On the identification of soil saturation excess from satellite data (SENSE CDT)

Extreme storms, that were once 1-in-100 year events, will become annual by 2050 [1]. Such a rate of change means that flood forecasting tools based on historic data and prior experience, as are currently used throughout the UK, will no longer be fit for purpose. In 2019, the investment needed to combat flood damage in England alone was estimated at an average of £1bn per annum until 2065, [2]. The floods of early 2020, coupled with unexpected waterway infrastructure failures such as the Toddbrook dam, [3], demonstrate that even this investment may not be sufficient. Determining those areas most at risk from flooding is critical to make the most efficient use of available resources. Despite the economic and social relevance of floods, flood risk prediction studies are limited [4]. Satellite-based studies are relevant as they provide a cost-effective approach to this type of problem, where the large-scale nature of the phenomena involved proscribe the use of other techniques. Identifying flooding from space has been a topic of relevance for the past decade. Projects such as FAME, [5], eSurge, [6], or FAST, [7], focused on coastal flooding and flood risk assessment through modelling and Earth Observation (EO). However, in these projects EO was mainly used to validate numerical models of flooding, rather than as a predictive tool. Flood prediction is also a key goal of ESA and NASA [8, 9]. Data from their respective Soil Moisture Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) satellite systems is used to estimate empirically soil water contents and to identify those areas at risk of saturation. However, these estimates rely upon historical data and do not account for the soil hydraulic history itself, nor for the cumulative effects of climate change. Moreover, information provided by these satellites has a very low resolution: a problem for site-specific studies where resolution in the range of metres is needed.

This PhD project will develop a method to couple EO and soil data to forecast the risk of soil saturation excess. This will provide the first step towards identifying regions with excessive flooding risk under future climates. This research includes the use of big data from multiple in situ sources, multispectral remote satellite data, and machine learning techniques.

The project comprises four key objectives:

– Objective 1: Build a database of existing in situ soil moisture and climate data in areas at risk of saturation excess;

– Objective 2: Develop a model to predict soil water storage capacity at depth from satellite soil surface moisture data;

– Objective 3: Develop methods to increase soil water mapping measurement resolution. High-resolution digital elevation models to be included;

– Objective 4: Predict and forecast runoff risk from coupled soil, meteorological and satellite data for selected regions.

Methodology: Satellite data (SMOS, SMAP) will provide surface soil water content time histories. Soil hydraulic properties will be estimated in the first instance from literature sources, e.g. the USDA Unsaturated Soils Hydraulic Database, to test and develop model functionality. Producing a functioning model (in Python or Matlab) constitutes the first milestone. These results will be compared to in situ observations. Methods used to increase data resolution, for example those based on soil texture heterogeneity or topography, will be examined for use with Scottish soil data. SAR data (i.e. Sentinel-1) will be used to complement SMOS and SMAP information at a higher resolution (10 m). Microwave information will be compared to that derived from the use of neural networks trained to derive relationships between soil water content data and multi-spectral properties [10]. Data used for this include high and super-high-resolution multispectral satellites (i.e. Sentinel-2 and Pleiades through the UKSA Data Procurement initiative).

Check https://eo-cdt.org/ for more information on funding and application process.