No Substitute For Ground Truthing
Flood risks within FloodiQ.com are created by combining publicly available data from multiple sources:
- Elevation data from the United States Geological Survey and county governments
- Historic tide gauge readings from the National Oceanic and Atmospheric Administration (NOAA)
- Local sea level variation from NOAA
- Storm surge predictions from the National Weather Service / NOAA
- Sea level rise predictions from the United States Army Corps Of Engineers
- Property details from state and county governments offices
Because the models we developed leverage so many different data sources that have different timelines for updating data, it’s important for us to do a lot of in-field research and measuring to help verify that the outputs of our models are accurate. This also helps us determine how we can optimize them to be even stronger and what might be necessary from a recalibration standpoint.
In order to validate the methodology behind the flood inundation level predictions shown on FloodiQ.com, this flooding season the First Street Foundation data team (Sharai Lewis-Gruss, Jeremy Porter, and I) made visits to South Florida and Hampton Roads, Virginia. During each trip we took measurements of the water depth in areas that were inundated with seawater from the king tides, areas like the ones shown in these videos:
The measurement data we collected (using ArcCollector) included timestamps in GMT and geographic coordinates that we then used to map the data points in geographic information (GIS) software. This allowed for a visual picture of the time and location of each water depth measurement and to have the data in a format for further geoprocessing.
The timestamps also allowed us to match each of the measurements we collected with the readings at all the nearby tide stations.
This is necessary because the tide changes rapidly throughout our collection periods, so we needed to know the exact reading of each gauge at the time each measurement is collected.
In order to match timestamps across the ~2000 measurements we collected, we needed to write scripts to download all the by minute gauge readings from the NOAA Tides and Currents and USGS Water Service APIs. Below are some examples.
We then utilized point sampling procedures, a technique for extracting the value of images at specific locations, on water level (MHHW) and elevation data files .tif) to determine both the land elevation at each point we collected and the typical high water levels of the water nearby. We could then compare each observed point to the prediction calculated from each gauge station
The comparisons are then used to build a calibration matrix to better predict the water level at any point given the distance between that point and the nearby stations.
We are still processing the data at each point, but already we have decided that we can better calibrate our prediction models by utilizing more NOAA tide stations and USGS gauge stations into our models. This will allow for more accurate risk assessments for individuals using our FloodiQ.com tool. We will continue to measure and calibrate each flood season to ensure we’re offering users the most accurate assessments possible.
