Filters

2017 - 201912020 - 20212

More filters

2019 - 201912020 - 20222
Reset filters

Settings

search.filters.applied.f.access: openAccess × Has files: Yes × Subject: machine learning × Subject: prediction ×

Search Results

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    A data-driven approach to identify controls on global fire activity from satellite and climate observations (SOFIA V1)
    (München : European Geopyhsical Union, 2017) Forkel, Matthias; Dorigo, Wouter; Lasslop, Gitta; Teubner, Irene; Chuvieco, Emilio; Thonicke, Kirsten
    Vegetation fires affect human infrastructures, ecosystems, global vegetation distribution, and atmospheric composition. However, the climatic, environmental, and socioeconomic factors that control global fire activity in vegetation are only poorly understood, and in various complexities and formulations are represented in global process-oriented vegetation-fire models. Data-driven model approaches such as machine learning algorithms have successfully been used to identify and better understand controlling factors for fire activity. However, such machine learning models cannot be easily adapted or even implemented within process-oriented global vegetation-fire models. To overcome this gap between machine learning-based approaches and process-oriented global fire models, we introduce a new flexible data-driven fire modelling approach here (Satellite Observations to predict FIre Activity, SOFIA approach version 1). SOFIA models can use several predictor variables and functional relationships to estimate burned area that can be easily adapted with more complex process-oriented vegetation-fire models. We created an ensemble of SOFIA models to test the importance of several predictor variables. SOFIA models result in the highest performance in predicting burned area if they account for a direct restriction of fire activity under wet conditions and if they include a land cover-dependent restriction or allowance of fire activity by vegetation density and biomass. The use of vegetation optical depth data from microwave satellite observations, a proxy for vegetation biomass and water content, reaches higher model performance than commonly used vegetation variables from optical sensors. We further analyse spatial patterns of the sensitivity between anthropogenic, climate, and vegetation predictor variables and burned area. We finally discuss how multiple observational datasets on climate, hydrological, vegetation, and socioeconomic variables together with data-driven modelling and model–data integration approaches can guide the future development of global process-oriented vegetation-fire models.
  • Thumbnail Image
    Item
    Neural partial differential equations for chaotic systems
    ([London] : IOP, 2021) Gelbrecht, Maximilian; Boers, Niklas; Kurths, Jürgen
    When predicting complex systems one typically relies on differential equation which can often be incomplete, missing unknown influences or higher order effects. By augmenting the equations with artificial neural networks we can compensate these deficiencies. We show that this can be used to predict paradigmatic, high-dimensional chaotic partial differential equations even when only short and incomplete datasets are available. The forecast horizon for these high dimensional systems is about an order of magnitude larger than the length of the training data.
  • Thumbnail Image
    Item
    Learning from urban form to predict building heights
    (San Francisco, California, US : PLOS, 2020) Milojevic-DupontI, Nikola; Hans, Nicolai; Kaack, Lynn H.; Zumwald, Marius; Andrieux, François; de Barros Soares, Daniel; Lohrey, Steffen; PichlerI, Peter-Paul; Creutzig, Felix
    Understanding cities as complex systems, sustainable urban planning depends on reliable high-resolution data, for example of the building stock to upscale region-wide retrofit policies. For some cities and regions, these data exist in detailed 3D models based on real-world measurements. However, they are still expensive to build and maintain, a significant challenge, especially for small and medium-sized cities that are home to the majority of the European population. New methods are needed to estimate relevant building stock characteristics reliably and cost-effectively. Here, we present a machine learning based method for predicting building heights, which is based only on open-access geospatial data on urban form, such as building footprints and street networks. The method allows to predict building heights for regions where no dedicated 3D models exist currently. We train our model using building data from four European countries (France, Italy, the Netherlands, and Germany) and find that the morphology of the urban fabric surrounding a given building is highly predictive of the height of the building. A test on the German state of Brandenburg shows that our model predicts building heights with an average error well below the typical floor height (about 2.5 m), without having access to training data from Germany. Furthermore, we show that even a small amount of local height data obtained by citizens substantially improves the prediction accuracy. Our results illustrate the possibility of predicting missing data on urban infrastructure; they also underline the value of open government data and volunteered geographic information for scientific applications, such as contextual but scalable strategies to mitigate climate change.