Stanford Urban Risk Framework (SURF)

Inspiration

Throughout the 2017-18 academic year, a collaborative research and design course at Stanford University developed a comprehensive risk framework with stakeholder engagement for sea level rise in the San Francisco Bay Area. We have implemented and built upon this analytical framework through a purpose-built suite of Python tools - the Stanford Urban Risk Framework (SURF).

What it does

The SURF tool projects averaged annualized losses (AAL) and uncertainty associated with sea level rise and coastal flooding. A built-in exposure model uses tax assessor data and construction costs to value structures and contents. To evaluate hazard, zonal statistics functions find building flood inundation across many storms of varying return periods. A vulnerability model is then applied to calculate the damage associated with these storms. AAL is then projected across multiple sea level rise projections.

How we built it

SURF incorporates many data sources to construct our hazard, exposure, vulnerability, and risk models. An initial data prep module is run using ArcPy to calculate depths of flooding for OpenStreetMaps building footprints over many flood scenarios. Building use types (single family home, apartment, grocery store, etc.) are spatially assigned via tax assessor parcel data. After geospatial operations are complete, the model runs primarily on Pandas. A parallelized object-oriented exposure model creates an instance of every building in the study area - assigning descriptors such as use type and area. Replacement costs of exposed assets are calculated based on their usage type and square footage using RSMeans data. The vulnerability model calculates damage from floods based on U.S. Army Corps depth-damage functions. Using a Monte Carlo model, we calculate AAL and uncertainty over a study period using sea level rise projections from Kopp et al 2017 - built upon IPCC emissions scenarios: RCP 2.6, 4.5, and 8.5.

Challenges we ran into

Getting many different data sources to come together under a single model was challenging - we spent a lot of time building databases and troubleshooting our functions for reading them. Additionally, the scale of our analysis included 23 flood scenarios and 18,000 affected buildings in San Mateo County, CA - our initial ad-hoc scripts ran into memory errors while processing large dataframes.

Accomplishments that we're proud of

Creating the object-oriented exposure model to replace our previous ad-hoc script improved SURF to a great degree. It more easily allows for future hazard such as earthquakes to be added and also avoids memory errors.

What we learned

We dramatically improved our understanding of object-oriented programming and were exposed to the challenges of software development.

What's next for Stanford Urban Risk Framework (SURF)

SURF will soon incorporate equity functions to calculate indirect effects. Based on Census and Bureau of Labor Statistics data, we are currently developing a model which calculates financial stress across income brackets. The results could be serve as a predictor for bankruptcy or displacement associated with natural hazards. A case study of damages and equity implications for sea level rise in San Mateo County will be submitted in early 2019.