WORKING PAPER - Last Updated: August, 2018 / Page
Economic and Social Costs of Sea Level Rise in San Mateo County
A Project-Based Learning Report by the Stanford Sustainable Urban Systems Initiative
Originally Completed: June, 2018
Contributing Students: Anyansi, I., Arashiro, T., Barns, K., Bhaiya, D., Bhattacharjee, H., Chu, S., Donaldson, A., Evans, M., Fischer, S., Kasmalkar, I., Kozlow, V., Lagron, C., Loos, S., McIntosh, T., Miao, Y., Mullet, B., Olson, K., Pang, J., Roberts, K., Santiago, A., Srivastava, C., Belanger, S., Vadalkar, V., Wang, H., Whiteley, B., Wong, Y., Xiang, X., Karam, Youssef., Yu, D., Zhu, Y.
Teaching Team: Baker, J., Balbi, M., Bick, I., Cain, B., Leckie, J., Ortolano, L., Ouyang, D., Serafin, K., Suckale, J.
DISCLAIMER: The following report was produced by students at Stanford University as part of fulfilling their course requirements. It should be viewed as an academic course report and not professional consultation. All results should be considered preliminary and not professionally reviewed. If you wish to copy and distribute substantial portions of this report, you may do so only with written permissions of the teaching team. Any inquiries should be addressed to email@example.com.
Average Annualized Loss
Association of Bay Area Governments
American Community Survey
The California Department of Transportation
Content Structure Value Ratio
First Floor Elevation
Global Multi-resolution Terrain Elevation Data
Intergovernmental Panel on Climate Change
Longitudinal Employer-Household Dynamics Origin-Destination Employment Statistics
Metropolitan Transportation Commission
Our Coast Our Future
Priority Development Areas
Representative Concentration Pathways
Sea Level Rise
San Mateo County
Sanitary Sewer Overflow
Stanford Urban Resilience Initiative
SUS (Stanford SUS)
Sustainable Urban Systems Initiative
United States Army Corps of Engineers
The Intergovernmental Panel on Climate Change (IPCC) estimates that the global sea level will rise by up to two meters in the coming century. This increase poses a significant threat to coastal communities around the world, especially for the regions with infrastructures bordering the shoreline. In the San Francisco Bay Area (hereafter, Bay Area), San Mateo County (SMC) is expected to be heavily impacted by flooding due to sea level rise (SLR) between now and 2040. In this report we investigate this impact by estimating the financial cost of flood damage to buildings in the county and quantifying indirect losses that may exacerbate those damages. We also investigate the potential damage to already vulnerable populations, including community members whose primary language is not English and low-income households.
This report documents the results of the Sustainable Urban Systems project class (SUS Project) at Stanford University. It summarizes our findings based on a combination of literature review, data gathering and analysis, and community outreach. Our study uses a variety of third party software, as well as a risk tool developed by the Stanford Urban Resilience Initiative (SURI), to quantify the expected effects of SLR on SMC. These tools enable the overlay of a map of building footprints with flood maps and the calculation of the damage to each building based on empirical depth-to-damage curves. We find that in our conservative SLR scenario, flooding will likely cost the county an average of $410 million each year from now to 2040 if no adaptation strategies are adopted.
The report also investigates the cascading effects of SLR. SLR can affect critical infrastructure, transportation systems, and businesses. As communities in the Bay Area continue to grow, SLR can also influence development patterns in the region. In addition, the report acknowledges the social implications of SLR in SMC by investigating which communities are disproportionately affected. The report also outlines the limitations of the study and suggests directions for further research into the effects of SLR in the Bay Area.
Sea levels are expected to rise significantly from now to 2040. Forward-looking governments and public agencies are beginning to investigate how to prepare their constituents for the effects of SLR. In September 2017, SMC, along with a number of local agencies, partnered with students at Stanford University to investigate the potential effects of SLR on the county. Our goal was to develop a set of modeling tools and actionable information that quantifies potential losses for SMC. Our intention is to provide local government staff, policy makers and residents with information that will facilitate exploration of hazard-related issues and concerns. This report analyzes the financial and socio-economic impacts of SLR in the County, as well as the effects of SLR on the broader Bay Area.
The context of our work was a three-quarter Stanford project course sequence that ran from September 2017 through June 2018. Throughout the course we met with a number of community partners to enhance our understanding of the the losses that SMC may face. Our overall goal was to create a body of research to inform next steps that might be taken by local governments and residents as well as to elaborate on impacts and vulnerabilities across different stakeholder groups.
The first part of our analysis focused on direct damage to buildings and building contents. However, these direct damages may be the tip of the iceberg; flood events can have effects on critical infrastructure systems, transportation networks, and businesses. Floods may also aggravate socioeconomic vulnerabilities for certain communities, deepening inequalities present in SMC.
In broad terms, this report includes estimates for the direct and indirect effects of SLR to coastal flooding in the Bay Area. Specifically, we include estimates for the following:
Uncertainty is present in all the inputs used in models for conducting policy analysis, and our models are no different in that regard. Our goal is to quantify uncertain futures in a way that provides policymakers with actionable information. Notwithstanding that many of our results are quantitative, they are inherently uncertain. Our numerical results must be viewed as approximations. Though our numerical results are not definitive, we believe our methodology can provide a foundation for conversations that are evidence-based.
The focus of our study was SMC, but many of the impacts areas were beyond the County boundaries. Our analysis highlights the far-reaching network effects of floods; e.g., it clarifies how floods of roads in SMC will disrupt commuters in both city centers around the Bay as well as regions that are further inland, such as Daly City.
As global sea levels rise over the coming decades, SMC will face an increasing risk of flood hazard. The Sea Level Rise Vulnerability Assessment report produced by the SMC Office of Sustainability highlights major challenges in the near future, such as the safety and well-being of vulnerable communities, and the disruption of major infrastructure systems including road networks, utilities and airports. The SMC study involved identifying the communities, infrastructures and regions exposed to flooding, assessing their vulnerabilities and the potential impacts of SLR, and providing information to reduce these impacts.
In line with the goals of the aforementioned SMC assessment, the SUS Project investigated the effects of flooding and SLR by estimating the direct losses to buildings in the county and quantifying indirect losses caused by disruptions to utilities, road networks, and businesses while also exploring effects in the form of disproportionate impacts on vulnerable communities.
As mentioned, this report is the result of a year long SUS project-based class at Stanford University that aims to assess the impacts of SLR in the Bay Area. During the Autumn 2017 academic quarter, the SUS Project produced baseline assessments of risk for the regions along the Bay starting from the city of Burlingame down to San Jose. In the Winter 2018 quarter, we refined these assessments of direct losses to structures due to flood inundation, and explored potential indirect impacts to low-income households, regional businesses, and public buildings. Guidance and suggestions from our community partners, including a number of city staff and policymakers, prompted us to refine our analysis by focusing our investigation within SMC.
The SMC Office of Sustainability has been a key partner for us throughout our investigation. Based on their requests and feedback, we conducted our Spring 2018 quarter study with the following objectives in mind:
Our overarching goal for the Spring 2018 quarter was to create a risk assessment approach that built upon our Autumn 2017 and Winter 2018 analyses to yield results that are more comprehensive, descriptive and actionable. One of our long-term goals is to develop a set of analysis tools that could be used directly by planners and policymakers in SMC and other Bay Area locations.
Climate change and SLR are among the major challenges of the 21st century. Communities all over the world, including the Bay Area, are collaborating and taking action to reduce the impact of SLR. Stanford University has public service as part of its mission and aims to be part of this collaboration and assist communities by providing academic support as we collectively address the challenges ahead.
Our first measure of flood risk to SMC comes in the form of direct economic losses. Direct economic losses, as defined here, are economic damages that occur to building structures and their contents. In order to calculate direct economic loss, we use a measure called Average Annualized Loss (AAL), defined as the average expected dollar loss to property over our study period of 2020-2040. AAL is a measure of the risk an area will face, and is a function of three measures: hazard -- the inundation depths linked to flooding; exposure -- the dollar value of assets subject to flood inundation; and vulnerability, as reflected in a relationship between depth of inundation and damage to structures and their contents.
Figure 4.1 Example house inundation under 1 year flood (left), 20 year flood (middle) and 100 year flood (right).
Figure 4.1 above shows an illustrative house under three flooding scenarios (assuming that there is no SLR at the current time). In the 1-year flood event, there is no flooding that occurs in the area in which the house is located. In the 20-year and 100-year flood event, flooding is observed. For the purpose of this example, the inundation depths are summarized in Table 4.1.
As detailed in Appendix A, with different base amounts of SLR, these inundations would increase. Our calculations utilized maps produced by AECOM for the Adapting to Rising Tides initiative. However, these floodmaps are also imperfect and have many limitations, which are detailed in Appendix B.
Exposure is the dollar value of the structure and content of buildings and is an input to our analysis. We use replacement costs to represent these dollar values. The calculation procedure accounts for local construction costs ($/sq.ft.), building floor area (sq.ft.), and building type. (The procedure is detailed in Appendix A.4). For example, the structural replacement value is estimated to be approximately $550,000 for an average-sized single-family residential building in SMC.
The content value is calculated by multiplying the structure replacement value by the content-structure value ratio (CSVR) provided in a local study prepared by the U.S. Army Corps of Engineers (USACE, 2011). The CSVR estimated for residential buildings is 0.5; therefore, the content replacement value for an average single family home is is $275,000.
Figure 4.2 Depth-Damage Curve for Single Family Residential building Structure (blue) and Contents (orange), from USACE San Francisquito Creek flood report.
Within the analytic framework we employed to assess AAL, a building’s physical vulnerability is defined as the damage to the building given a level of flood inundation. To measure the damage in response to inundation, we used depth-damage curves, which show the percent damage to an asset given a certain depth of inundation. The depth-damage curves are derived empirically from USACE studies of past flood events. Shown in Figure 4.2 is the depth-damage curve for SFR structures and contents. In order to obtain the structure and content damage, the damages are interpolated from Figure 4.2 for the given flood inundation depths. This is summarized in the Table 4.1 below.
Table 4.1 Sample damages for structure and content of a single-family home, shown as percentages. Note that content damage percentages are lower than structural percentages, and that damages increase as return period increases.
As noted above, AAL is defined as the expected loss per year due to flooding, averaged over a period of time (years). Take, for instance, Figure 4.3 below which shows the loss per year for an arbitrary region over many years. In most years, one can expect the losses to be relatively similar, and occasionally there will be years in which very small storms (or no storm at all) occur; those years will result in even lower losses. Sometimes, one may see a rare, but extreme storm which results in exceptionally large losses. If one were to record all the losses for a number of years and take the average across them all, the result would be the AAL.
Figure 4.3 Example of loss due to flooding per year, over many years.
Decisions had to be made about how to handle uncertainties in flooding events caused by winds along the coast of the Bay as well as SLR. To handle SLR, we examined the projections of the IPCC and decided to use their projection labelled Representative Concentration Pathway (RCP) 2.6 (see Appendix A). For the coastal flooding due to winds and tides, we employed the scenarios given by the AECOM SMC study (2016). AECOM lays out future coastal storm events in terms of probability of exceedance (e.g., the 10 year flood is expected to be equalled or exceeded with a probability of 0.10 in any year, whereas the 100 year flood is expected to be equalled or exceeded with a probability of 0.01 in any year). AAL offers a robust method for quantifying the effects of flooding by taking into account the probability of storm events being equalled or exceeded and the associated losses from them. AALs are often used by insurance companies to structure their insurance policy for a given peril (e.g., earthquake, flooding).
AAL can enable policy makers to make decisions in a robust, quantitative manner. Traditionally, the direct benefits of measures to protect against flooding are taken as the losses avoided, which is what AAL measures. Since AAL does not include any of the indirect losses discussed in Chapters 5-8, it provides a lower bound on benefits. Decision makers can use the impact on AAL when considering the cost of alternative flood mitigation actions.
Our value for AAL for SMC from 2020 to 2040 was $410M: $270M in structural damage plus $140M in content damage. Mapping out the results across the county shows that damage is spread across the bayfront cities, but is highest in Redwood City (see Figure 4.4).
Figure 4.4 Map of incorporated cities in SMC that can expect damage due to SLR by 2040. The redder cities, such as Redwood City can expect more damage while uncolored cities can expect no damage.
Chapters 5-8 of this report consist of explorations of the many potential indirect losses that follow the direct economic losses described in Chapter 4, which can be considered “the tip of the iceberg”. There are many areas of indirect loss we were unable to cover in the course of the SUS Project, but this preliminary work provides a framework for refining and adding to our understanding of indirect losses in future work.
The term “critical infrastructure” is commonly used by governments to describe assets essential for a society to function. In the context of flooding in SMC, critical infrastructure includes: road networks, airports and public transport systems; electrical power stations and grid elements; water supply and wastewater disposal systems (including related pumping stations); telecommunication systems; and public health and safety systems, including hospital access and the ability to deploy police and fire safety personnel. Given the limited time available, we decided to focus on the infrastructure elements of immediate concern to SMC staff, namely wastewater systems, electrical substations and road networks. For this investigation, which is preliminary in nature, we decided to focus on a single city (San Mateo City). This narrow focus allowed for more accurate ground-truthing of our information.
In the event of a major storm, it is important to look at the electrical system as a whole. Of the 43 substations in SMC, 16 are in the flood zone associated with 48 inches of inundation, as shown in Figure 5.1. Accounting for the fact that most substations place critical infrastructure at least 12 inches above the ground, these 16 substations would be flooded by 60 inches of total water, which is a level of inundation that is conceivable during the 2020-2040 study period. Losing some or all of these 16 substations would have significant effects on the safety and the economy of the region. For example, in the City of San Mateo, a large flood could knock out approximately 49% of the city’s available power generation. This approximation takes into account the relative sizes (in kV) of each of its four substations. For the City of San Mateo, this power loss would amount to approximately $30.4 million in lost economic activity in a large flood (Rose et al, 2007).
Figure 5.1 Map of the 48 electrical substations in SMC, overlaid with a 48-inch flood scenario (purple). White lines represent municipal city borders. 16 substations are within the flood zone.
Wastewater collection, treatment and disposal systems consist of the pipe networks, pumps, and treatment plants that transport and treat a community’s municipal and industrial wastewater. Each of the system’s component parts plays a critical role to the physical health of a community, and each can be affected by SLR in SMC. One of the particular concerns of San Mateo City staff we worked with was the possibility of SLR to increase the number of sanitary sewer overflows (SSOs, i.e., untreated sewage discharged from a sanitary sewer into the environment prior to reaching wastewater treatment facilities). These overflows have been linked to increased health risks. Our analysis of SSOs in the City of San Mateo suggest that the majority of SSO events in the past were not storm-related, and the number of SSOs would not significantly increase due to SLR. Further details are given in Appendix C.
In sewer networks not served entirely by gravity flows, pump stations are used, and the pumps are controlled to deliver wastewater flows to the wastewater treatment plant and to deliver treated wastewater to receiving waters (in this case, San Francisco Bay). Wastewater treatment plants are often located at low elevations near coastlines to minimize the cost of collecting wastewater and discharging treated effluent (Hummel, et al., 2018). In San Mateo City, both sanitary pumps and wastewater treatment plants are at low elevations and thus particularly vulnerable to SLR. Indeed, these low elevation locations are most susceptible to SLR. In the City of San Mateo, two pump stations stand to be inundated at 48 inches of SLR, which, after accounting for the approximately 18 inches of freeboard given to pump station entrances, puts the two pump stations just within the maximum reasonable flood expected by 2040 (see Figure 5.2).
Figure 5.2 Map of all wastewater pump stations in the City of San Mateo, overlaid with a 48 inch flood scenario (purple). Note that two pump stations are within the flood zone at 48 inches.
Similarly, one wastewater treatment plant becomes inundated at 54 inches. The American Society of Civil Engineers (ASCE) code requires that all critical wastewater infrastructure, including necessary electrification, must sit at least 2 feet above the base flood elevation (equivalent to 60 inches for San Mateo’s plant). This would make the plant susceptible to flooding in an extreme event. However, the code is not entirely relevant to an assessment of the plan’s vulnerability. The plant is over 75 years old, and many components still in use were installed before the current code requirements were in place, casting doubt on the plant’s ability to function after more moderate floods.
We conducted a commute disruption analysis to understand how commute patterns in the Bay Area would be influenced by coastal flooding. Impacts of flooding on the road networks will be cascading in the sense that road closures will alter the ability of employees to reach their workplaces, which will have subsequent impacts on businesses and result in associated economic losses. We did not attempt to analyze the additional adverse effects of road closures, such as the impacts on daily life (shopping, dropping children at schools) or the inability of emergency and public transport vehicles to function effectively, among many others.
Vehicle traffic in the Bay Area is already stressed, having grown 80% since 2010 (Baldassari, 2017). This situation will be exacerbated by disruptions from flooding, which can lead to increased commute distances and travel times, lost productivity, and reduced economic output. Our analysis is directed toward characterizing the aforementioned disruptions to traffic flows with high spatial and temporal resolution.
The traffic model we built characterizes flood disruption by computing the added travel time and total economic loss experienced by commuters and businesses in the 9 counties that comprise the Bay Area (Metropolitan Transportation Commission, MTC, 2018). The model is detailed in Section D of the Appendix. We conducted the analysis for a number of flooding scenarios, ranging from 0 to 66 inches of inundation, with a spatial resolution of one Census block and temporal resolution of one hour.
The nine-county scope of our analysis recognizes that commute patterns cover the entire Bay Area. Thus, this broad scope makes the analysis more useful than focusing on a single county or set of neighboring counties. We focused our analysis on six flooding scenarios, ranging from 12-66 inches, in addition to a baseline case with no flooding.
We assessed the impact of flooding by running our model using three sources of input data. The first input was a road network file obtained from MTC’s Travel Model One program, which is MTC’s comprehensive traffic simulation model for the Bay Area. The second input was an origin-destination matrix, which contained the volume of commuters as well as their commute origins and destinations. This was obtained from the Longitudinal Employer-Household Dynamics Origin-Destination Employment Statistics (LODES) California 2015 dataset, which generates the aforementioned origin-destination commute data from the 2010 census, administrative records, and survey data. Finally, our third input was flood maps from AECOM that included the aforementioned 12-66 inch flooding scenarios.
In addition to these sources, the California Department of Transportation (Caltrans) and the Global Multi-resolution Terrain Elevation Data (GMTED) elevation datasets were used to conduct an elevation adjustment of the road network. Processing of these datasets is described in Section D.2 of the Appendix.
Our commute disruption model was designed to simulate car traffic only. Therefore it only accounts for commuters who drive alone or carpool to their workplace. Commuters using public transit, including buses, and other modes of transport (commercial vehicles and trucks) are excluded from the analysis. Thus, this model provides a conservative simulation of the road network, since there would be more vehicles on the road from other modes of transportation. Our LODES origin-destination dataset aggregates commuters from all commute modes; in order to adjust this data to the relevant commute modes, we estimate the proportion of commuters carpooling and driving alone as described in Section D.2.4 of the Appendix.
Our model simulates the home-to-work commute in the morning, defined as the period from 5am to 11am. We assume that all of the trips that are being made are for going from home to work. Section D.2.3.3 of the Appendix details the process for dividing the morning period into one-hour windows. Peak traffic hours were assumed to be 7am-8am, 8am-9am, and 9am-10am, and off-peak hours are assumed to be 5am-6am, 6am-7am, and 10am-11am. Since the morning period is long, distinguishing between the different hours allows for more precise representation of traffic patterns.
Based on feedback from the Association of Bay Area Governments (ABAG), we simulated two different road closure scenarios. We assumed that road links were closed to traffic at either 3 inches or 1 inch of inundation, and produced results for both scenarios. For partial inundation below these thresholds, we assumed that traffic speed would follow Pregnolato et al’s empirical formula (Pregnolato 2017) and reduce quadratically with the inundation depth, as detailed in Section D.2.3.
The output from our traffic model is travel time on each road link for a given flood depth, traffic demand, and road closure scenario. For each of the scenarios, we compute travel time, travel distance, and accessibility of the shortest time path for each origin-destination pair present in LODES.
After adjusting the commute data of total commuters to the population of car commuters, we then analyze the scale of commute disruption compared to a baseline condition in which there is no flooding. The analysis is performed for the entire San Francisco Bay Area for consistency. However, it is possible to break the analysis down to county or smaller administrative levels, as detailed in Section D.3.3 of the Appendix.
One dimension of our results concerns the total economic losses from an employee’s perspective. We used hourly wage by income group to estimate commuters’ financial losses from being late to work and unable to get to work. AAL per year from 0, 6, 12, 18 to 24 inches SLR was derived, and finally an estimation for the RCP 2.6 scenario for 2020-2040 horizon was provided. Due to the uncertainties associated with traffic simulation and loss estimation, economic loss numbers should not be interpreted as precise values. The details of our calculation methodology and its limitations are illustrated in Section D.3.2 of the Appendix.
We assessed the performance of the road network system on the basis of travel time in minutes. Compared to vehicle miles traveled, which is an alternative metric we considered, we think travel time is more robust and reliable for the following reasons:
Figure 6.1 summarizes model results by plotting the proportion of commuters versus their driving time from home to work under different scenarios. Result for peak traffic and off-peak traffic are synthesized to represent the overall morning traffic condition on a day of flooding, as described in Section D.3.2.2 of the Appendix. It can be seen that the histogram shifts to the right and down as flooding levels increase for both road closure scenarios displayed in Figure 6.1. The rightward shift indicates that the driving time increases for a given proportion of commuters; for example, in Figure 6.1 (a) 60% of commuters take less than 25 minutes to commute to work when there is no flooding, but that number increases to 50 minutes with 48 inches of flooding. Similarly, the downward shift indicates that not all commuters are able to complete their commutes due to road closures from flooding.
In the case of three less-extreme scenarios (AECOM 12, 24 and 36 inch inundation scenarios), we can see that road closure at 1 or 3 inches yields significantly different results; for example, the commute disruption impact brought by AECOM 24 when roads are closed at 1 inch is higher than that of AECOM 36 when roads are closed at 3 inches. While a more restrictive road closure condition (i.e. 1 inch of inundation) causes more severe commute disruption, it may decrease the probability of accidents, which is a potential source of cascading disruption that is not captured by our model.
Figure 6.1 Histogram of commuter travel time in minutes for (a) road closure at 1 inch of inundation, and (b) road closure at 3 inches of inundation. The y axis indicates the proportion of commuters whose driving time is less than the corresponding value on the x axis for a given point on the curve. When a curve shifts right, it therefore indicates that driving time increases for a given proportion of commuters. When a curve shifts down, it indicates that not all commuters are able to complete their commute due to flooding.
We estimated economic losses by assigning dollar values to the added travel time and lost work hours for each of the AECOM flood maps. The methodology for dollar value assignment and AAL calculation is detailed in Section D.3.2 of the Appendix. The economic loss for road closures at 1 and 3 inches of inundation is displayed in Figure 6.2, which indicates that the economic impact of different road closure thresholds is significant. For example, closing roads at 1 inch of inundation with 0 inches of SLR results in an AAL comparable to closing roads at 3 inches of inundation with 24 inches of SLR. Additionally, Figure 6.2 (b) indicates that AAL increases steadily with higher SLR scenarios, and losses are caused by employees being both “late to work” and “unable to work”.
This trend is less significant in Figure 6.2(a), where we apply a more restrictive road-closure scheme. As sea level increases and more roads get closed, more commuters are blocked from reaching their destination. Thus, there is lower overall traffic demand, which may reduce average commute time for those who are still on the road. As a result, while loss from “unable to work” still rises with higher sea level, loss from “late to work” stays static.
Figure 6.2 An dollar amount estimation of AAL for commute disruption for (a) road closure at 1 inch of inundation and (b) road closure at 3 inches of inundation.
Using the RCP 2.6 scenario, which gives the probability of exceeding a certain sea level for the decades 2020, 2030, and 2040, we calculated the expected commute disruption AAL for the 2020-2040 period. Figure 6.3, i.e. the 1 inch road closure case, displays the commute disruption AAL results by county as well as for the entire Bay Area. In the RCP 2.6 SLR scenario, the entire Bay Area is expected to see a commute disruption AAL of $223 million, while commuters working in SMC are expected to experience a commute disruption AAL of $42 million. As Figure 6.3 demonstrates, San Francisco County, Santa Clara County and Alameda County will experience most of the commute disruption.
The magnitude of commute disruption calculated here is comparable to the loss from direct building damage, indicating the importance of commute disruption in analyzing risk of SLR faced by the area. A detailed breakdown of the AALs by county for different RCPs is given in Section D.3.2.3 of the Appendix.
Figure 6.3 AAL from commute disruption for RCP 2.6 for the 2020-2040 horizon, for road closure at 1 inch scenario, for the 9 counties in the Bay Area.
Our commute disruption model advances MTC’s Travel Model One by refining the spatial resolution of the origin-destination matrix from what MTC refers to as “super districts” to the higher-resolution census block level. Additionally, our model considers an hourly window of travel, and uses terrain elevation data to study the impact of road inundation on traffic disruption, resulting in a more complete flood disruption analysis. Our results indicate that the scale of commute disruption is significant, in terms of the proportion of commuters affected as well as in terms of dollar value losses, with the latter being comparable to direct building damages. By comparing results for road closures at 1 and 3 inches of inundation, we highlight the importance of road closure thresholds used by relevant stakeholders. Finally, by displaying results at the county level, we indicate areas to be prioritized in SLR adaptation and resiliency programs.
One major finding of our analysis was that despite a majority of exposed assets being residential buildings, commercial buildings actually make up 51% of the contribution to AAL due to direct damage to building structure and content (more details on these findings can be found in Appendix E). As shown in Figure 7.1, the highest areas of building structure and content damage within SMC centers on the low-lying commercial areas in Redwood City, Burlingame, and East Palo Alto. Aggregating this to the entire county, we get an estimated $206 million in direct economic loss AAL to commercial buildings.
Figure 7.1 Direct losses (AAL for building structures and contents) sustained by commercial buildings in SMC.
Our hypothesis was that this direct economic loss was really just the first step in understanding the larger impact that flooding could have on the commercial sector. Next, we explore indirect losses specifically as they could relate to businesses with the county.
Disruption of commercial operations was one of the topics we explored to gain an initial understanding of indirect commercial losses due to flooding. The research literature on this subject is sparse. Our effort represents an initial attempt and many assumptions were made in the process of examining operational disruption. Further research into how different segments of the business sector are affected differently by downtime and disruption is necessary to arrive at more refined estimates of operational loss as well as more thorough numbers on calculating net revenue and variable costs.
Using available data for estimating sales volume at the business level, we aggregated this data to the building level in order to allow us to compare it with the calculated AAL. For each of these businesses, we estimated days of disruption by taking the percentage of damage to total value for the building and approximating the amount of time the business would be closed based on this percentage of damage. This number totaled ~8k business-days per year for SMC. There are more details on this calculation in Appendix E.2.8.
We next calculated sales volume at the block group level by aggregating the building level sales volume to the appropriate block group. Then, we subtracted the calculated employee wages at the block group level from sales volume in order to get a net profit/loss. This number was then divided by 250, which was the assumed average days of operation for businesses per year, which gave us a daily net profit/loss. This resulted in an estimate of $152 million in net profit per day in all of SMC.
Finally, we divided the net profit by the total # of businesses affected which was ~16k to get an average net profit per business that could serve as a potential daily profit per business, which was ~$9,600. Then we multiplied the average potential daily profit by the total days displaced to get an estimated $76 million in operational disruption for all of SMC.
Building on the findings found from the commute team, we also estimated the total economic loss caused by employee commute disruption. We divided that disruption into two main components: disruption caused by delays, and disruption caused by inaccessibility to the workplace. Following the commute team’s framework and scenario-based approach (results in Figure 6.3), every commuter had an economic impact to businesses from either of those two components. We then summed the individual values, and aggregated them to the county level, resulting in an economic loss of $35 million caused by delays and $7 million caused by inaccessibility, and thus a total employee commute disruption loss of $42 million.
The results show that the total indirect losses calculated sum up to $120 million, with $76 million attributed to operational disruption loss and $42 million to commute disruption loss. This is slightly more than half of the total direct commercial AAL. These are significant results, and warrant further studies into indirect commercial losses.
Figure 7.2 Total direct and indirect commercial losses for SMC for RCP 2.6 in the time frame of 2020-2040. Our calculated indirect costs are 57% of the direct costs.
In the context of flooding, socially vulnerable communities are those less able to recover from floods because of demographic factors, such as low income and poor English language proficiency. As a result of demographic factors, some members of communities are less able to prepare for, respond to, and recover from shocks, such as flood events, than their more resilient counterparts. Understanding the distribution of damage from sea level rise and flooding across different segments of a community is an important step in creating a successful plan for resiliency that accounts for challenges faced by vulnerable populations. The discussion below aims to address two questions: Are flood damages distributed equitably among groups characterized by different demographics? And, does expected flood damage exacerbate pre-existing inequities experienced by more vulnerable populations?
For purposes of measuring the degree to which vulnerability to flooding was distributed equitably, we selected several metrics: English language proficiency, income level, education level, gender, age, gender of the household head, and race/ethnicity. Combinations of one or more of these measures may compound with other factors to intensify vulnerability and worsen impacts.
We evaluated the equity of the distribution of flood damage by comparing the relative proportions of a given demographic at large with the relative proportions of that same demographic experiencing damage. Consider, for example, income level. Flood damage would be distributed equitably across income level if the proportion of low-income households experiencing flood damage were the same as the proportion of low-income households in the population as a whole.
In order to determine the breakdown among those experiencing damage, we first collected demographic data from the U.S. Census at the smallest level of granularity available: block group. We then calculated the percentage of each block group experiencing damage (in our case, uncompensated AAL, explained in Section F.3.1 in the Appendix) from flooding due to SLR. We used the total counts of each demographic in the block group to determine the fraction of each demographic experiencing damage. At this point in the analysis it was necessary to make an assumption because, at the block group level of data, assertions cannot be made about the socioeconomic characteristics of individual households. We assumed that within a block group, all households with a given demographic characteristic have an equal likelihood of experiencing damage.
Consider, for example, the income variable for an arbitrary block group in which there was flooding. Using flood maps, it was possible to identify the number of parcels in the block group experiencing damage, but it was not possible to identify which of those parcels experiencing damage was a low-income household because we only have income data aggregated at the block group level. The assumption we made is that the fraction of low-income households experiencing damage within the block group is equal to the fraction of low income households in the block group. Thus if 50% of the household within a block group are classified as low income and 10% of the households in the block group experience flood damages, then 50% of the households experiencing damage are assumed to be low income households. This would constitute 5% of all households within the block group.
We made comparisons across two spatial levels: SMC and individual cities within the County. SMC as a whole contains around 770,000 people, and by 2040, 22,000 of these people will likely be experiencing damage under the RCP 2.6 Scenario. Figure 8.1, which is for the metric English language proficiency, shows that for SMC as a whole: 54% of the population speak only English; 38% primarily speak another language, but speak English “well” or “very well”; and 8% speak English “not well” or “not at all”. The lower bar in Figure 8.1 concerns the proportions of people in SMC with different levels of English language proficiency who experienced damage. When we compare figures for the County as a whole to those experiencing damage, only 42% of those experiencing damage speak only English, 43% speak English well or very well, and 14% speak English not well or not at all. In a scenario in which damages were distributed equitably, the proportions of those with different English proficiencies would be the same in the flood region as for the County as a whole. However, for SMC as a whole, 20% fewer English-only speakers will be experiencing damage than we would expect. For individuals who primarily speak another language, 12% more individuals with high English proficiency, and 75% more individuals with low or no English proficiency, will experience damage than we would expect.
The results in Figure 8.1 point to an inequitable distribution of flood damage among those with different levels of English proficiency. This inequity can be represented by an index that measures the average percent difference from expected outcome for each subgroup. Using the English language proficiency example, if damage were distributed equitably we would expect 54% of the households experiencing damage to be English-only speakers. The discrepancy between the County-wide English-only result and the English-only result for damaged parcels is 42% divided by 54%, which is 78%, or 22% less than 100% which would represent no skew. This type of percentage can be repeated for each of the other two categories in Figure 8.1. The average of these percent differences (in absolute value terms) is 0.36.
We were able to do this same English-language proficiency calculation for each city in SMC (Table 8.1). We also were able to calculate this type of inequity measure for a number of other metrics, such as education level and age. It is important to note that the inequity measure reflects inequity within a defined geographic area. Thus, for example, low inequity numbers in Table 8.1 for East Palo Alto simply indicates that households within different demographic categories (e.g. English proficiency) experience damage in roughly the same counts. The figures in the table for any one jurisdiction cannot be compared to those in any other; the comparisons relate to what would be expected if flood damage was distributed equitably within a jurisdiction. East Palo Alto has the second highest count of those experiencing damage, and the second highest count of those experiencing damage whose primary language is not English. This inequity measure should not be considered in isolation, but rather in conjunction with the total counts of vulnerable populations experiencing damage in various locations and the total amount of damage being experienced by households with these demographic characteristics
Figure 8.1 English proficiency of residents in San Mateo County as a proportion of total population. Note that those who speak English “not well” or “not at all” are more likely to experience damage.
Table 8.1 Table of social inequity scores for cities in SMC experiencing damage due to SLR. Each city receives a different social inequity score for seven different metrics. Red boxes are more inequitable scores, and green boxes are more equitable scores.
We are also able to calculate the amount of damage experienced by each demographic. In Figure 8.2, we observe that lower income groups in SMC receive a disproportionately high amount of flood damage, and in fact more flood damage than higher income brackets. This is problematic as these groups are more susceptible to susceptible to cascading impacts that accompany financial shocks.
Figure 8.2 Uncompensated loss and discretionary income for ten income brackets in SMC. The lowest two income brackets have discretionary incomes less than the expected uncompensated AAL due to SLR.
The influence of household income on the ability to respond to flood damages is examined by accounting for the discretionary income of the households in various income brackets. Discretionary income is defined as the income a household has left after paying for taxes and necessities such as food, housing, transportation and utilities. We estimate discretionary income for households in different household income brackets using general Bureau of Labor Services Data on consumer expenditures for the U.S. as a whole (Stoffel, 2015). Figure 8.2 is a graph of discretionary incomes for various income brackets. The two lowest deciles have negative discretionary income. We interpret this result to mean that individuals or households with these income levels are receiving welfare benefits or other aid to compensate for shortfalls in their income.
Figure 8.2 shows uncompensated AAL values together with discretionary income. For the two lowest income brackets, the uncompensated AAL is greater than annual discretionary income. This indicates that households with income less than $25,000 annually may not be able to afford to repair the flood damage to their homes. In total, over 2,200 households may not be able to afford to replace the losses to their residential structures and contents due to flooding. Additionally, for households earning less than $35,000 annually, the difference between their discretionary income and AAL is only $800 on average, which puts these households at a tipping point. Our estimates rely on the conservative RCP 2.6 scenario. In more severe scenarios, we may see households in the next higher income brackets being pushed over this tipping point. If a low-income household chooses to repair their home with their discretionary income, they may have to sacrifice some of their necessities. Alternatively, they might choose to delay repairs and live with flood damage. This strategy may have long-terms costs of health and other social factors, ultimately lowering their quality of life. Note that homeowners in this situation may also be permanently displaced from their homes. These individuals and families may relocate or may become homeless in this scenario. In future research we hope to refine these calculations by relying on data at a more granular level.
An important aspect of our analysis is also being clear about what we were not able to accomplish. One barrier was limited access to high quality, granular data. Our social vulnerability work lacked information at the household level, requiring us to use averages at the block group level. Similarly, the output of electrical substations was given in ranges, not specific values, limiting their accuracy.
The temporal and spatial scope of our investigation also limited our results. The SLR probabilities used in this report were drawn from the conservative IPCC RCP 2.6 scenario and should not be seen as certain. However, our work was limited to a concise planning period of 2020 to 2040 - through which sea level rise scenarios do not greatly diverge. Our decision to limit most of our analysis to one county also limited our results. Many network and indirect effects of sea level rise are not bound by political boundaries. However, due to the limited time and manpower available, we believe that our focus on a single county still provides useful, if imperfect, information on those effects.
The commute disruption modeling effort represents a first look at disruption only, and a number of important issues remain to be explored. Specifically, the impact of flood disruption on vulnerable communities could not be gauged since public transit and vehicle ownership were not incorporated into our model. Additionally, recent changes in commuter behavior such as the use of ridesharing apps were not accounted for comprehensively in our datasets. The impact of such changes in behavior, in addition to the incorporation of public transit, provide important starting points for further analysis. Additional limitations of our analysis, including limitations of input datasets are detailed in Section D.2.3 of the Appendix. Ultimately, we envision this analysis as providing a starting point for future, more refined studies on commute disruption.
Other factors that limited the depth and breadth of this paper were time and resources. While this project involved input from a variety of dedicated individuals in academia, public government, and local communities, there were occasions when we were unable to complete all that we had hoped.
The quantitative risk assessment approach developed through the SUS Project over the last three academic quarters provides a robust and flexible framework for hazard analysis. One focus of further study in the Summer 2018 academic quarter is the continued formalization of this framework into an open-source Python tool which can be used directly by analysts, planners, and policymakers. This tool will build upon the probabilistic methods applied in the course for calculation and aggregation of direct losses, while also adding optional modules for quantification of uncertainty and social equity. We are also interested in incorporating tools which allow planners to compare vulnerabilities associated with competing development scenarios. Building upon our experience with SMC, we are planning a Bay-wide case study which will use our formalized SUS risk framework to quantify direct and indirect losses from sea level rise. We will also explore the role of public-private-academic partnerships in providing the necessary data, funds, and expertise to ensure sustainable and equitable disaster planning.
The Bay Area, and specifically SMC, faces significant risk due to expected rises in sea level. That risk presents itself in the form of $410 million in direct economic losses to structures and their contents from 2020 to 2040. Moreover, these direct losses will be accompanied by a multitude of cascading effects that reflect the complexity inherent in responding to SLR. Interruptions to critical infrastructure systems have the potential to affect residents beyond the immediate flood zones. Likewise, the already congested Bay Area transportation network is easily stressed by closures of key roadways that are susceptible to sea level rise. Businesses may see operations disrupted as commercial properties flood and commuters are unable to get to work. Moreover, these impacts will not affect all groups in the same ways. Certain populations, such as lower-income populations and those with less english proficiency, will have particular difficulties in responding to flood damages. The work presented here was limited in scope and depth due to time and capacity constraints, but it nonetheless provides a useful starting point for future studies into the risks of SLR.
This paper is a product of the Stanford Sustainable Urban Systems Project, a year-long course in which approximately 30 students contributed. First, we would like to our teaching team: Derek Ouyang, Prof. Jenny Suckale, Prof. Leonard Ortolano, and Prof. James Leckie. The project would not have been possible without their constant leadership, feedback, and support. We would also like to thank Avery Bick, the teaching assistant for the course who spent more hours on this project than anyone else. A special thanks must go out to the more than 15 community partners we have engaged with over the course of the year, with particular thanks going to Jasneet Sharma, Hilary Papendick, Darcy Forsell, and Jimmy Vo. Lastly, we would like to thank Prof. Jack Baker and Mariano Balbi in the Stanford Urban Resilience Initiative for developing the initial risk tool that formed the core of this work and for sharing their transportation model.
Adapting to Rising Tides (ART, 2017, April). Regional Sea Level Rise Maps. www.adaptingtorisingtides.org/
AECOM (2016, May). Sea Level Rise & Overtopping Analysis for San Mateo County’s Bayshore.
AECOM (2016, June). City of San Mateo Pump Station and Force Main Condition Assessment.
Baldassari, E. (2017). Bay Area traffic congestion up 80 percent since 2010: Report. Retrieved from https://www.mercurynews.com/2017/09/18/report-traffic-on-major-freeways-has-grown-80-percent-since-2010/
Bay Conservation and Development Commission (BCDC, 2018, January 4). Adapting to Rising Tides Bay Area.
Boeing, G., (2017). OSMnx: New Methods for Acquiring, Constructing, Analyzing, and Visualizing Complex Street Networks. Computers, Environment and Urban Systems 65, 126-139. doi:10.1016/j.compenvurbsys.2017.05.004
Policy statistics. (2018). Bsa.nfipstat.fema.gov. https://bsa.nfipstat.fema.gov/reports/1011.htm
California, State of. “SSO Report Form.” California Integrated Water Quality System, California Environmental Protection Agency, 28 Mar. 2018, ciwqs.waterboards.ca.gov/ciwqs/readOnly/PublicReportSSOServlet?reportAction=criteria&reportId=sso_main.
City of San Mateo Clean Water Program. Nutrient Removal and Wet Weather Flow Management Upgrade and Expansion Project. City of San Mateo, 2017.
County of San Mateo, California State Coastal Conservancy, and Arcadis (2018, March). Sea Level Rise Vulnerability Assessment. Sea Change SMC.
Department of Homeland Security Federal Emergency Management Agency (2010). Multi-hazard Loss Estimation Methodology Model Technical Manual HAZUS MR5. http://deltacouncil.ca.gov/sites/default/files/2015/05/Multi-hazard%20Loss%20Estimation%20Methodology.pdf
Enrique, A., Foster, I., and LaDow, W. (2018, March). Adaptation Planning for Sea Level Rise in San Mateo County: An Examination of 11 Bayside Cities. Stanford Public Policy Senior Practicum Project.
Federal Highway Administration (FHA, 2010, December 14). Commute flows between Bay Area counties.
Federal Emergency Management Agency, (FEMA, 2013). HAZUS Flood Model Technical Manual.
Federal Emergency Management Agency, (FEMA, 2010, December 1). Home Builder's Guide to Coastal Construction. FIMA Risk Reduction. https://www.fema.gov/media-library/assets/documents/6131
Federal Emergency Management Agency, (FEMA, 2017). Policy Information by State. Accessed in February 2018.
Dafermos, Stella. C. and Sparrow, F.T. (1969). The Traffic Assignment Problem for a General Network. J. of Res. of the National Bureau of Standards, 73B, pp. 91-118.
Gary, G., Árvai, J., Cayan, D., DeConto, R., Fox, J., Fricker, H.A., Kopp, R.E., Tebaldi, C., & Whiteman, E.A. (2017). Rising Seas in California: An Update on Sea-Level Rise Science. California Ocean Science Trust.
Graham, M. ,Kutzbach, M., and McKenzie, B., 2014. Design Comparison of LODES and ACS Commuting Data Products. Working Papers 14-38, Center for Economic Studies, U.S. Census Bureau.
Hammond, M.J., Chen, A.S., Djordjevic, S., Butler, D., Mark, O. (2015). Urban Flood Impact Assessment: A State of the Art Review.
Hummel, M. A., Berry, M. S., & Stacey, M. T. (2018). Sea level rise impacts on wastewater treatment systems along the US coasts. Earth's Future, 6(4), 622-633.
Kopp, R. E., Horton, R. M., Little, C. M., Mitrovica, J. X., Oppenheimer, M. , Rasmussen, D. J., Strauss, B. H. and Tebaldi, C. (2014), Probabilistic 21st and 22nd century sea‐level projections at a global network of tide‐gauge sites. Earth's Future, 2: 383-406. doi:10.1002/2014EF000239.
Longitudinal Employer-Household Dynamics (LEHD, 2015). LEHD Origin-Destination Employment Statistics.
MTC/ABAG Analytical Modeling Wiki - Time Periods. (2018). Retrieved from http://analytics.mtc.ca.gov/foswiki/bin/view/Main/TimePeriods
Metropolitan Transportation Commission (MTC, n.d.). Nine Bay Area Counties.
Metropolitan Transportation Commission and Association of Bay Area Governments (MTC & ABAG, 2017, Aug). Data Repository.
Metropolitan Transportation Commission and Association of Bay Area Governments (MTC & ABAG, 2017, July 26). Plan Bay Area 2040.
Milanes, C., Kadir T., Lock, B., Monserrat, L., Pham, N., Randles, K., and California Environmental Protection Agency (2018). Indicators of Climate Change in California.
Miller, M. K. (2014). Seismic risk assessment of complex transportation networks (Doctoral dissertation, Stanford University). https://purl.stanford.edu/hx023kk0983.
Office of Pavement Management & Performance (OPM&P, 2016). Elevation data, California. Caltrans. http://www.dot.ca.gov/hq/tsip/gis/datalibrary/Metadata/Elevation.html
OpenStreetMap Contributors. OpenStreetMap database.
Our Coast Our Future. Northern California Flood Maps.
Pregnolato, Maria, et al. "The impact of flooding on road transport: A depth-disruption function." Transportation research part D: transport and environment 55 (2017): 67-81.
Pregnolato, M., Ford, A., Wilkinson, S. M., & Dawson, R. J. (2017). The impact of flooding on road transport: A depth-disruption function. Transportation research part D: transport and environment, 55, 67-81.
Ritz, C., Edwards, T. L., Durand, G., Payne, A. J., Peyaud, V., and Hindmarsh, R. C. A. (2015). Potential sea-level rise from Antarctic ice-sheet instability constrained by observations.
Nature. 528 115–8.
Rose, Adam, et al. “Business Interruption Impacts of a Terrorist Attack on the Electric Power System of Los Angeles: Customer Resilience to a Total Blackout.” Risk Analysis, vol. 27, no. 3, 2007, pp. 513–531., doi:10.1111/j.1539-6924.2007.00912.x.
Rositer, K. (2011). What are Census Blocks? Census Blogs.
Sarmiento, C., & Miller, T. R. (2006). Costs and consequences of flooding and the impact of the National Flood Insurance Program. Calverton, MD: Pacific Institute for Research and Evaluation.
SimplyAnalytics (2017). EASI US Census 2010 Vehicle Data, SimplyAnalytics database.
Stoffel, B. (2015). The Average American Household's Discretionary Income, by Age and Pay: Where Do You Stand? The Motley Fool. Retrieved 16 March 2018, from https://www.fool.com/investing/general/2015/10/18/the-average-american-households-discretionary-inco.aspx
Sweet, W.V., Kopp, R.E., Weaver, C.P., Obeysekera, J., Horton, R.M., Thieler, E.R. & Zervas, C. (2017). Global and Regional Sea Level Rise Scenarios for the United States. NOAA Technical Report NOS CO-OPS 083. NOAA/NOS Center for Operational Oceanographic Products and Services.
Turner & Townsend (2017). International Construction Market Survey 2017.
U.S. Army Corps of Engineers (USACE, 2003, October 10). Memorandum (EGM) 04-01, Generic Depth-Damage Relationships for Residential Structures with Basements.
U.S. Army Corps of Engineers (USACE, 2009, June 22). Memorandum for Planning Committee of Practice.
U.S. Army Corps of Engineers (USACE, 2011, November 18). San Francisquito Creek – Preliminary Flood Damage Analysis.
U.S. Census Bureau (2013). American Community Survey 2009-2013 Commute Mode Survey Table 2.
U.S. Census Bureau (2016). American Community Survey 2016 1-Year Estimates of Commute Characteristics.
US Census Bureau, C. (2018). OnTheMap Help and Documentation - Longitudinal Employer-Household Dynamics. Retrieved from https://lehd.ces.census.gov/applications/help/onthemap.html#!data_sources
US Department of Commerce. “Bureau of Economic Analysis.” U.S. Bureau of Economic Analysis (BEA), 4 June 2018, www.bea.gov/
U.S. Geological Sciences (USGS, 2011). Global Multi-resolution Terrain Elevation Data 2010 https://pubs.usgs.gov/of/2011/1073/pdf/of2011-1073.pdf
U.S. Census Transportation Planning Package (US CTPP, 2017). Vital Signs: Commute Patterns - Bay Area.
Vista 2035 East Palo General Plan (2016, October 4).
Wang, P., Hunter, T., Bayen, A. M., Schechtner, K., and Gonzalez, M.C. (2012). Understanding road usage patterns in urban areas. Scientific Reports; 2, doi:10.1038/srep01001.