Earth day 2023 visualize the predicted vulnerabilities
Clark Labs, a research and development company based in Worcester, Massachusetts, is at the forefront of using geospatial intelligence to predict vulnerability change. Through their cutting-edge research, they are harnessing the power of Earth observation data, specifically the maps provided by the European Space Agency (ESA), to create predictive models that can forecast changes in a given area’s vulnerability to various environmental stresses. By using advanced machine learning algorithms and statistical models, Clark Labs is able to analyze a range of environmental factors, such as land cover, climate, and topography, and predict how these factors will interact to affect vulnerability over time. This innovative approach has significant implications for environmental management and disaster response, as it enables decision-makers to foresee and prepare for potential vulnerabilities before they become major problems. In this article, we will explore how Clark Labs is leveraging geospatial intelligence and ESA maps to revolutionize our understanding of vulnerability change.
Three different models for the most common use-cases
Geospatial intelligence has become an invaluable tool for predicting vulnerability change in various regions around the world. Clark Labs, a research and development company, has developed three distinct models for predicting vulnerability change on a country, region, and global scale. Each model has its unique strengths and can be used in different scenarios. In this blog post, we will explore when to use each of the models and provide a proof of concept using ArcGIS and the Living Atlas.
The Country Model
The country model is ideal for predicting vulnerability change at a national level. This model takes various environmental and socio-economic factors specific to a particular country in account. It is especially useful for governments and non-governmental organizations (NGO) looking to plan and implement policies that address specific vulnerabilities in a country. For example, a government can use this model to predict changes in vulnerability to water scarcity in a particular region and then plan and implement policies to address the issue.
The Regional Model
When we need to recognize the change of vulnerability within a particular geographic region, often at the administrative level 1 or 2, the regional model offers a common precision. This model is ideal for regional planners and organizations interested in understanding how different areas within a region are likely to be affected by environmental and socio-economic factors. For example, an organisation working on disaster response in a particular geographic region can use this model to predict areas that are likely to be affected the most during a natural disaster, such as a hurricane or earthquake.
The Global Model
This is the most high-level model being ideal for predicting vulnerability change on a global scale. Usually, comprehensive sustainable multinational acting agencies prefer the global model and address various environmental and socio-economic factors that impact vulnerability across the world. It is exceptionally useful for international organizations and governments that are interested in understanding how different regions and countries are likely to be affected by environmental and socio-economic factors. For example, the United Nations can use this model to predict changes in vulnerability to climate change across the world and then plan and implement policies to address the issue.
Domain specific use-case addressing water scarcity
Demonstrating the usefulness of land cover models, we need a domain specific use-case using pre-processed datasets from ArcGIS Living Atlas. To address water scarcity, we should focus on the country model and use it to predict changes in vulnerability to water scarcity in a specific region. Some African countries like Kenya are water-scarce countries due to its low supply of renewable freshwater and have also a growing population.
Step 1: Gather Data — The first step is to collect data on environmental and socio-economic factors that impact water scarcity in Kenya. We can use data from ArcGIS Living Atlas, and other spatial-enabled Open Data sources, which provide a vast collection of geospatial data on various environmental and socio-economic factors.
Step 2: Create a Model — The next step is to create a domain specific model using the cleaned datasets in step 1. We can directly use the country level land-cover classifications developed by Clark Labs and adapt it to the specific environmental and socio-economic factors that impact water scarcity in Kenya.
Step 3: Run the Model — The third step is to run the model and generate a map that shows changes in vulnerability to water scarcity in Kenya over time. The Kenyan government and NGOs can use the gained insights working in the region to plan and implement policies to address the issue.
Green Infrastructure
Clark Labs’ models for predicting vulnerability change on a country, region, and global scale are valuable tools for decision-makers interested in understanding how different areas are likely to be affected by environmental and socio-economic factors. By using geospatial intelligence common practices with public available Open Data, we can set up the stage designing a proof of concept that demonstrates the usefulness of these models, and provides decision-makers with the information they need to plan and implement policies that address specific vulnerabilities in their regions.
The process of gathering land cover data involves the use of satellite imagery to classify land use and land cover. The European Union executes a Copernicus land monitoring service where we can access pre-processed imagery as a GeoTIFF file or classified polygon features being shipped as a geopackage. Each geospatial features represent a different land cover classification.
The GeoTIFF file is round about 200 MBs and classifies the raster pixels using 44 different classes. Accessing the pixel values of a raster image can be done using GDAL and numpy in Python. Numpy offers high efficient and effective data analysis and manipulation. The two dimensional numpy array representing the land cover raster band has a dimension of (46000, 65000). The geopackage file contains the classified geospatial features of Europe and needs round about 9 GBs of disk space. Some of these polygon shapes are highly complex and consist of many thousand vertices representing the natural shape of the underlying feature.
Designing a web service offering land cover classification using latitude and longitude poses several challenges. These include the need for accurate and up-to-date data, ensuring compatibility with various data formats and systems, and the need for efficient geospatial processing and delivery of data. Usually, geospatial experts use highly optimized imagery software offering image services capabilities for identifying raster values using latitude and longitude locations.
To create an easy-to-use web app for data scientists, engineers should focus on providing an intuitive web interface, like the Landcover Change Web App, that allows users to quickly access and analyze land cover data. This could include features such as interactive maps, data visualization tools, and the ability to filter and search for specific value patterns. It is also important to ensure optimizing the web app for performance and reliability, even under heavy usage. Supporting wide adoption, there is a need for web service offering a consumption-based business model with an easy to integrate API.
