Banking and Insurance

What Tasks Does Remote Sensing Solve:

• Bank loans, leases, rentals, and spatial projects after assessing objects, territories, and clients using remote methods. For example, obtaining a preferential loan for a farmer in Europe for fuel and fertilizers after presenting remote sensing data and crop yield forecasts for their agricultural plot.
• Insurance for urban planning, agriculture, climate change, and flood prevention creates demand for ML/AI in remote sensing data. The most advanced sector in banking and insurance, where remote sensing is legally recognized as a reliable, legally confirmed method of providing financial services, is agriculture. Small farm owners produce a significant portion of the world's food and will require up to $180 billion in working capital and loans to continue feeding the world. Interestingly, these small farm owners make up about 31% of the world's population and do not have financial records in banks, making it almost impossible for them to receive support from financial institutions.
• Spatiotemporal analysis and ML/AI methods based on studying satellite images before and after a specific event help to understand which areas were affected, the intensity of material damage and the extent of losses for insurers.
• Flood analysis. Insurance companies can use remote sensing to quickly assess affected regions and then efficiently predict the estimated claim amount for individual properties.
• Deploying and scaling such solutions at the enterprise level requires insurers to consider variations in data quality across different regions. Specific markets may need to account for different climate zones, rainfall patterns, and diverse local data sources. Areas with a large number of private and individual houses may be more suitable for loss assessment using remote sensing technology. In urban areas, interference from high-rise buildings can sometimes create challenges in acquiring flood imagery, though methods are being developed to overcome these issues. For example, insurers can complement remote sensing data with hydrological data. Satellite images, for instance, can provide visibility of river flow and floodplains in areas with hydrological data, while real-time hydrological data can address satellite revisit timing issues during peak floods and signal interference in urban areas.
• Predictive industry performance models based on the analysis of remote sensing data and GIS visualizations to determine industry indicators (revenues, managed funds, registered clients).
• Using geospatial data and satellite imagery to assess damages and prioritize claim processing, allowing insurance agents to verify claims by comparing images taken before and after incidents.
• Risk assessment. Analyzing high-resolution images taken in various weather conditions to accurately evaluate risks surrounding an object.
• Premium strategy. Satellite imagery provides information that helps determine pricing strategy and insurance premiums.

Advantages of Using Remote Sensing

• As banks continue to integrate satellite, UAV, and remote sensing technologies into their workflows, small farm owners, for instance, will have an opportunity to receive financial assistance. This will give banks confidence in working with this demographic group.
• Satellite data is useful not only for farmers. Financial institutions can also use this data to assess agricultural crop yields and remotely evaluate field values. This would help them assess risks to facilitate lending to small farmers.
• Accurate risk assessment using remote sensing methods is one of the biggest challenges banks face when issuing loans. They need to know whether a business owner or a small company will be able to repay the loan. Therefore, banks must review all necessary documents and understand the extent to which the surrounding environment poses risks. This also applies to areas with high emergency risk, such as earthquakes, floods, fires, chemical plants, refineries, etc.
• To verify information in documents provided by farmers, banks traditionally need to send field agents to farms. This process can be resource-intensive in terms of time and labor. However, with satellite images and other remote sensing technologies, this task can be performed remotely. By using crop monitoring platforms, financial institutions have easy access to historical farm data, vegetation indices, farm productivity, etc. This gives them confidence in working with farmers. It’s a win-win situation that reduces costs on both sides.
• Alternative credit scoring - people with low income, living in rural areas and small towns, often lack good credit history, making it difficult for banks to assess their value and provide loans. However, this issue can be solved simply using satellite imagery. These images provide banks with historical data on agricultural lands and construction sites, which banks can use to evaluate property values.
• Remote sensing has been recognized as an effective tool and indicator for extracting socioeconomic metrics such as building size, distance between houses, street width, and lack of trees/greenery (vegetation index). These indicators, along with several GIS criteria and weighted overlay analysis, have been used to identify suitable locations for food banks.
• Lack of transparency – some clients may provide false or forged documents about their earnings and income just to obtain loans from banks. There is insufficient data for a proper assessment—banks cannot rely solely on reviewing documents to evaluate the value of small business owners. They need to send agents to verify the actual situation, which can take a lot of time and resources.
• Modern risk analysis technologies are needed—traditional risk assessment tools need to be replaced with digital methods to keep up with constant changes in the financial sector.
• Industrial finance. Factories, warehouses, retail stores, and major industrial storage facilities financed by banks can be geotagged. Regular inspections using UAVs and mobile 3D cameras should reflect the actual physical inventory on-site. If significant discrepancies are found between the output data and inventory reports provided by borrowers, controllers may request a detailed on-site inspection.
• Infrastructure financing. To monitor large-scale infrastructure projects such as roads, mines, ports, telecommunications, power grids, and housing developments, 3D models based on satellite imagery of project sites, supplemented with UAV surveys, will help measure and assess actual physical progress over time.
• Revenue forecasting. Food insecurity is widespread in many developing countries, where rain-fed agriculture is a primary food production method. Climate change has put them at even greater risk due to constant threats of droughts and floods that often lead to lower yields. However, remote sensing allows experts to obtain accurate crop yield estimates, giving governments time to respond and address food shortages.
• GIS applications in banking. GIS data can be used to create charts, maps, and 3D models of the Earth's surface, including hills, mountains, landslides, trees, buildings, streets, rivers, etc. GIS provides a visual representation of data, shows relationships between locations, and helps determine the best locations for new retail stores, shopping centers, and customer accessibility.
• Satellite imagery enables many organizations to make informed decisions, increase resilience, and better manage risks in a changing world.

Application Examples:

Remote Sensing and GIS for Food Banks

Charitable organizations that collect food from producers and suppliers and distribute it to those in need (hereafter referred to as Food Banks) play a crucial role in ensuring economic sustainability and social cohesion (food preservation, waste reduction).

Objective: Demonstrate how Earth remote sensing (ERS) and geographic information systems (GIS) can be used to address social and environmental challenges faced by food banks.

Solution: In many cases, socioeconomic data is collected through qualitative surveys/interviews and lacks spatial reference. Moreover, even this data may not be available to researchers in the same format as in this study. Remote sensing has proven to be an effective tool for obtaining socioeconomic indicators such as building size, distance between houses, street width, and lack of trees/greenery (vegetation index). These indicators, along with multi-criteria and weighted GIS overlays, were used to identify suitable locations for food banks. The study aligns with UN Sustainable Development Goals #2, #11, and #12. The beneficiaries of this study may include charitable organizations, food producers/consumers (restaurants, hotels, individuals), and environmental services (water supply, energy, waste management).

Result: Suitable locations for food banks were identified based on indirect indicators. It is believed that a census at the neighborhood level would be more accurate, although this study is based on the district level. Expats relocate based on various factors such as rent, job changes, or the need for a larger home. This creates a challenge when developing a GIS database for food bank users (spatial and temporal changes). The study results can be used for environmental conservation, cost savings, and achieving UN Sustainable Development Goals (zero hunger, sustainable cities and communities, responsible consumption and production).

Financing Small Farmers

Utilizing new technologies for remote data collection and analysis of potential clients. Financial service providers who recognize the opportunity to reach financially isolated individuals in rural areas can leverage new technologies for remote data collection and analysis of potential clients.

Objective: Develop credit scoring algorithms to facilitate financing for small farmers. Determine whether information obtained from satellite data can be used as predictive factors for risks such as yield and income.

Solution: A deep learning model was created to predict agricultural crop yields in Kenya. Financial service providers serving small farmers in Kenya can use models based on satellite imagery to assess relevant indicators, such as crop timing and yield magnitude, with minimal marginal costs.

Other organizations in the agricultural value chain—resource suppliers, exporters, and traders—can use detailed information on crop type and density or population distribution to make informed decisions regarding crop selection, expected yields, warehouse locations, price forecasting, and more.

Development organizations and public sector entities (Ministry of Agriculture, Economic Development Agency, etc.) recognize the potential of financial accessibility and are increasingly interested in big data and analytics with geospatial applications.

Result: Models were developed for automatic classification of agricultural crops (corn, beans, and potatoes) for the summer and fall seasons (the seasons for which sufficient imagery was obtained). Over 20 different models with varying degrees of success were tested. The best-performing model used an InceptionV3 convolutional neural network architecture and red, green, and blue channels (visual assets) to predict corn yield based on imagery from the summer of 2016.

Multi-Scale Remote Sensing for Agricultural Insurance Support: From Medium-Term to Instant Payouts

The attractiveness depends on the ability to convince farmers of the necessity of insurance coverage and the insurer's ability to better calibrate (and potentially reduce) premiums applied to farmers.

Objective: Find new ways to monitor crops to make insurance policies covering yield losses more attractive to farmers.

Solution: A methodological approach was developed to support agriculture-focused insurance strategies based on time-series analysis of free multispectral satellite data. The main idea is to link agricultural yield in the medium and short term to calibrate insurance premiums, considering both temporal trends and spatial distribution of biomass production by crops. The study area was selected in the Piedmont region (northwestern Italy) as a test case for methodology validation.

Result: A product providing 16-day composite NDVI maps with a geometric resolution of 250 m from 2000 to 2018 proved effective for describing medium-term yield trends at both regional and macro-agricultural levels. Time-series NDVI maps obtained from Copernicus Sentinel-2 data, with higher geometric resolution (10 m), allowed for more detailed field-level studies, refining insurance risk assessment and linking it to local crop conditions. A simple yet highly operational mathematical model was proposed for calibrating annual insurance premiums at the regional level.

Farmers can be offered more suitable insurance contracts, encouraging them to adopt this type of coverage for their activities. In other words, the insurance company can attract new clients, while farmers can protect themselves with fair and transparent pricing.

Remote Sensing Technology in Claims Assessment: Lessons from the 2021 East Coast Australia Floods

Objective: Flood risk assessment. Optimize the property insurance claims process.

Solution: Satellite imagery was used to detect floods and measure the scale of losses, integrating data on water depth, flood area, and event duration. This included both inland (stormwater and riverine) flooding and coastal (storm surge) flooding. Street-level flood information was utilized as input data for faster claims settlement and for improving vulnerability curves in risk models.

Principal Component Analysis (PCA) visualization facilitated a three-dimensional flood risk assessment while the flooding event was still unfolding. In the case of the New South Wales flood, using event footprints and overlaying them with portfolio exposure data showed that estimated losses were lower than initially expected. If the insurance company had relied solely on publicly available information, the estimate would likely have been conservative—possibly extending over a longer period.

Result: Damage assessment based on remote sensing image analysis provided a realistic damage estimation in a relatively short time frame, as well as greater confidence in the results. Deploying and scaling such solutions at the enterprise level requires insurers to consider differences in data quality across various regions. Specific markets may need to account for different climate zones, precipitation patterns, and diverse local data sources. Areas with a high number of private and individual houses may be more suitable for loss assessment using remote sensing technology. In urban areas, interference from tall buildings can sometimes create challenges in acquiring flood imagery, though methods are being developed to address these issues. For example, insurers can complement remote sensing data with hydrological data. Satellites can provide visibility of river flow and floodplains in areas with hydrological data, while real-time hydrological data can address the issue of satellite revisit time during peak floods and signal obstructions in urban areas.