Biomass burning in southern Africa has brought significant challenges to the research society as a fundamental driver of climate and land cover changes. Burned area mapping approaches have been developed that generate large-scale low and moderate resolution products made with different satellite data. This consequently afford the remote sensing community a unique opportunity to support their potential applications in e.g., examining the impact of fire on natural resources, estimating the quantities of burned biomass and gas emissions. Generally, the satellite-derived burned area products produced with dissimilar algorithms provide mapped burned areas at different levels of accuracy, as the environmental and remote sensing factors vary both spatially and temporally. This study focused on the inter-comparison and accuracy evaluation of the 500-meter Moderate Resolution Imaging Spetroradiomter (MODIS) burned area product (MCD45A1) and the Backup MODIS burned area product (hereafter BMBAP) across the main-fire prone South African biomes using reference data independently-derived from multi-temporal 30-meter Landsat 5 Thematic Mapper (TM) imagery distributed over six validation sites. The accuracy of the products was quantified using confusion matrices, linear regression and subpixel burned area measures. The results revealed that the highest burned area mapping accuracies were reported in the fynbos and grassland biomes by the MCD45A1 product, following the BMBAP product across the pine forest and savanna biomes, respectively. Further, the MCD45A1 product presented higher subpixel detection probabilities for the burned area fractions <= 50% than the BMBAP product, which appeared more reliable in detecting burned area fractions > 50% of a MODIS pixel. Finally the results demonstrated that the probability of identifying a burned area within a MODIS pixel is directly related to the proportion of the MODIS pixel burned and thus, highlights the relevance of fractional burned area during classification accuracy assessment of lower resolution remotely-sensed products using data with higher spatial resolution. / Dissertation (MSc)--University of Pretoria, 2011. / Geography, Geoinformatics and Meteorology / unrestricted

Study object. The research was made in two forest stands: one was affected by fire and another without fire damage. A sample plot was in a pine stand damaged by fire a year ago, which is is in Kaunas forest enterprise, in Kulautuva forest department. The stand is in square No. 169 and the plots numbers are 14 and 15. The damaged area takes 0,4 hectare. It is 100% pine stand with a site index Nbl. Study aim. To investigate the low fire influence on pine stand litter entomocomplex fluctuation: on their number and seasonal dynamic, on their distrubution by trophical groups, on species abundance and variety in a fire damaged stands. Methods. The research was made in 2004 since april till october. The Barberis ground trap was used in a research. The trap is made from a 150 ml volume and 6,5 cm diameter plastic cruet. The traps were digged into ground to the top of the trap. 1/3 of the trap was filled with a 10% formalin soak. In that case the insects were saved of decomposing and birds. The traps were placed in three zones: in burned area, chequerwise in the place between damaged and undamaged area, and in undamaged area. There were 6 satelites in each area holding 4 traps in each. All together it was 72 traps placed. The distance between the areas was 20 m and the distance between satelites – 10 m. The collected insects were dried and recognized in the labs of Lithuanian University of Agriculture and Forest Research Institute. Results. The total amount of collected insects in 6... [to full text]

Submitted by Cláudia Bueno (claudiamoura18@gmail.com) on 2015-12-04T17:06:29Z No. of bitstreams: 2 Tese - Fernando Moreira de Araújo - 2015.pdf: 6967405 bytes, checksum: 6a3a47f12667e66913e2ad2c041d7e21 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2015-12-07T11:21:38Z (GMT) No. of bitstreams: 2 Tese - Fernando Moreira de Araújo - 2015.pdf: 6967405 bytes, checksum: 6a3a47f12667e66913e2ad2c041d7e21 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2015-12-07T11:21:38Z (GMT). No. of bitstreams: 2 Tese - Fernando Moreira de Araújo - 2015.pdf: 6967405 bytes, checksum: 6a3a47f12667e66913e2ad2c041d7e21 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2015-03-03 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Fire is an important ecological and environmental disturbance phenomenon in the Cerrado biome, which is influenced by climatic, ecological, cultural and economic factors. Burning biomass promotes change in the structure and composition of the soil, accelerates the erosion and sedimentation processes, promotes seed germination, renewal of grasses and releases large amounts of trace gases and aerosols into the atmosphere. This research seeks to understand the location patterns and the recurrence of burned areas to the Cerrado biome according to their different geographical regions from the land cover and land use class, and climatic variables. In particular, we evaluated quantitatively the performance of the MODIS MCD45A1 product for the Cerrado biome, in accordance with the different land use and land cover classes depending on the landscape, using as reference 130 randomly selected burned area polygons, and four randomly selected medium resolution orbital images, both from the month of September. In addition, we analyzed the burned areas spatial and temporal patterns in Brazil, considering both the territorial limits of the six biomes (i.e. Amazônia, Caatinga, Cerado, Mata Atlântica, Pampa and Pantanal) and their main land cover and land use classes (PROBIO MMA). At last, for burned areas greater than 2 km², we evaluated their patterns (spatial and temporal) and frequency of recurrence (burned areas that intersect more than 50%), as well as their interanual and seasonal climatic variations from vegetation index (EVI), precipitation (TRMM) and evapotranspiration (MOD16) images. These images were analyzed in association with the four groups of recurrence (burned areas with recurrence every four years) to determine the different burning regimes and the pre and post burning vegetation responses. The MCD45A1 polygons (September) had good relationship with the reference mapping (Landsat base) (r² = 0.92) and also with the hotspots (MOD14 and MYD14 - monthly data) (r² = 0.89). The omission error found is high, but this is associated with landscape structure patterns, as found for the four reference images. The Product MCD45A1 even with spatial and temporal limitations promotes temporal mapping consistency and provides understanding the behavior and impacts of the burned on the Cerrado landscape. Among the Brazilian biomes, Cerrado is the one with the highest record of fires between 2002 and 2010 (73%), followed by Amazon (14%), while 81% of the records of the burned areas occur on the remnant vegetation cover class. The records of the burning recurrences in the Cerrado focus on the remnant class, 59% whereas the frequency of recurrence presented a higher proportion up to 3 times for the period 2000-2013. For the four groups of recurrences analyzed after the first fire record, 2002, the vegetative vigor records declines, especially after 2010, on average 4% for the remnant classes and 3.5% for the anthropogenic. The minimum distance between the fragments presents that 42% of recurrences analyzed are between 1-2 kilometers, providing greatest spatial concentration. / O fogo é um importante fenômeno ecológico e de perturbação ambiental no bioma Cerrado, o qual sofre influência de fatores climáticos, ecológicos, culturais e econômicos. A queima da biomassa promove a alteração da estrutura e composição do solo, acelera os processos erosivos e de sedimentação, promove a germinação de sementes, renovação das gramíneas e libera grandes quantidades de gases traços e aerossóis para a atmosfera. A presente pesquisa busca entender os padrões de localização e de recorrência das áreas queimadas para o bioma Cerrado, segundo suas diferentes regiões geográficas a partir das classes de uso da terra e cobertura vegetal remanescente e variáveis climáticas. Em específico busca-se avaliar quantitativamente a performance do produto MODIS MCD45A1 para o bioma Cerrado, de acordo com as distintas classes de uso e cobertura da terra em função da paisagem com base em 130 polígonos de áreas queimadas e 4 imagens orbitais de resolução espacial média, ambas amostras localizadas no mês de setembro. Outro interesse é de analisar os padrões espaciais e temporais da distribuição das áreas queimadas no território brasileiro considerando os limites territoriais dos biomas brasileiros e seus principais tipos de cobertura vegetal natural e antrópica (base PROBIO MMA), para tanto foi considerado os limites territoriais dos 6 biomas brasileiros (i.e. Amazônia, Caatinga, Cerrado, Mata Atlântica, Pampa e Pantanal). Por fim, avaliar os padrões (espaciais e temporais) e as frequências de recorrências das queimadas no Cerrado, com base em áreas queimadas >= 2km² e interseções >= 50% para as recorrências, bem como entender como a sazonalidade e variação interanual climática a partir dos dados de índice da vegetação (EVI), precipitação (TRMM) e evapotranspiração (MOD16) determinam os diferentes regimes de queima das várias classes de uso da terra e cobertura vegetal remanescente, para tanto foram analisados quatro grupos de recorrências com recortes temporais distintos. Os polígonos do produto MCD45A1 (setembro) apresentaram boa relação com o mapeamento de referência (base Landsat) (r² = 0.92) e também com os focos de calor (MOD14 e MYD14 - dados mensais) (r² = 0.89). O erro de omissão encontrado é alto, mas este está associado aos padrões de estrutura da paisagem, como o encontrado para as quatro imagens de referência. O Produto MCD45A1 mesmo apresentando limitações espaciais e temporais promove mapeamento com série temporal consistente, fornecendo entendimento do comportamento e os impactos do fogo sobre a paisagem do Cerrado. Entre os biomas brasileiros, o Cerrado é o que apresenta o maior registro de queimadas entre 2002 a 2010 (73%), seguido do Amazônico (14%), enquanto 81% dos registros das áreas queimadas ocorrem sobre as classes de cobertura vegetal remanescente. O registro das recorrências de queimadas no Cerrado se concentra sobre as classes remanescentes, 59%, já a frequência da recorrência apresenta maior proporção até 3 vezes para o período de 2000 a 2013. Para os quatro grupos de recorrências analisados, após o primeiro registro de fogo, 2002, o vigor vegetativo registra queda, principalmente após o ano de 2010, sendo em média 4% para as classes remanescentes e 3,5% para as antrópicas. A distância mínima entre os fragmentos mostra que 42% de recorrências analisadas estão entre 1 a 2 quilômetros, conferindo maior concentração espacial.

Wildfire activity has increased in recent years in the United States, endangering both environment and society. Appropriate management of this phenomenon is only achievable with a thorough understanding of the critical factors influencing wildfire activity in each region. In three essays, I use statistical and mathematical models to examine wildfires and propose solutions to mitigate their impact on society. In the first essay, I focused on building a systematic framework for modeling wildfire as a coupled human-natural system. I employ system dynamics modeling, which was previously applied in various fields, including healthcare, sustainability, and disaster mitigation. I show how, in the absence of exogenous factors such as temperature or lightning, the human perception of fire danger may establish a feedback loop that can yield significant trends such as fluctuation or even fluctuation with rising amplitude when linked with the natural system. This conclusion is counter-intuitive, given that the human contribution to wildfire is typically described in the literature using constant or semi-constant variables. Additionally, I analyzed the impact of three important fire protection measures on reducing burning rates (prescribed burning, enhancing immediate suppression accomplishment, and regulating the rate of WUI growth). The research concludes that appropriately integrating several policies can result in a synergistic effect that is greater than the sum of the effects of the individual policies. The second essay calibrates the model built in the first essay and examines wildfire trends across the contiguous United States. The simulation results closely match the real data, and the model serves as a foundation for data-driven policy research. To be more precise, I fit the model to each state separately and then compare the model's goodness of fit. Following that, I examine the influence of various policies and scenarios on wildfire behavior. In the scenario, I examine the effect of maintaining constant temperatures and precipitation levels relative to the average values for these variables over the last century. For the policy analysis, I examine the influence of three policies on each state (prescribed burning, increasing immediate suppression achievement, and regulating the rate of WUI development). Here, I provide state-specific suggestions about the primary factors that contribute to wildfires and the most effective policies for each state. In the third essay, I have implemented the Oregon wildfire history dataset and integrated it with two other aerial datasets, including meteorological data gathered by weather stations located around the state and counties. Then, using hierarchical modeling on over 10,000 wildfire ignitions, I developed a classification system for determining if a given fire has the potential to grow major or not. However, utilizing a huge dataset and a variety of resources presents several obstacles, such as the presence of missing data. I imputed the missing numbers using a sophisticated mathematical approach called "Predictive Mean Matching". / Doctor of Philosophy / Wildfire activity has increased in recent decades in the United States, which put many people in danger. Climate change, the Settlement of people in the Wildland Urban Interface, and an increase in vegetation density each play a role in this increase. In this dissertation, we discuss the wildfire in the United States in three essays. In the first essay, we develop a mathematical model to show how humans and nature affect wildfire activity in any area. We then test different major wildfire management policies on the hypothetical situation to compare the outcome of these policies together. In the second essay, we use the model developed in the essay (with some minor changes) to model the wildfire activity in 11 states of the U.S. which has the most wildfire activity in recent years. First, we show that our model can replicate the wildfire activity in each state. Second, we test the effect of wildfire mitigation policies on each state. This essay proposes state-specific policy recommendations based on the main reasons for the increase in wildfire activity in each state. Finally, in the third essay, we develop a statistical model to predict the existence of large wildfires in the next month in Oregon counties. We use climate, land, and fire history data to develop a warning system. Policymakers can use this system to move the fire suppression resources to counties with a high probability of experiencing large wildfires over the next month. Finally, all essays aim to enhance our understanding of the reasons for the increase in wildfire activity in recent years and suggest finding the appropriate way to deal with this change to reduce the effect of wildfire on human life.

Fires occur extensively across Zambia every year, a problem recognized as a major threat to biodiversity. Yet, basic tools for mapping at a spatial and temporal scale that provide useful information for understanding and managing this problem are not available. The objectives of this research were: to develop a method to map the spatio-temporal seasonal fire occurrence using satellite imagery, to develop a technique for estimating missing data in the satellite imagery considering the possibility of change in land cover over time, and to demonstrate applicability of these new tools by analyzing the fine-scale seasonal patterns of landscape fires in eastern Zambia. A new approach for mapping burned areas uses multitemporal image analysis with a fuzzy clustering algorithm to automatically select spectral-temporal signatures that are then used to classify the images to produce the desired spatio-temporal burned area information. Testing with Landsat data (30m resolution) in eastern Zambia showed accuracies in predicting burned areas above 92%. The approach is simple to implement, data driven, and can be automated, which can facilitate quicker production of burned area information. A profile-based approach for filling missing data uses multitemporal imagery and exploits the similarity in land cover temporal profiles and spatial relationships to reliably estimate missing data even in areas with significant changes. Testing with simulated missing data from an 8-image spectral index sequence showed highly correlated (R2 of 0.78-0.92) and precise estimates (deviations 4-7%) compared to actual values. The profile-based approach overcomes the common requirement of gap-filling methods that there is gradual or no change in land cover, and provides accurate gap-filling under conditions of both gradual and abrupt changes. The spatio-temporal progression of landscape burning was evaluated for the 2009 and 2012 fire seasons (June-November) using Landsat data. Results show widespread burning (~ 60%) with most fires occurring late (August-October) in the season. Fire occurrence and burn patch sizes decreased with increasing settlement density and landscape fragmentation reflecting human influences and fuel availability. Small fires (< 5ha) are predominant and were significantly under-detected (>50%) by a global dataset (MODIS Burned Area Product (500m resolution)), underscoring the critical need of higher geometric resolution imagery such as Landsat imagery for mapping such fine-scale fire activity. / Ph. D.

Bien que le feu soit reconnu comme un facteur d’influence dans la dynamique de végétation des savanes, son rôle n’est pas clairement défini. Cette thèse aborde le problème de l’étude de la relation entre l’usage des feux et la dynamique de végétation. L’approche choisie repose sur l’analyse de séries temporelles d’images de télédétection à moyenne résolution spatiale. Les savanes étudiées sont situées sur le bassin versant de Marovoay au nord-ouest de Madagascar. Dans la mesure où il n’existe pas de consensus quant aux méthodes à utiliser, les savanes de Madagascar offrent un contexte particulier, en raison de la dégradation très prononcée du couvert végétal et des changements recherchés, pour tester les méthodes existantes et en proposer des nouvelles. Le premier objectif de ce travail est d’identifier le régime des feux à travers le suivi des variations spatio-temporelles des surfaces brûlées en milieu de savane. Pour cela, une méthode de cartographie des surfaces brûlées a été développée : elle est basée sur le calcul d’un indicateur annuel indiquant le passage d’un feu pendant la saison sèche et d’un indicateur saisonnier traduisant la période de passage du feu. Cette méthode, appliquée au site d’étude, a permis de produire une série temporelle de données utilisées pour caractériser le régime des feux à partir de deux paramètres, la période d’occurrence et la fréquence de passage du feu. En parallèle, le deuxième objectif consiste à caractériser la dynamique de végétation par l’analyse des variations spatio-temporelles de l’activité végétale. Deux approches de détection des changements, basées sur le traitement de série temporelle de NDVI, ont été testées. La première repose sur l’analyse des variations inter annuelles d’un indicateur phénologique traduisant l’activité végétale pendant la phase de croissance des savanes. La deuxième utilise une technique de décomposition temporelle pour extraire la tendance d’une série de NDVI. Dans les deux cas, les résultats ont permis de caractériser la dynamique de végétation à travers trois classes d’évolution de l’activité végétale (séries progressive, régressive ou stable). Ces résultats ont été évalués par comparaison avec ceux issus de techniques de détection des changements basées sur l’analyse diachronique d’images à haute résolution spatiale. Enfin, dans la dernière étape du travail, nous avons étudié les relations entre les informations relatives aux régimes des feux et à la dynamique de végétation en utilisant des modèles de régression multivariée. L’objectif est d’estimer l’importance et le rôle du feu dans la dynamique de végétation. Les résultats ont amené à trois conclusions : a) Le feu est un facteur de maintien des savanes ; b) Dans les situations où la pression liée aux activités anthropiques est faible, le feu, en particulier par la fréquence de son usage, est un facteur déterminant de la dynamique de végétation ; c) Dans les autres situations, l’interprétation des résultats est complexe et difficile, très certainement en raison de l’interaction de multiples facteurs anthropiques. / Fire is recognized to be an essential factor that explains savanna vegetation dynamics. But its role is not clearly defined. This work investigates the problem of studying the relation between fire usage and vegetation dynamic. This is addressed through the analysis of time series of medium spatial resolution remotely sensed images. We studied the savanna localized on the Marovoay watershed, on the northwest part of Madagascar. As no consensus exists on the adapted methods, the savanna of Madagascar offers a particular context to test existing methods or develop new ones, because of the advanced vegetation cover degradation and the nature of change to be detected. The first objective of this work is to identify fire regime by analyzing the spatio-temporal variations of burned areas in savanna areas. To this end, a burned area mapping method was developed. It is based on the definition of an annual indicator, which indicates the occurrence of a fire during the dry season, and a seasonal indicator, which gives the information on the date of the fire event. Applied to the study site, this method has lead to the production of a time series of data used for the characterization of the fire regime through two parameters, the period of occurrence and the frequency of fire. In parallel, the second objective consists on characterizing vegetation dynamic by monitoring spatiotemporal variations of vegetation activity. Two change detection approaches based on NDVI time series, were tested. The first consists on analyzing the inter annual variations of a phenological indicator related to vegetation activity during the active growth season. The second uses a temporal decomposition technique to extract the trend from an NDVI time series. In both cases, the vegetation dynamic is characterized through three classes linked to the evolution of the vegetation activity (progressive, regressive or stable series). Results were evaluated by a comparison with the results obtained from a diachronic change detection technique based of high spatial resolution images. Finally, in the last part of this work, we investigated the relation between fire regimes and vegetation activity change classes using multivariate regression models. The objective is to analyze to determine the importance and the role of fire into the vegetation dynamic. Results leaded to three conclusions: a) Fire is a factor that maintains savanna; b) In areas where pressure due to anthropogenic activities is low, fire frequency represents a determinant factor to explain vegetation dynamic; c) In others areas, the interpretation of results appears to be complex and difficult, probably because of the high level of interactions between multiple environmental factors.

Les incendies sont des perturbations intrinsèques au fonctionnement des écosystèmes agissant sur la distribution des espèces, les cycles biogéochimiques et les interactions biosphère/atmosphère. La distribution spatiale et temporelle des surfaces brûlées résulte principalement d'un compromis entre la disponibilité en biomasse combustible et son état hydrique. Ainsi, caractériser et modéliser un régime d'incendie repose donc sur une évaluation principalement de la saisonnalité et l'intensité de la sécheresse. Dans les écosystèmes fortement soumis aux incendies comme les savanes, biome le plus affecté au niveau global, les informations disponibles ne sont issues que des observations issues de la télédétection globale. Malgré que ces types de données ont été largement utilisés depuis les années 2000 pour la quantification des émissions de carbone par les modèles globaux, les incertitudes restent grandes entre les différents produits mis à la disposition des scientifiques. De même les processus utilisés dans la modélisation dynamique des surfaces incendiées, couplés aux modèles de végétation repose sur des équations empiriques de quantification de la sécheresse. Pour des écosystèmes comme les savanes du Brésil, où les informations incendies sont uniquement obtenues par les informations issues de la télédétection globale, et où la variabilité des types fonctionnels de la végétation affecte le bilan hydrique des sols, évaluer les incertitudes sur ces variables est un pré requis important pour une modélisation appropriée des processus. Cette thèse aborde donc la caractérisation de la variabilité saisonnière des incendies et ses incertitudes dans les différents biomes du Brésil à partir des principales sources d'information sur les surfaces brulées issues de la télédétection globale (GFED4, GFED4s, MODIS MCD45A1, ESAFIRECCI) et des principaux indicateurs de sécheresse développés par les différents services forestiers. La même analyse est aussi appliquée aux surfaces incendiées issues des principaux modèles globaux de dynamique de végétation (DGVM) couplés avec un modèle incendie, afin d'identifier si les relations feux/sécheresse observées sur les produits satellitaires étaient conservées dans le modèles globaux. Après une analyse des processus utilisés dans les DGVMs, et notamment les nouveaux développements sur la caractérisation des patches de feux, la deuxième partie de mon travail a cherché à évaluer la capacité des produits globaux issus de la télédétection à méso-échelle (500m) à conserver la morphologie des patchs de feux observés à résolution plus fine pour une utilisation dans la caractérisation du régime des incendies et l'évaluation des DGVMs. La dernière partie de ce travail utilise une approche de biogéographie fonctionnelle basé sur les traits morphologiques des feux pour décrire un assemblage d'événements incendies et les comparer entre différents produits de télédétection pour en évaluer leurs similitudes ou leurs différences, ou entre différents biomes pour une nouvelle approche de pyrogéographie comparative / Fires are intrinsic disturbances in ecosystems functioning, and affecting species distribution, biogeochemical cycles and biosphere/atmosphere interactions. The spatial and temporal pattern of burnt areas is mainly the consequence of a compromise between fuel biomass availability and its water status. Characterizing and modelling a fire regime then relies on a reliable spatial and temporal evaluation of mainly seasonality and intensity of drought. In fire-prone ecosystems as savannas, the most fire affected biome globally, the only available information on burnt areas are derived from global remote sensing. Despite this information has been widely used for a global assessment of carbon emissions from fires since 2000, uncertainties remain important among data sources available for the scientific community. As well, processes used in the dynamic modelling of burned areas, embedded in vegetation models, are based on empirical equations quantifying the seasonal pattern of drought. For ecosystems like Brazilian savannas where fire information only rely on global remote sensing products, and where the variability in vegetation functional types affects the soil water budget, evaluating uncertainties on these variables is a prerequisite for an adequate modelling of processes. This thesis then aimed at characterizing the fire seasonality and the associated uncertainties in the different Brazilian biomes based on the main sources of global burned area (GFED4, GFED4s, MODIS MCD45A1, ESA FIRECCI) and a panel of drought indices developed by forest services worldwide. This thesis also investigated the same relationships analysis but with burned areas derived from the major dynamic global vegetation models (DGVM) coupled with fire modules, in order to test whether the burned area/drought relationships based on remote sensing are conserved in global models. After a review of the main processes computed in DGVMs, and particularly the new developments on the characterization of fire patches, the second part of this research aimed at evaluating the ability of global remote sensing products at 500m resolution to conserve patch morphologies captured at finer resolutions, for better characterizing fire regimes and evaluate DGVMs performances. The last part of this work used a functional biogeography approach based on fire morphological traits to describe an assemblage of fire events and to compare them between different remote sensing products, in order to evaluate their similarities and discrepancies, or between biomes for a new approach of comparative pyrogeography.

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>

I flera års tid har antalet svåra och okontrollerade skogsbränder ökat i antal. Det finns ett behov av att detektera skogsbränder med hjälp av satelliter som har högre tidsupplösning samt högre geometrisk upplösning än de satelliter som är i bruk idag. I detta examensarbete utfördes dels en studie av bildframställningen hos, och dels designen av omloppsbanan till en mikrosatellitkonstellation, som använder sig av nära-infraröd (NIR) och kortvågig infraröd (SWIR) avbildningsteknik med förstärkt geometrisk upplösning och tidsupplösning för övervakning av skogsbränder. En översikt av den nuvarande tekniken visade olika alternativ för kamera: två kommersiella sensorer med ett spektralområde på 0.7 till 1.7 μm och geometrisk upplösning på 140 m respektive 112 m; en kommersiell sensor med ett spektralområde på upp till 2.2 μm och geometrisk upplösning på 168 m; samt ett förslag till sensor med högre geometrisk upplösning på 20 m eller 50 m, som erhålls genom att öka brännvidden. Flera slutsatser nåddes med hänsyn till avbildningen: lämpligheten hos linserna som hittades till respektive sensor verifierades, och det upptäcktes att förvrängningen till följd av jordens rotation ökade med förlängd exponeringstid, vilket även skedde för förhållandet mellan signal och störningar. Ett förslag till en cirkulär, solsynkron, polär omloppsbana med ett dagligen upprepande mönster lades fram. En förenklad metod för beräkning av halva storaxeln tillämpades, och därmed bestämdes en höjd på 561 km och lutning på 97.64°. I enlighet med detta uppskattades antalet satelliter som krävs för både global och regional (Sverige) täckning för alla avbildningsalternativ. För global täckning beräknades antalet satelliter som krävs för att uppnå en geometrisk upplösning på 140.25 m till 15 stycken, medan en geometrisk upplösning på 50 m krävde 84 stycken satelliter. För regional täckning (Sverige) krävdes å andra sidan endast 6 stycken satelliter för att uppnå en geometrisk upplösning på 140.25 m, och 32 stycken för en geometrisk upplösning på 50 m. Detta examensarbete har visat hur många satelliter som behövs för antingen global eller regional täckning så att skogsbränder kan detekteras med högre geometrisk upplösning samt högre tidsupplösning. / For several years, the occurrence of more severe and uncontrolled wildfires has been increasing. There is a need to detect wildfires with a higher spatial and temporal resolution than the ones currently provided by operational satellites. In this thesis, an imager study and orbital design of a microsatellite constellation using near infrared (NIR) and shortwave infrared (SWIR) imagers for wildfire monitoring, with enhanced spatial and temporal resolution, was conducted. After a state of the art review, different alternatives of imaging systems were discerned: two commercial sensors with spectral range of 0.7 to 1.7 μm and spatial resolution of 140 m and 112 m, respectively; one commercial sensor with spectral range up to 2.2 μm and spatial resolution of 168 m; and one sensor proposal with higher spatial resolution of 20 m or 50 m, achieved by increasing the focal length. Several conclusions were reached with regards to the imagers: the appropriateness of lenses found for each sensor was confirmed, the Earth rotation distortion was found to increase as the exposure time is extended, as did the signal-to noise ratio. A proposal for a circular sun-synchronous polar orbit with a daily repeating pattern was made. Applying a simplified method to calculate the semi-major axis, an altitude of 561 km and inclination of 97.64° were determined. Accordingly, the number of satellites for both global and regional (Sweden) coverage was estimated for all imager alternatives. For global coverage, the necessary number of satellites to achieve a spatial resolution of 140.25 m was calculated to be 15 satellites, whereas for a spatial resolution of 50 m the number of satellites increased to 84. On the other hand, for regional coverage (Sweden), the number of satellites to achieve a spatial resolution of 140.25 m were 6, and for a spatial resolution of 50 m the number of satellites was 32. This thesis shows how many satellites are required for either global or regional coverage, considering different imager configurations, to detect wildfires with a higher spatial and temporal resolution.