Diploma thesis deals with the land cover classification in Sidama region of Ethiopia and 2 kebeles, Chancho and Dangora Morocho. High resolution Sentinel-2 and very high resolution PlanetScope satellite images are used. The development of the classification algorithm is done in the Google Earth Engine cloud based environment. Ten combinations of the 4 most important parameters of the Random Forest classification method are tested. The defined legend contains 8 land cover classes, namely built-up, crops, grassland/pasture, forest, scrubland, bareland, wetland and water body. The training dataset is collected in the field during the fall 2020. The classification results of the two data types at two scales are compared. The highest overall accuracy for land cover classification of Sidama region came out to be 84.1% and kappa index of 0.797, with Random Forest method parameters of 100 trees, 4 spectral bands entering each tree, value of 1 for leaf population and 40% of training data used for each tree. For the land cover classification of Chancho and Dangora Morocho kebele with the same method settings, the overall accuracy came out to be 66.00 and 73.73% and kappa index of 0.545 and 0.601. For the classification of Chancho kebele, a different combination of parameters (80, 3, 1, 0.4) worked out better...

Thesis (M.Sc. (Geography)) -- University of Limpopo, 2020 / Wetland vegetation provides a variety of goods and services such as carbon sequestration, flood control, climate regulation, filtering contamination, improve and maintain water quality, ecological functioning. However, changes in land cover and uses, overgrazing and environmental changes have resulted in the transformation of the wetland ecosystem. So far, a lot of focus has been biased towards large wetlands neglecting wetlands at a local scale. Smaller wetlands continue to receive massive degradation by the surrounding communities.Therefore, this study seeks to assess and map wetland vegetation as an indicator of ecological productivity on a small scale. The Sentinel-2 MSI image was used to map wetland plant species diversity and above-ground biomass (AGB). Four key diversity indices; the Shannon Wiener (H), Simpson (D), Pielou (J), and Species richness (S) were used to measure species diversity. A multilinear regression technique was applied to establish the relationship between remotely sensed data and diversity indices and AGB. The results indicated that Simpson (D) has a high relationship with combined vegetation indices and spectral band, yielding the highest accuracy when compared to other diversity indices. For example, an R² of 0.75, and the RMSE of 0.08 and AIC of -191.6 were observed. Further, vegetation AGB was estimated with high accuracy of an R² of 0.65, the RMSE 29.02, and AIC of 280.21. These results indicate that Maungani wetland has high species abundance largely dominated by one species (Cyperus latifidius) and highly productive. The findings of this work underscore the relevance of remotely sensed to estimate and monitor wetland plant species diversity with high accuracy.

Příloha k disertační práci: Abstrakt v AJ (Mgr. Josef Laštovička) Abstract This dissertation thesis deals with the study of forest ecosystems in the central Europe with the time series of multispectral optical satellite data. These forest ecosystems have been influenced by biotic and abiotic disturbances for the last decade. The time series of the satellite data with high spatial resolution allow the detection and analysis of forest disturbances. This thesis is mainly focused primally on free available Landsat and Sentinel-2 data, these two data types were compared. From methods, the difference time series analyses / algorithms were used. The whole thesis can be divided into two main parts. The first one analyses usability of classifiers for detection of forest ecosystems with per-pixel and sub-pixel methods. Specifically, the Neural Network, the Support Vector Machine and the Maximum Likelihood per-pixel classifiers were used and compared for different types of data (for data with high spatial resolution - Landsat or Sentinel-2; very high spatial resolution - WorldView-2) and for classification of protected forest areas. The Support Vector Machine were selected as the most suitable method for forest classifications (with most accurate outputs) from the list of selected per-pixel classifiers. Also, Spectral...

Healthy forests and soils are crucial for the very existence of mankind as they provide food, clean water and air, shade and protection against floods and storms. With their photosynthetic carbon storage ability, they mitigate climate change and fertilise and stabilise soils. Unfortunately, deforestation and the loss of fertile soils are the bleak reality and among the world’s most pressing challenges. Over the past decades Kenya has faced severe deforestation, but efforts are being undertaken to reverse deforestation, revegetate degraded land and combat erosion. Satellite remote sensing technology becomes increasingly useful for vegetation monitoring as the data quality improves and the costs decrease. This thesis explores the potential of free open access Sentinel-2 data for vegetation monitoring through Random Forest land cover classification and post-classification change detection on Rusinga Island, Kenya. Different single-date and multi-temporal predictor datasets differentiating respectively between five and four classes were examined to develop the most suitable model. The classification achieved acceptable results when assessed on an independent test dataset (overall accuracy of 90.06% with five classes and 96.89% with four classes), which should however be confirmed on the ground and could potentially be improved with better reference data. In this study, change detection could only be analysed over a time frame of two years, which is too short to produce meaningful results. Nevertheless, the method was proven conceptually and could be applied in the future to monitor land cover changes on Rusinga Island.

Semi-natural grasslands are amongst the most biodiverse ecosystems in Europe, though their importance they are experiencing a declining trend. To monitor and assess the health of these ecosystems is generally costly, personnel demanding and time-consuming. With satellite imagery and machine learning becoming more accessible, this can offer a cheap and effective way to gain ecological information about semi-natural grasslands.This thesis explores the possibilities to predict plant species richness in semi-natural grasslands with high resolution satellite imagery through machine learning. Five different machine learning models were employed with various subsets of spectral- and geographical features to see how they performed and why. The study area was in southern Sweden with satellite and survey data from the summer of 2019.Geographical features were the features that influenced the machine learning models most. This can be explained by the geographical spread of the semi-natural grasslands, as well as difficulties in finding correlations in the relatively noisy satellite data. The most important spectral features were found in the red edge- and the short-wave infrared spectrums. These spectrums represent leaf chlorophyll content and water content in vegetation, respectively. The most accurate machine learning model was Random Forest when it was trained using with all the spectral- and geographical features. The other models; Logistic Regression, Support Vector Machine, Voting Classifier and Neural Network, showed general inabilities to interpret feature subsets containing the spectral data.This thesis shows that with deeper knowledge about the satellite-biodiversity relationship and how to apply it with machine learning have the possibilities of cheaper, more efficient and standardized monitoring of ecologically valuable areas such as semi-natural grasslands.

Monitoring carbon stock has emerged as a critical environmental problem among several worldwide organizations and collaborations in the context of global warming and climate change. This study seeks to provide a remote sensing solution based on three types of data, to explore the feasibility and reliability of estimating aboveground biomass (AGB) in order to improve the efficiency of monitoring carbon stock. The study attempted to investigate the potential of using Google Earth Engine (GEE), and the combinations of different datasets from Sentinel-1 (SAR), Sentinel-2 multispectral imagery, and LiDAR data to estimate AGB, by using the random forest algorithm (RF). Two models were proposed: the first one (Model 1) detected the AGB temporal changes from 2016 to 2021 in Southern Sweden; while the second one (Model 2) focused on Hultsfred municipality and studied the influence of different variables including the canopy height. Besides, six experimental groups of variables were tested to determine the performance of using different types of remote sensing data. We validated these two models with the observed AGB, and the findings showed that the combination of SAR polarization, multisprectral bands, vegetation indices able to estimate AGB for Model 1. In addition, Model 2 showed that further using the canopy height data can further improve the estimation.  We also found out that the spectral bands from Sentinel-2 contributed the most to AGB estimation for Model 1 in terms of: bands B3 (Green), B4 (Red), B5 (Red edge), B11 (SWIR), B12 (SWIR); and, vegetation indices of RVI, DVI, and EVI. On the other hand, for Model 2, B1(Ultra blue), B4 (Red), EVI, SAVI, and the canopy height are the most crucial variables for estimating AGB. Besides, the radar backscatter values using VV and VH modes from Sentienl-1 were both important for Models 1 and 2. For Model 1, the experimental group with the best accuracy was the group that used all variable combinations from Sentinel-1 and 2, and its   was 0.33~0.74. For Model 2, the group that used all the variables, in addition to the canopy height performed the best, where its   is 0.91. These therefore showed the benefit of integrating different remote sensing data sources.  In conclusion, this study showed the potential of using RF and GEE to estimate AGB in Southern Sweden. Furthermore, this study also shows the possibility of handling large dataset for a large scale area, at the resolution of 10 m, and producing time series AGB maps from 2016 to 2021. This can help enhance our understanding of AGB temporal changes and carbon stock detection in Southern Sweden, that can provide valuable insights for forest management and carbon monitoring.

Inland water has for a long time showed vast stress due to eutrophication, mainly caused by increased levels of phosphorus. Applying remote sensing as a tool for monitoring water parameters has long been used. In the past, inland watercourses measurements have proven to be challenging, often due to the limitations of satellite missions' spectral resolution or difficulties in implementing the appropriate methodology. This project investigates the potential to use a high-resolution satellite mission, Sentinel 2, to monitor phosphorus with the proxies total suspended matter (TSM) and turbidity in two smaller watercourses, Fyrisån and Sävjaån, in Uppsala, Sweden. From April to November, a period spanning three years (2018, 2019, and 2021), empirical modeling was employed to conduct investigations. The three years all represent different weather patterns and discharge velocities. The bands 2 to 8 were investigated individually and together to see if there was a potential using a single band correlation or multiple to correlate with turbidity or TSM. The two optically active water parameters are known to have a high correlation with the non-optically active phosphorus. There was no correlation found between the proxies and each band individually for any of the years investigated. Using a multi regression analysis both 2018 and 2019 showed high correlation for TSM, and 2019 for turbidity. While the results for 2021 were not significant for any of the proxies. The conclusion indicates that with right surrounding factors it’s possible to use TSM and turbidity as a proxy for phosphorus when using Sentinel 2 in these smaller watercourses. Nevertheless, further studies are needed to investigate how the proxy and the nutrient acts together with satellite data for peaks etc. before using Sentinel 2 results as a direct interpretation.

Stora arealer skog behöver övervakas. Att göra detta på ett kostnadseffektivt sätt är något som skogssektorn efterfrågar. Syftet med studien var att undersöka möjligheten att skatta skogliga variabler med satellitbilder från Sentinel 2. Korrelationen mellan granskogens uppmätta reflektans i satellitbilder från Sentinel 2 och uppmätta variablerna i fält har beräknats och analyserats. Resultatet visar att styrkan i korrelation skiljer sig mellan olika rumsliga upplösningar, vilken tid på året satellitbilderna är tagna, vilka spektrala band och vegetationsindex som används samt vilka skogliga variabler som avses uppskattas. Att använda enskilda satellitbilders värden från Sentinel 2 ger inte tillräckligt tillförlitliga data för att uppskatta skogliga variabler.

Granbarkborre är en av Sveriges mest destruktiva skadeinsekter som angriper granskog. Insekten har medfört förödande konsekvenser för granskog, framför allt sedan2018 där stora arealer granskog nästan har eliminerats. Insekten trivs i varmt ochtorrt klimat. Växthuseffekten i form av värmeböljor och perioder av minskad nederbörd tros gynna denna skadliga insekt då de kan fortplanta sig flera gånger och erövra nya områden under en enda sommarsäsong. En vital och nödvändig åtgärd vid bekämpning av skadeinsekter är att föra bort angripna träd innan granbarkborren lämnar barken. Dock är det nästan omöjligt attundersöka all granskog på det traditionella sättet, det vill säga till fots eftersom detär mycket tids- och resurskrävande. I det tidiga skedet visar det angripna trädet ingabetydande färgförändringar i det synliga spektrumet inom fjärranalys, vilket försvårar tidig upptäckt. Men för att försöka göra detta möjligt ämnar det här arbetet undersöka skillnaderna hos friska och angripna träd i tid, där det användes band i detosynliga spektrumet som ShortWave Infrared. Detta användes bland annat i form avbandkombinationer, som Atmospheric Penetration och Agriculture. Dessutom utfördes empiriska experiment på olika vegetationsindex (VI) som var NormalizedSimple Ratio, Enhanced Vegetation Index, Green Chlorophyll Vegetation Index,Normalized Difference Vegetation Index, Normalized Difference Moisture Indexoch Normalized Distance RED and SWIR. I denna studie användes satellitbilder från Sentinel-2 över studieområdet i Mellansverige under månaderna maj-september från 2020 till juli 2022. Inrapporterade dataför angrepp av granbarkborren i studieområdet hämtades från databasarkiven GlobalBiodiversity Information Facility och Holmen AB. Skogsstyrelsens öppna karttjänstanvändes för att erhålla data över Sveriges skogsarter, för att säkerställa att studieområdet bestod av granskog. Genom att utföra empiriska experiment av de olika VI och bandkombinationer sompresenteras i denna studie kunde några indikationer utmärkas. På grund av problematiken med de olika påverkande faktorerna, som bland annat lokalt klimat i kombination med tröskelvärden, var det svårt att fastställa en fullständig bedömning. Vårslutsats visar att de använda vegetationsindex och de två bandkombinationer tillsammans med den spatiala upplösningen, som Sentinel-2 erbjuder, inte uppnår det someftersträvas i denna studie. Anledningen till detta var att möjligheten att identifieraenstaka sjuka träd i studieområdet saknades. / The spruce bark beetle is one of Sweden's most destructive pests that attack spruceforests. The insect has had devastating consequences for spruce forests, especiallysince 2018 where large areas of spruce forest have been almost eliminated. The insect thrives in warm and dry climates. The greenhouse effect in the form of heatwaves and periods of reduced rainfall is believed to favor this harmful insect as theycan reproduce several times and conquer new areas in a single summer season. A vital and necessary measure in combating pests is to remove infested trees beforethe spruce bark beetle leaves the bark. However, it is almost impossible to examineall the spruce forest in the traditional way, which is on foot, because it is time- andresource-consuming. In the early stage, the infested tree shows no significant colorchanges in the visible spectrum in remote sensing, which makes early detection difficult. To try to make this possible, this work intends to investigate the differences inhealthy and infested trees in time, where bands in the invisible spectrum such asShortWave Infrared were used. This was used, among other things, in the form ofband combinations, such as Atmospheric Penetration and Agriculture. In addition,empirical experiments were performed on different vegetation indices (VI) whichwere Normalized Simple Ratio, Enhanced Vegetation Index, Green ChlorophyllVegetation Index, Normalized Difference Vegetation Index, Normalized DifferenceMoisture Index and Normalized Distance RED and SWIR. In this study, satellite images from Sentinel-2 were used over the study area in central Sweden during the months of May-September from 2020 to July 2022. Reported data for attacks by the spruce bark beetle in the study area were retrievedfrom the Global Biodiversity Information Facility and Holmen AB database archives.The Forestry Agency's open map service was used to obtain data on Sweden's forestspecies, to ensure that the study area consisted of spruce forest. By performing empirical experiments of the different VI and band combinationspresented in this study, some indications could be distinguished. Due to the problems with the various influencing factors, such as local climate in combination withthreshold values, it was difficult to establish a complete assessment. Our conclusionshows that the used vegetation indices and the two band combinations together withthe spatial resolution offered by Sentinel-2 do not achieve what is sought in thisstudy. The reason for this was that the possibility of identifying individual diseasedtrees in the study area was missing.

Remote sensing data has great potential for wildfire detection and monitoring with enhanced spatial resolution and temporal coverage. Earth Observation satellites have been employed to systematically monitor fire activity over large regions in two ways: (i) to detect the location of actively burning spots (during the fire event), and (ii) to map the spatial extent of the burned scars (during or after the event). Active fire detection plays an important role in wildfire early warning systems. The open-access of Sentinel-2 multispectral data at 20-m resolution offers an opportunity to evaluate its complementary role to the coarse indication in the hotspots provided by MODIS-like polar-orbiting and GOES-like geostationary systems. In addition, accurate and timely mapping of burned areas is needed for damage assessment. Recent advances in deep learning (DL) provides the researcher with automatic, accurate, and bias-free large-scale mapping options for burned area mapping using uni-temporal multispectral imagery. Therefore, the objective of this thesis is to evaluate multispectral remote sensing data (in particular Sentinel-2) for wildfire detection, including active fire detection using a multi-criteria approach and burned area detection using DL models.        For active fire detection, a multi-criteria approach based on the reflectance of B4, B11, and B12 of Sentinel-2 MSI data is developed for several representative fire-prone biomes to extract unambiguous active fire pixels. The adaptive thresholds for each biome are statistically determined from 11 million Sentinel-2 observations samples acquired over summertime (June 2019 to September 2019) across 14 regions or countries. The primary criterion is derived from 3 sigma prediction interval of OLS regression of observation samples for each biome. More specific criteria based on B11 and B12 are further introduced to reduce the omission errors (OE) and commission errors (CE).        The multi-criteria approach proves to be effective in cool smoldering fire detection in study areas with tropical &amp; subtropical grasslands, savannas &amp; shrublands using the primary criterion. At the same time, additional criteria that thresholds the reflectance of B11 and B12 can effectively decrease the CE caused by extremely bright flames around the hot cores in testing sites with Mediterranean forests, woodlands &amp; scrub. The other criterion based on reflectance ratio between B12 and B11 also avoids the effects of CE caused by hot soil pixels in sites with tropical &amp; subtropical moist broadleaf forests. Overall, the validation performance over testing patches reveals that CE and OE can be kept at a low level  (0.14 and 0.04) as an acceptable trade-off. This multi-criteria algorithm is suitable for rapid active fire detection based on uni-temporal imagery without the requirement of multi-temporal data. Medium-resolution multispectral data can be used as a complementary choice to the coarse resolution images for their ability to detect small burning areas and to detect active fires more accurately.        For burned area mapping, this thesis aims to expound on the capability of deep DL models for automatically mapping burned areas from uni-temporal multispectral imagery. Various burned area detection algorithms have been developed using Sentinel-2 and/or Landsat data, but most of the studies require a pre-fire image, dense time-series data, or an empirical threshold. In this thesis, several semantic segmentation network architectures, i.e., U-Net, HRNet, Fast- SCNN, and DeepLabv3+ are applied to Sentinel-2 imagery and Landsat-8 imagery over three testing sites in two local climate zones. In addition, three popular machine learning (ML) algorithms (LightGBM, KNN, and random forests) and NBR thresholding techniques (empirical and OTSU-based) are used in the same study areas for comparison.        The validation results show that DL algorithms outperform the machine learning (ML) methods in two of the three cases with the compact burned scars,  while ML methods seem to be more suitable for mapping dispersed scar in boreal forests. Using Sentinel-2 images, U-Net and HRNet exhibit comparatively identical performance with higher kappa (around 0.9) in one heterogeneous Mediterranean fire site in Greece; Fast-SCNN performs better than others with kappa over 0.79 in one compact boreal forest fire with various burn severity in Sweden. Furthermore, directly transferring the trained models to corresponding Landsat-8 data, HRNet dominates in the three test sites among DL models and can preserve the high accuracy. The results demonstrate that DL models can make full use of contextual information and capture spatial details in multiple scales from fire-sensitive spectral bands to map burned areas. With the uni-temporal image, DL-based methods have the potential to be used for the next Earth observation satellite with onboard data processing and limited storage for previous scenes.    In the future study, DL models will be explored to detect active fire from multi-resolution remote sensing data. The existing problem of unbalanced labeled data can be resolved via advanced DL architecture, the suitable configuration on the training dataset, and improved loss function. To further explore the damage caused by wildfire, future work will focus on the burn severity assessment based on DL models through multi-class semantic segmentation. In addition, the translation between optical and SAR imagery based on Generative Adversarial Network (GAN) model could be explored to improve burned area mapping in different weather conditions. / Fjärranalysdata har stor potential för upptäckt och övervakning av skogsbränder med förbättrad rumslig upplösning och tidsmässig täckning. Jordobservationssatelliter har använts för att systematiskt övervaka brandaktivitet över stora regioner på två sätt: (i) för att upptäcka placeringen av aktivt brinnande fläckar (under brandhändelsen) och (ii) för att kartlägga den brända ärrens rumsliga omfattning ( under eller efter evenemanget). Aktiv branddetektering spelar en viktig roll i system för tidig varning för skogsbränder. Den öppna tillgången till Sentinel-2 multispektral data vid 20 m upplösning ger en möjlighet att utvärdera dess kompletterande roll i förhållande till den grova indikationen i hotspots som tillhandahålls av MODIS-liknande polaromloppsbanesystem och GOES-liknande geostationära system. Dessutom krävs en korrekt och snabb kartläggning av brända områden för skadebedömning. Senaste framstegen inom deep learning (DL) ger forskaren automatiska, exakta och förspänningsfria storskaliga kartläggningsalternativ för kartläggning av bränt område med unitemporal multispektral bild. Därför är syftet med denna avhandling att utvärdera multispektral fjärranalysdata (särskilt Sentinel- 2) för att upptäcka skogsbränder, inklusive aktiv branddetektering med hjälp av ett multikriterietillvägagångssätt och detektering av bränt område med DL-modeller. För aktiv branddetektering utvecklas en multikriteriemetod baserad på reflektionen av B4, B11 och B12 i Stentinel-2 MSI data för flera representativa brandbenägna biom för att få fram otvetydiga pixlar för aktiv brand. De adaptiva tröskelvärdena för varje biom bestäms statistiskt från 11 miljoner Sentinel-2 observationsprover som förvärvats under sommaren (juni 2019 till september 2019) i 14 regioner eller länder. Det primära kriteriet härleds från 3-sigma-prediktionsintervallet för OLS-regression av observationsprover för varje biom. Mer specifika kriterier baserade på B11 och B12 införs vidare för att minska utelämningsfel (OE) och kommissionsfel (CE). Det multikriteriella tillvägagångssättet visar sig vara effektivt när det gäller upptäckt av svala pyrande bränder i undersökningsområden med tropiska och subtropiska gräsmarker, savanner och buskmarker med hjälp av det primära kriteriet. Samtidigt kan ytterligare kriterier som tröskelvärden för reflektionen av B11 och B12 effektivt minska det fel som orsakas av extremt ljusa lågor runt de heta kärnorna i testområden med skogar, skogsmarker och buskage i Medelhavsområdet. Det andra kriteriet som bygger på förhållandet mellan B12 och B11:s reflektionsgrad undviker också effekterna av CE som orsakas av heta markpixlar i områden med tropiska och subtropiska fuktiga lövskogar. Sammantaget visar valideringsresultatet för testområden att CE och OE kan hållas på en låg nivå (0,14 och 0,04) som en godtagbar kompromiss. Algoritmen med flera kriterier lämpar sig för snabb aktiv branddetektering baserad på unika tidsmässiga bilder utan krav på tidsmässiga data. Multispektrala data med medelhög upplösning kan användas som ett kompletterande val till bilder med kursupplösning på grund av deras förmåga att upptäcka små brinnande områden och att upptäcka aktiva bränder mer exakt. När det gäller kartläggning av brända områden syftar denna avhandling till att förklara hur djupa DL-modeller kan användas för att automatiskt kartlägga brända områden från multispektrala bilder i ett tidsintervall. Olika algoritmer för upptäckt av brända områden har utvecklats med hjälp av Sentinel-2 och/eller Landsat-data, men de flesta av studierna kräver att man har en förebränning. bild före branden, täta tidsseriedata eller ett empiriskt tröskelvärde. I den här avhandlingen tillämpas flera arkitekturer för semantiska segmenteringsnätverk, dvs. U-Net, HRNet, Fast- SCNN och DeepLabv3+, på Sentinel- 2 bilder och Landsat-8 bilder över tre testplatser i två lokala klimatzoner. Dessutom används tre populära algoritmer för maskininlärning (ML) (Light- GBM, KNN och slumpmässiga skogar) och NBR-tröskelvärden (empiriska och OTSU-baserade) i samma undersökningsområden för jämförelse. Valideringsresultaten visar att DL-algoritmerna överträffar maskininlärningsmetoderna (ML) i två av de tre fallen med kompakta brända ärr, medan ML-metoderna verkar vara mer lämpliga för kartläggning av spridda ärr i boreala skogar. Med hjälp av Sentinel-2 bilder uppvisar U-Net och HRNet jämförelsevis identiska prestanda med högre kappa (omkring 0,9) i en heterogen brandplats i Medelhavet i Grekland; Fast-SCNN presterar bättre än andra med kappa över 0,79 i en kompakt boreal skogsbrand med varierande brännskadegrad i Sverige. Vid direkt överföring av de tränade modellerna till motsvarande Landsat-8-data dominerar HRNet dessutom på de tre testplatserna bland DL-modellerna och kan bevara den höga noggrannheten. Resultaten visade att DL-modeller kan utnyttja kontextuell information fullt ut och fånga rumsliga detaljer i flera skalor från brandkänsliga spektralband för att kartlägga brända områden. Med den unika tidsmässiga bilden har DL-baserade metoder potential att användas för nästa jordobservationssatellit med databehandling ombord och begränsad lagring av tidigare scener. I den framtida studien kommer DL-modeller att undersökas för att upptäcka aktiva bränder från fjärranalysdata med flera upplösningar. Det befintliga problemet med obalanserade märkta data kan lösas med hjälp av en avancerad DL-arkitektur, lämplig konfiguration av träningsdatasetet och förbättrad förlustfunktion. För att ytterligare utforska de skador som orsakas av skogsbränder kommer det framtida arbetet att fokusera på bedömningen av brännskadornas allvarlighetsgrad baserat på DL-modeller genom semantisk segmentering av flera klasser. Dessutom kan översättningen mellan optiska bilder och SAR-bilder baserad på en GAN-modell (Generative Adversarial Network) undersökas för att förbättra kartläggningen av brända områden under olika väderförhållanden. / <p>QC 20210525</p>