Drought, being one the most complex natural hazards, has a significant impact on society. To mitigate the impact and risk, it is crucial to be able to forecast drought, which is a challenging task. Nowadays, with technology innovations, large amounts of remote sensing data is available on the cloud. Meanwhile,machine learning and data-driven techniques have proven to be reliable data processing methods in a number of fields. In this study, the author aims to predict agricultural drought in Sweden with machine learning (ML) models. Several ML models, including random forest, decision tree, linear regression, support vector machine, ARIMA, ANN, and CNN, are employed to find out the best performing model. Seven hypothesized factors are used and tested with RFE for features analysis. Three data arrangement methods are explored for the best possible way to arrange the dataset. In the result section, the author concludes that soil moisture is the most important feature, ARIMA and random forest models are the most reliable algorithms, and the temporal method is more suitable for short-term predicting. / Torka, som är en av de mest komplexa naturliga riskerna, har en betydande inverkan på samhället. För att mildra påverkan och risken är det avgörande att kunna förutse torka, vilket är en utmanande uppgift. Nuförtiden, med tekniska innovationer, finns stora mängder fjärranalysdata tillgänglig på molnet. Under tiden har maskininlärning och datadrivna tekniker visat sig vara tillförlitliga databehandlingsmetoder inom ett antal områden. I denna studie syftar författaren till att förutsäga jordbrukstorka i Sverige med modeller för maskininlärning (ML). Flera ML-modeller, inklusive slumpmässig skog, beslutsträd, linjär regression, stödvektormaskin, ARIMA, ANN och CNN, används för att ta reda på den bästa modellen. Sju hypotesfaktorer används och testas med RFE för funktionsanalys. Tre dataarrangemangsmetoder utforskas för bästa möjliga sätt att ordna datamängden. I resultatavsnittet drar författaren slutsatsen att markfuktighet är den viktigaste egenskapen, ARIMA och slumpmässiga skogsmodeller är de mest tillförlitliga algoritmerna, och den tidsmässiga metoden är mer lämpad för korttidsförutsägelse.

The seagrass extent has been declining globally. The human activities that are most likely to cause seagrass loss are those which affect the water quality and clarity. However, turbidity following coastal erosion is often left out from marine ecosystem vulnerability indices. This study quantified the coastal erosion for Tsimipaika Bay in northwestern Madagascar by using change detection analysis of satellite imageries. The annual coastal erosion data was then used to create an index for seagrass vulnerability to turbidity following coastal erosion. Considering that the height of seagrass species plays an important role in their survival following turbidity, the seagrass vulnerability index (SVI) was based on two factors; seagrass species height and their distance to the nearest possible erosion place. The results for the coastal erosion showed that the amount of erosion was particularly high in 1996, 2001 and 2009 for Tsimipaika Bay. The highest erosion occurred in 2001 with a land loss area of about 6.2 km2 . The SVI maps revealed that 40% of the seagrass communities had minimum mean SVI values in 2001 and 50% had the maximum mean SVI during the year 2009. This study showed that it is possible to use coastal erosion to measure seagrass vulnerability; however, the index requires configuration such as including the total amount of annual coastal erosion and incorporating bathymetric data. The entire project was built and automated in Jupyter Notebook using Python programming language, which creates a ground for future studies to develop and modify the project.

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...

Remote sensing has enabled unprecedented earth observation from space and has proven to be an invaluable tool for agricultural applications and crop management practices. Here we detect seasonal metrics indicating the start of the season (SOS), the end of the season (EOS) and maximum greenness (MAX) based on vegetation spectral signatures and the normalized difference vegetation index (NDVI) for a time series of Landsat-8, Sentinel-2 and PlanetScope imagery of potato, wheat, watermelon, olive and peach/apricot fields. Seasonal metrics were extracted from NDVI curves and the effect of different spatial and temporal resolutions was assessed. It was found that Landsat-8 overestimated SOS and EOS and underestimated MAX due to its low temporal resolution, while Sentinel-2 offered the most reliable results overall and was used to classify the fields in Aljawf. Planet data reported the most precise SOS and EOS, but proved challenging for the framework because it is not a radiometrically normalized product, contained clouds in its imagery, and was difficult to process because of its large volume. The results demonstrate that a balance between the spatial and temporal resolution of a satellite is important for crop monitoring and classification and that ultimately, monitoring vegetation dynamics via remote sensing enables efficient and data-driven management of agricultural system

Groundwater is a vital non-renewable resource that is being over exploited at an alarming rate. In Saudi Arabia, the majority of groundwater is used for agricultural activities. As such, the mapping of irrigated lands is a crucial step for managing available water resources. Even though traditional in-field mapping is effective, it is expensive, physically demanding, and spatially restricted. The use of remote sensing combined with advanced computational approaches provide a potential solution to this scale problem. However, when attempted at large scales, traditional computing tends to have significant processing and storage limitations. To address the scalability challenge, this project explores open-source cloud-based resources to map and quantify center-pivot irrigation fields on a national scale. This is achieved by first applying a land cover classification using Random Forest which is a machine learning approach, and then implementing a circle detection algorithm. While the analysis represents a preliminary exploration of these emerging cloud-based techniques, there is clear potential for broad application to many problems in the Earth and environmental sciences.

Skogsbränder är kraftfulla naturfenomen kapabla att åstakomma omfattande skada över stora ytor och medföra omfattande kostnader för sammhället både humanitärt, ekonomiskt och miljömässigt. Det finns därför ett starkt incitament att övervaka och förutspå skogsbränders utveckling och spridning. Traditionellt används kraftfulla skrivbordsklienter för att köra den simulerings-mjukvara som förutspår skogsbränder, vilket begränsar användningsområdet för simuleringar. Webbklienter är naturligt mobila och lättanvända. Genom att flytta bearbetningen till en server överförs majoriteten av arbetet från klienten. Det här projektet utvecklar ett server-klient baserat ramverk för att simulera skogsbränder, visualisera resultatet och hantera data för användning i skogsbrandsbekämpnings och -analys arbetsflöde. Både parametrarna som skickas till servern och simuleringsresultatet som returneras till klienten. Ramverket använder en kombination av HTTPS-kommunikation och websocket-teknologi för att kommunicera data mellan klienten och server i real-time genom Django-ramverket. Brandmodellen på den Kanadensiska empiriska brandmodellen Prometheus som är implementerad i programmeringsspråket Python. Det är optimerat för det svenska klimated för att enkelt kunnas fältsättas i en webbapplikation för svenska myndigheter. Webb-applikationen är tillgänglig genom mobila och stationära enheter där ramverket beräknar och visualiserar förutspådd fortspridning av skogsbrand i realtid. Skogsbrands moduleringsmodellen av applikationen är jämförd med skogsbränderna i Enskogen och Ängra närastaden Kårböle under sommaren 2018. Noggrannhetsbedömningen av modellen påvisar att den simulerade branden tenderar att innehålla den egentliga elden men är benägen att överskatta eldspridningen. Applikationen utvärderades även genom ett formulär om applikationens funktionallitet som skickades till en provgrupp av personer som arbetar med skogsbränder eller annat relevant område. Provgruppen var nöjd med applikationen och såg ett anvädningsområde för applikationen i sitt arbetsflöde. Mycket arbete återstår för att göra applikationen fältduglig genom integration av myndigheters datatjänster och andra databaser som innehåller riskobjekt, byggnader, kraftledningar e.g. Trots detta ansågs brandingejörer inom räddningstjänster en möjlighet att använda verktygen i dess nuvarande tillstånd om simuleringsresultatet anses korrekt nog för att fungera som underlag för beslut. Detta understryker behovet av en liknande applikation, med vidare funktionalitet och integration med data-system. / Wildfires are powerful natural forces capable of causing extensive damage to large areas of lands and induce a high societal cost in both humanitarian, economic and environmental terms. As such there is a strong incentive to track and predict wildfires' development and spread. Traditionally heavy desktop clients are required to run the simulation-software required to perform wildfire spread predictions, which limits their use and versatility. Conversely, web-based clients are lightweight and versatile by design. By moving the processing of the simulation to a server the bulk of the workload is removed from the client. This project aims to produce a server-client framework for simulating wildfires, visualising the result and handling the fire data for use in the workflow of wildfire suppression and analysis. Both the parameters sent to the server and the simulation result returned to the client. It utilises a combination of HTTPS-requests and websockets-technology to communicate data and information between the client and server in real-time through the Django framework. The fire simulation is based upon the Canadian empirical fire-model Prometheus. The implementation of the algorithm were adopted in the programming language python and optimized for the Swedish climate to be easily deployed in a web-application to be used by Swedish organisations. The web-application was accessible though mobile and stationary devices where the framework calculated and visualised the progression of the wildfire in real-time. The wildfire progression model of the application was compared to the wildfires Enskogen and Ängra, close to the town of Kårböle during the summer of 2018. The accuracy assessment of the fire progression model found that the simulated wildfire progression tend to contain the observed fire and prone to overestimate the wildfires progression. The application was evaluated though a questionnaire which was answered by a sample group composed of persons working with wildfires or wildfire related fields. The sample group were satisfied by the application and broadly found that the application could be implemented into their workflow.  Much work remain to operationalise the application, such as integration of municipal data sources and other databases containing resources, risk-objects, buildings, power-lines. In spite of this Fire-engineers in emergency services state a possibility for use of the application as is, if the simulations are deemed accurate enough and provide a better basis for decision making and measures. This underlines the need of an application such as this in the field, and with further functionalities and integration's with data-systems.

To ensure the highest possible temporal resolution of SAR data, it is necessary to use all the available acquisition orbits and paths of a selected area. This can be a challenge in a mountainous terrain, where the side-looking geometry of space-borne SAR satellites in combination with different slope and aspect angles of terrain can strongly affect the backscatter intensity. These errors/noises caused by terrain need to be eliminated. Although there have been methods described in the literature that address this problem, none of these methods is prepared for operable and easily accessible time series analysis in the mountainous areas. This study deals with a land cover-specific local incidence angle (LIA) correction method for time-series analysis of forests in mountainous areas. The methodology is based on the use of a linear relationship between backscatter and LIA, which is calculated for each image separately. Using the combination of CORINE and Hansen Global Forest databases, a wide range of different LIAs for a specific forest type can be generated for each individual image. The algorithm is prepared and tested in cloud-based platform Google Earth Engine (GEE) using Sentinel-1 open access data, SRTM digital elevation model, and CORINE and Hansen Global Forest databases. The method was tested...

No description available.

Agriculture and food security are interlinked. New technologies and instruments are making the agricultural system easy to operate and increasing the food production. Remote sensing technology is widely used as a non-destructive method for crop growth monitoring, climate analysis, and forecasting crop yield. The objectives of this study are to (1) monitor crop growth remotely, (2) identify climate impacts on crop yield, and (3) forecasting crop yield. This study proposed methods to improve crop growth monitoring and yield predictions by using remote sensing technology. In this study, we developed crop vegetative growth metrics (VGM) from the MODIS (Moderate Resolution Imaging Spectroradiometer) 250m NDVI (Normalized Difference Vegetation Index) and EVI (Enhanced Vegetation Index) data. We developed 19 NDVI and EVI based VGM metrics for soybean crop from a time series of 2000 to 2018, but the methods are applicable to other crops as well. We found VGMmax, VGM70, VGM85, VGM98T are about 95% crop yield predictable. However, these metrics are independent of climatic events. We modelled the climatic impacts on soybean crop from the time series data from1980-2019 collected from NOAA's National Climatic Data Center (NCDC). Therefore, we estimated the impacts of increase and decrease of temperature (maximum, mean, and minimum) and precipitation (average) pattern on crop yields which will be helpful to monitor climate change impacts on crop production. Lastly, we made crop yield forecasting statistical model across different climatic regions in USA using Google Earth Engine. We used remotely sensed MODIS Terra surface reflectance 8-day global 250m data to calculate VGM metrics (e.g. VGM70, VGM85, VGM98T, VGM120, VGMmean, and VGMmax), MODIS Terra land surface temperature and Emissivity 8-Day data for average day-time and night-time temperature and CHIRPS (Climate Hazards Group Infra-red Precipitation with station data) data for precipitation, from a time series data of 2000-2019. Our predicted models showed a NMPE (Normalized Mean Prediction error) with in a range of -0.002 to 0.007. These models will be helpful to get an overall estimate of crop production and aid in national agricultural strategic planning. Overall, this study will benefit farmers, researchers, and management system of U.S. Department of Agriculture (USDA).

Today, people are used to having technology as a constant aid. This also sets expectations that information should always be available. This, together with ongoing climate change that has led to more natural disasters, has laid the foundation for the need to change the methodology for how geographical data is collected, compiled and visualized when used for crisis communication. This study explores how authorities, at present, communicate with the public during a crisis and how this can be done in an easier and more comprehensible way, with the help of Geographical Information Systems (GIS). The goal is to present a new way of collecting, compiling and visualizing geographical data in order to communicate, as an authority, with the public during a crisis. This has been done using a case study with focus on wildfires. Therefore, most of the work consisted of the creation of a prototype, CMAP – Crisis Management and Planning, that visualizes fire-related data. The basic work of the prototype consisted of determining what data that exists and is necessary for the information to be complete and easily understood together with how the data is best implemented. The existing data was retrieved online or via a scheduled API request. Eventrelated data, which is often created in connection with the event itself, was given a common structure and an automatic implementation into the prototype using Google Fusion Tables. In the prototype, data was visualized in two interactive map-based sections. These sections focused on providing the user with the information that might be needed if one fears that they are within an affected location or providing the user with general preparatory information in different counties. Finally, a non-map-based section was created that allowed the public to help authorities and each other via crowdsource data. This was collected in a digital form which was then directly visualized in the prototype’s map-based sections. The result of this showed, among other things, that automatic data flows are a good alternative for avoiding manual data handling and thus enabling a more frequent update of the data. Furthermore, it also showed the importance of having a common structure for which data to be included and collected in order to create a communication platform. Finally, by visualizing of dynamic polygon data in an interactive environment a development in crisis communication that can benefit the public’s understanding of the situation is achieved. This thesis is limited to the functionality and layout provided by the Google platform, including Google Earth Engine, Google Forms, Google Fusion Tables etc / I dagens samhälle är människan van vid teknik som ett ständigt hjälpmedel. Detta sätter också förväntningar på att information alltid ska vara tillgänglig och uppdaterad. Detta tillsammans med pågående klimatförändringar som lett till fler och svårare naturkatastrofer har lagt grunden till att det finns ett behov av att förändra hur man samlar in, sammanställer och visualiserar geografiska data som används för kommunikation i en krissituation. Denna studie utforskar hur myndigheter, i dagsläget, kommunicerar med allmänheten vid en krissituation och hur detta kan göras på ett enklare och mer givande sätt med hjälp av GIS. Målet är att visa ett nytt sätt att samla in, sammanställa och visualisera geografiska data för att, som myndighet, kommunicera med allmänheten under en kris. Detta har gjorts som i en fallstudie med fokus på skogs- och gräsbränder. Merparten av arbetet bestod därför av framtagande av en prototyp, CMAP – Crisis Management and Planning som visualiserar brandrelaterade data. Grundarbetet till prototypen bestod av att fastställa vilken data som finns och är nödvändig för att informationen skulle bli lättförstådd och komplett samt hur denna bäst implementeras. Den existerande data som implementerades hämtades online eller via ett schemalagt anrop av APIer. Händelserelaterade data skapas ofta i samband med själva händelsen och därför skapades en gemensam struktur och direktimplementation till prototypen för denna data med hjälp av Google Fusion Tables. I prototypen visualiserades data i två interaktiva kartbaserade sektioner. Dessa sektioner fokuserade kring att förse användaren med den information som kan behövas om man befarar att man befinner sig på en drabbad plats eller att förse användaren med allmän förberedande information inom olika län. Slutligen skapades även en icke kartbaserad sektion som möjliggjorde att allmänheten kan hjälpa myndigheter och varandra genom ”crowdsource” data. Denna samlades in i ett digitalt formulär som sedan direkt visualiserades i prototypens kartbaserade delar. Resultatet av detta visade bland annat att automatiska dataflöden är ett bra alternativ för att slippa manuell hantering av data och därmed möjliggöra en mer frekvent uppdatering. Vidare visade det även på vikten av att ha en gemensam struktur för vilken data som ska inkluderas och samlas in för att skapa en kommunikationsplattform. Slutligen är visualisering av dynamiska polygondata i en interaktiv miljö en utveckling av kriskommunikation som kan gynna förståelsen för situationen hos allmänheten. Studien är begränsad till att skapa en plattform baserad på den inbyggda funktionaliteten och designen som erbjuds i Googles plattform, detta inkluderat Google Earth Engine, Google Formulär, Google Fusion Tables etc.