlands remains a major challenge, and much uncertainty exists in their remaining extent, historical breadth, and likely future trajectories. Dryland landscapes show large, fine-scale spatial heterogeneity in seasonal greenness patterns, driven in part by spatial variation in water availability. Riparian woodlands occur where water is concentrated in the landscape, either as aboveground streamflow or subsurface groundwater. In arid and semi-arid climates, this renders them phenologically distinctive from upland ecosystems. However, despite their importance and distinctiveness, there are currently no automated methods for delineating dryland riparian woodlands across regional extents in the cloud. Here we designed and implemented a cloud-based algorithm to retrieve dryland land surface phenology patterns from multispectral satellite imagery and conducted sensitivity analyses using real and simulated data to demonstrate that the approach is robust for MODIS, Sentinel-2, and Landsat over realistic ranges of noise and cloud cover. We then designed a series of random forest vegetation classifiers that integrate phenological and spectral information, vegetative structure from LiDAR, and topography from LiDAR or the Shuttle Radar Topography Mission. We implemented classifiers for three local study sites and then generalized our model to run regionally across the southwestern United States, with balanced accuracy for the riparian woodland class ranging from 94.5% to 97.5% when validated with local to regional datasets. Generally, phenological information proved more important than any other data source for mapping riparian woodlands, which showed more stability in interannual phenology than did upland vegetation types. To our knowledge, ours is the first regional, annual, automatically-generated and updated approach for mapping dryland riparian woodlands in the southwestern United States, paving the way for improved modeling and management efforts on watershed to regional scales. We also provide one of the first operational, exclusively cloud-based methods to extract dryland land surface phenology patterns using Landsat, Sentinel-2, MODIS, or other sensors, providing a framework for future studies investigating other aspects of long-term or spatial variation in dryland vegetative seasonality across the globe.

Delineating wildfire burns and regrowth using satellite imagery to assess moose (Alces alces) spatial responses to burns

AbstractClimate change is predicted to increase the frequency and size of wildfires. Wildfire burns can influence wildlife space use, but prior to understanding the relationship between wildlife and burns, the extent of wildfires and the regrowth of burned areas must be determined. We used remotely sensed, multispectral satellite imagery to calculate differenced normalized burn ratios (NBRs) to define areas burned by wildfire and assess temporal trends in burn frequency and extent in central British Columbia. We then used NBRs as an index of vegetation regrowth to model regrowth trajectories within burns. Next, we examined spatial responses to burns by a moose (Alces alces) population that recently declined in abundance. We utilized locations used by and available to collared moose to determine spatial responses as a function of years since wildfire and vegetation regrowth. We also assessed functional responses to burns by moose, dependent upon the proportion of burned areas available to each individual. We hypothesized that (h1) wildfire burns within our study area increased from 1985 to 2017, (h2) that moose increase their use of burns in comparison with availability as burns age as a result of vegetation regrowth, and (h3) that moose demonstrate functional responses to burns (i.e., selection of burns by individual moose declines as the amount of burned areas increases). We observed that the frequency and extent of burns are increasing within our study area. In contrast to h2, moose in summer and fall decreased the selection of burns as burns aged. We, however, observed that moose responded most positively to locations with intermediate vegetation regrowth (i.e., NBR values), which according to burn regrowth trajectories were most likely to be achieved 13 or more years post‐wildfire; given that most burns within our study area were ≤13 years of age, we predict that the use of burns will likely increase as vegetation regrows. Moose only selected for burns in fall, and we did not find support for the presence of a functional response. Our research demonstrated the utility of remotely sensed imagery and NBRs to define burn locations and to reveal current and probable future spatial responses to burns by a wildlife population of concern.

Satellite Synthetic Aperture Radar, Multispectral, and Infrared Imagery for Assessing Bridge Deformation and Structural Health—A Case Study at the Samuel de Champlain Bridge

A space-borne remote sensing method was applied, validated, and demonstrated in a case study on the Samuel de Champlain Bridge in Montreal, Canada. High-resolution C-band radar satellite imagery was analyzed using the Persistent Scatterer Interferometric Synthetic Aperture Radar technique to derive bridge displacements and compare them against theoretical estimates. Multispectral and long-wave thermal infrared satellite imagery acquired during the InSAR observation period and historical environmental data were analyzed to provide context for the interpretation and understanding of InSAR results. Thermal deformation measurements compared well with their theoretical estimates based on known bridge geometry and ambient temperature data. Non-thermal deformation measurements gave no evidence of settlement during the 2-year monitoring period, as would normally be expected for a newly constructed bridge with its foundation on bedrock. The availability of environmental data obtained from multispectral and thermal infrared satellite imagery was found to be useful in providing context for the bridge stability assessment. Ambient temperature measurements from thermal infrared satellite imagery were found to be a suitable alternative in cases where data from in situ temperature sensors or nearby weather stations are not available or not fit for purpose. No strong correlation was found between the river conditions and bridge deformation results from the InSAR analysis; this is partly due to the fact that most of these effects act along the river flow in the north–south direction, to which the satellite sensor is not sensitive.

Learning class-specific spectral patterns to improve deep learning-based scene-level fire smoke detection from multi-spectral satellite imagery

This paper introduces the Input Amplification (InAmp) module, a deep learning (DL) innovation designed to enable DL models to automatically learn class-specific spectral patterns for accurate fire smoke detection using satellite imagery. Traditional fire smoke detection methods depend on threshold values manually derived from satellite spectral bands, based on domain knowledge. These thresholds are challenging to generalise and integrate into existing DL models, which are not inherently equipped to learn such spectral patterns. InAmp addresses this gap by explicitly and automatically learning class-specific spectral patterns from multi-spectral satellite imagery. Functioning as an input pre-processing block, InAmp facilitates seamless integration with various DL architectures. The effectiveness of InAmp is demonstrated through its integration with Convolutional Neural Network (CNN) architectures and testing on two datasets: USTC_SmokeRS (MODIS, three bands) and Landsat_Smk (Landsat 5/8, six bands). Our results show improved fire smoke detection accuracy, and we also present how InAmp successfully extracts class-specific spectral patterns, showcasing its potential in advancing fire smoke detection based on satellite imagery.

Igniting Urbanization and Development of Mamfe Town through Road Infrastructure: Case of the Bamenda-Mamfe Highway

Road infrastructure is set to be a catalyst to urbanization and development in Sub-Saharan African towns. This article hypothesises that the Bamenda-Mamfe highway has contributed to the urban growth and development of Mamfe. The paper reveals that spatial growth, socio-economic development and trade of Mamfe town have been triggered by the highway. A multi-spectral Landsat satellite imagery of 1986 and 2020 were used to determine the spatial expansion of the town over a period of 34 years. Some 110 questionnaires were administered to households of 7 neighbourhoods in Mamfe to determine the outcome of expansion. Field observation and interview sessions were done to have information on the socio-economic development of the town. An in-depth secondary data collection was done to gather information on the rate of population growth in Mamfe. The data was processed to generate maps through the ArcGIS 10.4 and MapInfo soft wares and tables through SPSS 17 for results and analyses. Results from findings propound a rapid increase in the spatial expansion of Mamfe instigated by the construction of road as the urban space increased from 21% in 1986 to 59% in 2020. This was accompanied by a population upsurge of 119,000 inhabitants from 18,000 in 1987 to 137,000 inhabitants in 2020. There has been a remarkable increase in trade and commerce influenced by the construction of the road as people and goods can move in and out of the town with ease. The socio-economic development of Mamfe has as well improved with the presence of commercial centres, schools and health units. The development of Mamfe is however faced with planning challenges and urban ills which might escalate if not controlled. This paper recommends that stringent majors are taken to implement planning and proper security in Mamfe with the hope to render the town sustainable for living.

A Convolutional Neural Network approach for image-based anomaly detection in smart agriculture

The recent technological advances and their applications to agriculture provide leverage for the new paradigm of smart agriculture. Remote sensing applications can help optimise resources, making agriculture more ecological, increasing productivity and helping farmers to anticipate events that could not otherwise be avoided. Considering that losses caused by anomalies such as diseases, weeds and pests account for 20–40 % of overall agricultural productivity, a successful research effort in this area would be a breakthrough for agriculture. In this paper, we propose a methodology with which to discover and classify anomalies in images of crops, taken from a wide range of distances, using different Convolutional Neural Network architectures. This methodology also deals with several difficulties that usually appear in this kind of problems, such as class imbalance, the insufficient and small variety of images, overtraining or lack of models generalisation. We have implemented four convolutional neural network architectures in a high-performance computing environment, and propose a methodology based on data augmentation with the addition of Gaussian noise to the images to solve the above problems. Our approach was tested using two well-established open datasets that are unalike: DeepWeeds, which provides a classification of 8 weed species native to Australia using images that were taken at a distance of 1 m, and Agriculture-Vision, which classifies 6 types of crop anomalies using multispectral satellite imagery. Our methodology attained accuracies of 98 % and 95.3% respectively, improving the state-of-the-art by several points. In order to ease reproducibility and model selection, we have provided a comparison in terms of computational time and other metrics, thus enabling the choice between architectures to be made according to the resources available. The complete code is available in an open repository in order to encourage reproducibility and promote scientific advances in sustainable agriculture.

Land use land cover mapping and snow cover detection in Himalayan region using machine learning and multispectral Sentinel-2 satellite imagery

Land use land cover mapping and snow cover detection in Himalayan region using machine learning and multispectral Sentinel-2 satellite imagery

Boreal to temperate forests transition within Kunashir Island (the Kuril Archipelago): Insights from the novel vegetation map

The vegetation of the Kuril Islands remains insufficiently explored due to their remote location and harsh environment. Understanding the factors that influence vegetation distribution in this extensive volcanic chain of islands, spanning over a thousand kilometers, is important for global biogeography. Our paper presents the first vegetation map of Kunashir Island from the Kuril archipelago, compiled using Sentinel-2 medium-resolution multispectral satellite imagery, incorporating original field research data and the Random Forest machine-learning algorithm. The temperate broad-leaved forests are mainly composed of Acer mono subsp. mayrii and Quercus mongolica, predominantly limited to the southwestern coast, and true-boreal dark-coniferous Abies sachalinensis–Picea jezoensis forests located in the northeast part of Kunashir. The vegetation distribution pattern is associated with the warming effect of the Soja current on the east coast of Kunashir Island and the cooling effect of the Kuril current on the west coast. Kunashir's vegetation displays features of an ecotone between temperate and boreal zones. Still, the boundary between these zones should be drawn within the island, rather than between Kunashir and Hokkaido or Kunashir and Iturup, as suggested in some previous biogeographic generalizations.

COMPARATIVE ANALYSES OF SECONDARY ECOLOGICAL SUCCESSION FOLLOWING WILDFIRES IN THREE DISTINCT FOREST TYPES. A STUDY CASE FROM MOGUER, SPAIN

Wildfires have become increasingly prevalent and destructive in forest ecosystems worldwide, necessitating a comprehensive understanding of post-fire recovery dynamics for effective conservation and management. Remote sensing technology, coupled with vegetation indices such as Normalized Burn Ratio (NBR), Normalized Difference Vegetation Index (NDVI), Green Red Vegetation Index (GRVI), and Red Vegetation Index (RVI), offers a powerful means to investigate these processes. In this study, we utilize remote sensing techniques to conduct a comparative analysis of secondary ecological succession following wildfires in three distinct forest types (Coniferous, Sclerophyll and Mixed) of a forest affected by fire near Moguer, Spain. Through the acquisition and analysis of multispectral satellite imagery, we monitored changes in vegetation health and recovery across the region of interest. The NBR index allowed us to assess the severity and extent of wildfire damage, while NDVI quantified vegetation greenness and regrowth. GRVI and RVI provided insights into subtle variations in vegetation composition and health. We identified distinct temporal and spatial patterns in post-fire recovery among the different forest types by applying these indices for the period between 2017 and 2021. Our findings underscore the significance of understanding the diverse responses of these ecosystems to wildfires. While common recovery patterns emerged, such as an initial decrease in NDVI followed by regeneration, variations were observed in the timing and magnitude of recovery. These distinctions are attributed to differences in species composition, fire adaptations, and ecological processes specific to each forest type. In conclusion, the utilization of NBR, NDVI, GRVI, and RVI indices allows for a more nuanced evaluation of post-fire recovery dynamics.

Mapping the tidal marshes of coastal Virginia: a hierarchical transfer learning approach

ABSTRACT Coastal wetlands, especially tidal marshes, play a crucial role in supporting ecosystems and slowing shoreline erosion. Accurate and cost-effective identification and classification of various marsh types, such as high and low marshes, are important for effective coastal management and conservation endeavors. However, mapping tidal marshes is challenging due to heterogeneous coastal vegetation and dynamic tidal influences. In this study, we employ a deep learning segmentation model to automate the identification and classification of tidal marsh communities in coastal Virginia, USA, using seasonal, publicly available satellite and aerial images. This study leverages the combined capabilities of Sentinel-2 and National Agriculture Imagery Program (NAIP) imagery and a UNet architecture to accurately classify tidal marsh communities. We illustrate that by leveraging features learned from data abundant regions and small quantities of high-quality training data collected from the target region, an accuracy as high as 88% can be achieved in the classification of marsh types, specifically high marsh and low marsh, at a spatial resolution of 0.6 m. This study contributes to the field of marsh mapping by highlighting the potential of combining multispectral satellite imagery and deep learning for accurate and efficient marsh type classification.