The thesis presents the main points of standardization of remote sensing (RS) usage for Forensic Ecotechnique: forest and trees (FEft). Based on the study FEft expertise database, the types of expertise, which remote sensing data have already been used and the types where the use of remote sensing increases the exactness of solutions were analyzed. RS data are essential in a problematic situation when an examined object does not exist or has been changed at the time of the assignment. This situation may be solved by obtaining information about the investigated object via retrospective image data, as described in the model example. There are chosen categories of image data relevant to the needs of experts in this thesis. These categories of RS data and their products are described with an indication of their possible use in the expertise processing, including freely available or paid resources and archives of image data.

Land use and land cover changes (LULCC) including deforestation for agricultural land and others are elements that contribute on global environmental change. Therefore understanding a trend of these changes in the past, current, and future is important for making proper decisions to develop in a sustainable way. This study analyzed land use and land cover (LULC) changes over time for Tuy Duc district belonging to Dak Nong province based on LULC maps classified from a set of multidate satellite images captured in year 2003, 2006, 2009, and 2013 (SPOT 5 satellite images). The LULC spatio-temporal changes in the area were classified as perennial agriculture, cropland, residential area, grassland, natural forest, plantation and water surface. Based on these changes over time, potential LULC in 2023 was predicted using Cellular Automata (CA)–Markov model. The predicted results of the change in LULC in 2023 reveal that the total area of forest will lose 9,031ha accounting of 50% in total area of the changes. This may be mainly caused by converting forest cover to agriculture (account for 28%), grassland (12%) and residential area (9%). The findings suggest that the forest conversion needs to be controlled and well managed, and a reasonable land use plan should be developed in a harmonization way with forest resources conservation. / Thay đổi sử dụng đất và thảm phủ (LULCC) bao gồm cả việc phá rừng để phát triển nông nghiệp và vì các mục đích khác là tác nhân đóng góp vào biến đổi môi trường toàn cầu. Vì vậy hiểu biết về khuynh hướng của sự thay đổi này trong quá khứ, hiện tại và tương lai là quan trọng để đưa ra những quyết định dúng đắn để phát triển bền vững. Nghiên cứu đã phân tích LULCC trong thời gian qua dựa vào các bản đồ sử dụng đất và thảm phủ (LULC) đã được phân loại từ một loạt ảnh vệ tinh đa phổ được thu chụp vào năm 2003, 2006, 2009 (ảnh SPOT 5). Những thay đổi LULC theo thời gian và không gian trong khu vực được phân loại thành đất nông nghiệp với cây dài ngày, cây ngắn ngày, thổ cư, trảng cỏ cây bụi, rừng tự nhiên, rừng trồng và mặt nước. Dựa trên sự thay đổi này theo thời gian, LULC tiềm năng cho năm 2023 đã được dự báo bằng cách sử dụng mô hình CAMarkov. Kết quả dự báo LULCC năm 2023 đã cho thấy tổng diện tích rừng bị mất khoảng 9,031 ha chiếm 50% trong tổng số diện tích thay đổi. Điều này chủ yếu là do chuyển đổi từ rừng tự nhiên sang canh tác nông nghiệp (chiếm 28%), trảng cỏ cây bụi (12%) và khu dân cư (9%). Kết quả cho thấy việc chuyển đổi rừng cần phải được kiểm soát và quản lý tốt và một kế hoạch sử dụng đất hợp lý cần được xây dựng trong sự hài hòa với bảo tồn tài nguyên rừng.

An investigation into using attention mechanisms for better feature extraction in wildfire spread prediction models. This research examines the U-net architecture to achieve image segmentation, a process that partitions images by classifying pixels into one of two classes. The deep learning models explored in this research integrate modern deep learning architectures, and techniques used to optimize them. The models are trained on 12 distinct observational variables derived from the Google Earth Engine catalog. Evaluation is conducted with accuracy, Dice coefficient score, ROC-AUC, and F1-score. This research concludes that when augmenting U-net with attention mechanisms, the attention component improves feature suppression and recognition, improving overall performance. Furthermore, employing ensemble modeling reduces bias and variation, leading to more consistent and accurate predictions. When inferencing on wildfire propagation at 30-minute intervals, the architecture presented in this research achieved a ROC-AUC score of 86.2% and an accuracy of 82.1%.

South Africa’s growing elephant populations are concentrated in relatively small enclosed protected areas resulting in the over utilisation of the available food sources. Elephants and other herbivores as well as other natural disturbances such as fires and droughts play an important role in maintaining savannah environments. When these disturbances become too concentrated in a particular area the vegetation composition may be negatively affected. Excessive damage to the vegetation would result from exceeding the capacity of a protected area to provide food resources. The effect of the 120 elephants on the vegetation of Welgevonden Private Game Reserve, is not known. The rugged terrain of this reserve makes it a difficult, time consuming and labour intensive exercise to conduct ground studies. Satellite images can be used as a monitoring tool for vegetation change and improve the quantity and quality of environmental data to be collected significantly, allowing more informed management decision-making. This study evaluated the use of satellite imagery for monitoring elephant induced vegetation change on Welgevonden Private Game Reserve. The LANDSAT Thematic Mapper multispectral images, acquired at two yearly intervals from 1993 until 2007 were used. However, no suitable images were available for the years 1997, 2001 and 2003. A series of vegetation change maps was produced and the distribution of water sources and fire occurrences mapped. The areas of change were then correlated with the spatial distribution of water points and fire occurances, with uncorrelated areas of change. This was analysed using large animal population trends, weather data and management practices. On the visual comparison of the vegetation maps, it was seen that over this time period there was some decrease and thinning of woodland, but the most notable change was the increase of open woodland and decrease in grasslands. Using only the digital change detection for the period 1993 to 2007, a general increase in vegetation cover is seen. But this generalisation is misleading, since comparing the digital change detection to the vegetation maps indicates that while vegetation cover may have increased, significant changes occurred in the vegetation types. Most of the areas of significant change that were identified showed a strong positive correlation with burnt areas. The distribution of the water sources could not be directly linked to the vegetation change although rainfall fluctuations seemed to have accelerated vegetation changes. Years with high game counts, such as 1999, also coincide with very low rainfall making it difficult to differentiate between the effects of heavy utilisation of vegetation and low rainfall. Furthermore, many of the initial vegetation changes could be the result of land use changes due to the introduction of browsers, selective grazers and elephants that allow for more natural utilisation of the vegetation. Remote sensing makes it possible to successfully track changes in vegetation and identify areas of potential elephant induced vegetation change. Vegetation changes caused by disturbances, such as fire and anthropogenic activities, can be accounted for but it is not possible to conclude with a high level of certainty that the further changes seen are solely a result of elephant damage. Further work is required to reliably isolate elephant induced vegetation changes, as well as to establish the effects these changes have on the ecosystem as a whole. / Environmental Sciences / (M. Sc. (Environmetal Sciences))

I conducted a 12-month study of near-shore habitats in Lake Pontchartrain to assess spatiotemporal variation in the abundance of early life stage blue crabs (Callinectes sapidus). Collections were made using a 1 m2 throw trap and data showed that C. sapidus numbers varied over time and among sites. Two recruitment events occurred during the study. During the first recruitment in May-June, C. sapidus entered Lake Pontchartrain via the Inner Harbor Navigational Canal. In September-October, C. sapidus entered the Lake Pontchartrain via the Rigolets and Chef passes. My data suggest that C. sapidus utilize water circulation within the Lake Pontchartrain as a means of transportation throughout the estuary. MODerate-resolution Imaging Spectroradiometer (MODIS) 250 m data were analyzed to gain a large-scale view of suspended sediments patterns within Lake Pontchartrain and quantify water movement. Field sampling along with remote sensing proved to be beneficial when assessing estuarine-wide C. sapidus post-larval dispersal processes.

Given satellite images with accompanying pixel classifications and elevation data, we propose different solutions to object detection. The first method uses hierarchical clustering for segmentation and then employs different methods of classification. One of these classification methods used domain knowledge to classify objects while the other used Support Vector Machines. Additionally, a combination of three Support Vector Machines were used in a hierarchical structure which out-performed the regular Support Vector Machine method in most of the evaluation metrics. The second approach is more conventional with different types of Convolutional Neural Networks. A segmentation network was used as well as a few detection networks and different fusions between these. The Convolutional Neural Network approach proved to be the better of the two in terms of precision and recall but the clustering approach was not far behind. This work was done using a relatively small amount of data which potentially could have impacted the results of the Machine Learning models in a negative way.

South Africa’s growing elephant populations are concentrated in relatively small enclosed protected areas resulting in the over utilisation of the available food sources. Elephants and other herbivores as well as other natural disturbances such as fires and droughts play an important role in maintaining savannah environments. When these disturbances become too concentrated in a particular area the vegetation composition may be negatively affected. Excessive damage to the vegetation would result from exceeding the capacity of a protected area to provide food resources. The effect of the 120 elephants on the vegetation of Welgevonden Private Game Reserve, is not known. The rugged terrain of this reserve makes it a difficult, time consuming and labour intensive exercise to conduct ground studies. Satellite images can be used as a monitoring tool for vegetation change and improve the quantity and quality of environmental data to be collected significantly, allowing more informed management decision-making. This study evaluated the use of satellite imagery for monitoring elephant induced vegetation change on Welgevonden Private Game Reserve. The LANDSAT Thematic Mapper multispectral images, acquired at two yearly intervals from 1993 until 2007 were used. However, no suitable images were available for the years 1997, 2001 and 2003. A series of vegetation change maps was produced and the distribution of water sources and fire occurrences mapped. The areas of change were then correlated with the spatial distribution of water points and fire occurances, with uncorrelated areas of change. This was analysed using large animal population trends, weather data and management practices. On the visual comparison of the vegetation maps, it was seen that over this time period there was some decrease and thinning of woodland, but the most notable change was the increase of open woodland and decrease in grasslands. Using only the digital change detection for the period 1993 to 2007, a general increase in vegetation cover is seen. But this generalisation is misleading, since comparing the digital change detection to the vegetation maps indicates that while vegetation cover may have increased, significant changes occurred in the vegetation types. Most of the areas of significant change that were identified showed a strong positive correlation with burnt areas. The distribution of the water sources could not be directly linked to the vegetation change although rainfall fluctuations seemed to have accelerated vegetation changes. Years with high game counts, such as 1999, also coincide with very low rainfall making it difficult to differentiate between the effects of heavy utilisation of vegetation and low rainfall. Furthermore, many of the initial vegetation changes could be the result of land use changes due to the introduction of browsers, selective grazers and elephants that allow for more natural utilisation of the vegetation. Remote sensing makes it possible to successfully track changes in vegetation and identify areas of potential elephant induced vegetation change. Vegetation changes caused by disturbances, such as fire and anthropogenic activities, can be accounted for but it is not possible to conclude with a high level of certainty that the further changes seen are solely a result of elephant damage. Further work is required to reliably isolate elephant induced vegetation changes, as well as to establish the effects these changes have on the ecosystem as a whole. / Environmental Sciences / (M. Sc. (Environmetal Sciences))

The purpose of this research is to study the detection of temporal changes between two (or more) multimodal images satellites, i.e., between two different imaging modalities acquired by two heterogeneous sensors, giving for the same scene two images encoded differently and depending on the nature of the sensor used for each acquisition. The two (or multiple) multimodal satellite images are acquired and coregistered at two different dates, usually before and after an event. In this study, we propose new models belonging to different categories of multimodal change detection in remote sensing imagery. As a first contribution, we present a new constraint scenario expressed on every pair of pixels existing in the before and after image change. A second contribution of our work is to propose a spatio-temporal textural gradient operator expressed with complementary norms and also a new filtering strategy of the difference map resulting from this operator. Another contribution consists in constructing an observation field from a pair of pixels and to infer a solution maximum a posteriori sense. A fourth contribution is proposed which consists to build a common feature space for the two heterogeneous images. Our fifth contribution lies in the modeling of patterns of change by anomalies and on the analysis of reconstruction errors which we propose to learn a non-supervised model from a training base consisting only of patterns of no-change in order that the built model reconstruct the normal patterns (non-changes) with a small reconstruction error. In the sixth contribution, we propose a pairwise learning architecture based on a pseudosiamese CNN network that takes as input a pair of data instead of a single data and constitutes two partly uncoupled CNN parallel network streams (descriptors) followed by a decision network that includes fusion layers and a loss layer in the sense of the entropy criterion. The proposed models are enough flexible to be used effectively in the monomodal change detection case. / Cette recherche a pour objet l’étude de la détection de changements temporels entre deux (ou plusieurs) images satellitaires multimodales, i.e., avec deux modalités d’imagerie différentes acquises par deux capteurs hétérogènes donnant pour la même scène deux images encodées différemment suivant la nature du capteur utilisé pour chacune des prises de vues. Les deux (ou multiples) images satellitaires multimodales sont prises et co-enregistrées à deux dates différentes, avant et après un événement. Dans le cadre de cette étude, nous proposons des nouveaux modèles de détection de changement en imagerie satellitaire multimodale semi ou non supervisés. Comme première contribution, nous présentons un nouveau scénario de contraintes exprimé sur chaque paire de pixels existant dans l’image avant et après changement. Une deuxième contribution de notre travail consiste à proposer un opérateur de gradient textural spatio-temporel exprimé avec des normes complémentaires ainsi qu’une nouvelle stratégie de dé-bruitage de la carte de différence issue de cet opérateur. Une autre contribution consiste à construire un champ d’observation à partir d’une modélisation par paires de pixels et proposer une solution au sens du maximum a posteriori. Une quatrième contribution est proposée et consiste à construire un espace commun de caractéristiques pour les deux images hétérogènes. Notre cinquième contribution réside dans la modélisation des zones de changement comme étant des anomalies et sur l’analyse des erreurs de reconstruction dont nous proposons d’apprendre un modèle non-supervisé à partir d’une base d’apprentissage constituée seulement de zones de non-changement afin que le modèle reconstruit les motifs de non-changement avec une faible erreur. Dans la dernière contribution, nous proposons une architecture d’apprentissage par paires de pixels basée sur un réseau CNN pseudo-siamois qui prend en entrée une paire de données au lieu d’une seule donnée et est constituée de deux flux de réseau (descripteur) CNN parallèles et partiellement non-couplés suivis d’un réseau de décision qui comprend de couche de fusion et une couche de classification au sens du critère d’entropie. Les modèles proposés s’avèrent assez flexibles pour être utilisés efficacement dans le cas des données-images mono-modales.

<div>This dissertation addresses the problem of how to design a convolutional neural network (CNN) for giving semantic labels to the points on the ground given the satellite image coverage over the area and, for the ground truth, given the noisy labels in OpenStreetMap (OSM). This problem is made challenging by the fact that -- (1) Most of the images are likely to have been recorded from off-nadir viewpoints for the area of interest on the ground; (2) The user-supplied labels in OSM are frequently inaccurate and, not uncommonly, entirely missing; and (3) The size of the area covered on the ground must be large enough to possess any engineering utility. As this dissertation demonstrates, solving this problem requires that we first construct a DSM (Digital Surface Model) from a stereo fusion of the available images, and subsequently use the DSM to map the individual pixels in the satellite images to points on the ground. That creates an association between the pixels in the images and the noisy labels in OSM. The CNN-based solution we present yields a 4-8% improvement in the per-class segmentation IoU (Intersection over Union) scores compared to the traditional approaches that use the views independently of one another. The system we present is end-to-end automated, which facilitates comparing the classifiers trained directly on true orthophotos vis-`a-vis first training them on the off-nadir images and subsequently translating the predicted labels to geographical coordinates. This work also presents, for arguably the first time, an in-depth discussion of large-area image alignment and DSM construction using tens of true multi-date and multi-view WorldView-3 satellite images on a distributed OpenStack cloud computing platform.</div>