Property-level coastal hazard exposure and buyout-leaseback economic valuation outputs for five California coastal communities (Managed Shores group project of Bren School)
Abstract
This dataset contains property-level coastal hazard exposure and economic valuation outputs from the Managed Shores buyout-leaseback modeling framework, applied to five California coastal communities: Carpinteria, Isla Vista, King Salmon, Pacifica, and Stinson Beach. Data were generated by integrating USGS Coastal Storm Modeling System (CoSMoS) flood inundation and cliff retreat projections with residential property records to estimate optimal managed retreat timing and associated buyout costs under multiple sea level rise scenarios. The dataset includes derived hazard exposure timeseries (annual flood depth and cliff distance estimates by property), economic valuation outputs (optimal retreat year T*, buyout price, government net cost), and sensitivity analysis results across three sea level rise scenarios (Intermediate, Intermediate-High, High) and three government discount rates (2 %, 3 %, 4 %). All outputs are property-level and anonymized to parcel centroids. Monte Carlo storm simulation results are summarized as mean, median, and percentile distributions across 1,000 iterations per property per scenario. These data support research on proactive coastal managed retreat, climate adaptation finance, and coastal hazard economics. They may be of interest to coastal planners, municipal governments, researchers in environmental economics or climate adaptation policy, and state agencies including the California Coastal Commission.
Dataset DOI: 10.5061/dryad.msbcc2gcx
DESCRIPTION OF THE DATA AND FILE STRUCTURE
This dataset contains property-level coastal hazard exposure timeseries and economic valuation outputs generated by the Managed Shores buyout-leaseback modeling framework. Data cover five California coastal communities: Carpinteria, Isla Vista, King Salmon, Pacifica, and Stinson Beach. Outputs span a 74-year planning horizon from 2026 to 2100.
FILES AND VARIABLES
HAZARD EXPOSURE FILES
[site]/cosmos_annual_hazards.csv Annual coastal hazard exposure estimates for each property, interpolated from USGS CoSMoS discrete SLR projections using OPC 2024 sea level rise trajectories. Flood sites (Carpinteria, King Salmon, Stinson Beach) contain depth_m but not cliff_dist_m. Cliff sites (Isla Vista, Pacifica) contain cliff_dist_m but not depth_m. See Missing Values section below.
Columns:
- parcel_id: Unique property identifier
- year: Planning horizon year (1-74, corresponding to 2026-2100)
- scenario: SLR scenario (Intermediate, Intermediate-High, High)
- slr_m: Projected sea level rise (meters above MHHW)
- depth_m: Interpolated flood depth in meters (flood sites only; n/a for cliff sites)
- cliff_dist_m: Distance from property centroid to projected cliff edge in meters (cliff sites only; n/a for flood sites)
[site]/slr_projection_curves.csv Annual sea level rise trajectories for each OPC 2024 scenario (Intermediate, Intermediate-High, High) spanning 2026-2100.
[site]/cosmos_hazard_metrics.csv Property-level hazard extractions at discrete SLR levels directly from CoSMoS rasters. Also contains raw Redfin MLS listing fields carried through from property data processing. These fields are not used in the economic analysis and may contain n/a where listing data were incomplete.
[site]/storm_annual.csv Property-level flood depths under average annual storm conditions (w001, 94 % annual probability) at discrete SLR levels. Flood sites only.
[site]/storm_20yr.csv Property-level flood depths under 20-year storm return period (w020, 5 % annual exceedance probability) at discrete SLR levels. Flood sites only.
[site]/storm_100yr.csv Property-level flood depths under 100-year storm return period (w100, 1 % annual exceedance probability) at discrete SLR levels. Flood sites only.
VALUATION OUTPUT FILES
[site]/retreat_schedule_baseline_[site].csv Property-level optimal retreat timing and buyout cost estimates under baseline parameters (market discount rate = 5 %, rental yield = 5 %, Intermediate-High SLR scenario). Properties where retreat is not triggered within the 2026-2100 planning horizon are assigned n/a for all economic output columns.
Columns:
- parcel_id: Unique property identifier
- scenario: SLR scenario
- retreat_year: Optimal retreat year T* (years)
- mc_mean_year: Mean T* across Monte Carlo iterations (years)
- mc_median_year: Median T* across Monte Carlo iterations (years)
- mc_q05_year: 5th percentile T* (years)
- mc_q25_year: 25th percentile T* (years)
- mc_q75_year: 75th percentile T* (years)
- mc_q95_year: 95th percentile T* (years)
- buyout_price: Present value of rental income stream from year 0 to T* at discount rate 5% (USD)
- leaseback_revenue: Present value of rental income collected by government during leaseback period (USD)
- gov_net_cost: Total public cost (buyout price + NPV repairs minus leaseback revenue) (USD)
- gov_outcome: Categorical government fiscal outcome
- npv_rent_annual: Annual NPV of rental income stream (USD)
- npv_damages_annual: Annual NPV of expected flood damages (USD)
- npv_net_annual: Net annual NPV (rent minus damages) (USD)
- mean_annual_damages_str: Mean annual structure damages across Monte Carlo iterations (USD)
- sd_annual_damages_str: Standard deviation of annual structure damages (USD)
- q05_annual_damages_str: 5th percentile annual structure damages (USD)
- q95_annual_damages_str: 95th percentile annual structure damages (USD)
[site]/retreat_schedule_sensitivity_[site].csv Government-side economics across sensitivity parameter combinations (government discount rate = 2 %, 3 %, 4 %; all three SLR scenarios). Properties where retreat is not triggered within the planning horizon are assigned n/a for all economic output columns.
Columns:
- parcel_id: Unique property identifier
- scenario: SLR scenario
- delta_g: Government discount rate (decimal)
- retreat_year: Optimal retreat year T* (years)
- mc_mean_year: Mean T* across Monte Carlo iterations (years)
- buyout_price: Present value of rental income stream from year 0 to T* at discount rate 5 % (USD)
- npv_repairs: Present value of flood repair costs during leaseback period discounted at delta_g (USD)
- leaseback_revenue: Present value of rental income collected by government discounted at delta_g (USD)
- gov_net_cost: Total public cost (buyout price + NPV repairs minus leaseback revenue) (USD)
[site]/timing_sensitivity_baseline.csv T* distributions across all three SLR scenarios under baseline economic parameters.
[site]/timing_sensitivity_summary.csv Site-level summary of timing sensitivity results.
[site]/sensitivity_summary.csv Site-level summary of government economics across discount rate sensitivity runs.
[site]/mc_distributions_baseline.csv Monte Carlo damage distributions summarized by property and scenario.
[site]/mc_damage_distributions_baseline_[site].csv Full Monte Carlo flood damage distributions (1,000 iterations x property x scenario). Large file.
[site]/mc_tstar_distributions_baseline_[site].csv Full Monte Carlo T* distributions (1,000 iterations x property x scenario). Large file.
[site]/community_stats_baseline.json Site-level summary statistics in JSON format.
MISSING VALUES
Missing values in all CSV files are represented as n/a. Three categories of missing data are present:
- Beyond-horizon properties Affected files: retreat_schedule_baseline_*.csv, retreat_schedule_sensitivity_*.csv, timing_sensitivity_baseline.csv Properties where projected hazard exposure never triggers retreat within the 2026-2100 planning horizon are assigned n/a for all economic output columns including retreat_year, mc_mean_year, mc_median_year, mc_q05_year through mc_q95_year, buyout_price, leaseback_revenue, gov_net_cost, gov_outcome, npv_rent_annual, npv_damages_annual, npv_net_annual, and associated damage distribution columns.
- For cliff sites (Isla Vista, Pacifica): n/a indicates cliff distance never reaches the 10-meter safety threshold by 2100.
- For flood sites (Carpinteria, King Salmon, Stinson Beach): n/a indicates cumulative flood damages never exceed rental income within the planning horizon.
- Incomplete Redfin listing fields Affected files: cosmos_hazard_metrics.csv Columns inherited from Redfin MLS exports including SOLD DATE, HOA/MONTH, DAYS ON MARKET, NEXT OPEN HOUSE START TIME, NEXT OPEN HOUSE END TIME, STATUS, SOURCE, and MLS# are not available for all properties and are not used in the economic analysis. Missing values reflect incomplete listing data in the original MLS records.
- Site type mismatch Affected files: cosmos_annual_hazards.csv
- depth_m is n/a for cliff sites (Isla Vista, Pacifica) where no flood inundation data apply.
- cliff_dist_m is n/a for flood sites (Carpinteria, King Salmon, Stinson Beach) where no cliff retreat data apply.
CODE / SOFTWARE
All data files are provided in CSV format and can be opened with any spreadsheet software (Microsoft Excel, Google Sheets, LibreOffice Calc) or read programmatically in any language supporting CSV import. No specialized software is required to view the data.
The analysis was conducted in R (version 4.4.x) using RStudio. The following packages are required to reproduce the analysis:
Core: tidyverse, dplyr, tidyr, readr, purrr, stringr Spatial: sf, terra, raster, lwgeom, stars, units Data acquisition: tidycensus, rvest, httr Visualization/output: shiny, leaflet, plotly, DT, jsonlite
Analysis workflow - scripts should be run in the following order:
- extract_cosmos_metrics_unified.R — Extracts flood depth, wave height, and cliff retreat metrics from CoSMoS rasters at property centroid locations
- cosmos_process.R / cosmos_process_cliff.R — Processes raw CoSMoS extractions for flood and cliff sites respectively
- redfin_data_code_ms.R — Cleans property records, estimates rental values, and geocodes addresses
- interpolate_cosmos.R — Interpolates discrete CoSMoS SLR snapshots to annual hazard timelines using OPC 2024 projections
- monte_carlo_storms.R — Sourced automatically by the valuation script; defines storm sampling and NPV calculation functions
- cosmos_valuation_v3.R — Main economic model; calculates T*, buyout prices, and government net costs across all scenarios
- cosmos_shiny_v2.R — Launches interactive Shiny dashboard for exploring results
Note: Scripts must be run separately for each of the five case study sites by updating the case_name variable at the top of each script (e.g., case_name <- "carpinteria"). CoSMoS source rasters are not included in this deposit but are publicly available from the USGS CoSMoS data portal.
ACCESS INFORMATION
This dataset was derived from the following publicly available sources:
USGS Coastal Storm Modeling System (CoSMoS) Flood inundation and cliff retreat projections used to generate property-level hazard exposure timeseries. CoSMoS v3.0 Phase 2 (flood sites): https://www.usgs.gov/centers/pcmsc/science/coastal-storm-modeling-system-cosmos CoSMoS v3.1 (cliff sites): https://www.usgs.gov/centers/pcmsc/science/coastal-storm-modeling-system-cosmos License: Public domain (U.S. Government work)
California Ocean Protection Council (OPC) 2024 Sea Level Rise Projections Used to convert discrete CoSMoS SLR snapshots to annual hazard timelines. https://opc.ca.gov/updating-californias-sea-level-rise-guidance/ License: Public domain
Lincoln Institute of Land Policy — Land Value Dataset ZIP code-level land share estimates used to decompose property values into land and structure components. Albouy, D., Ehrlich, G., & Liu, Y. (2018). Housing demand, cost-of-living inequality, and the affordability crisis. Review of Economics and Statistics. License: Contact Lincoln Institute of Land Policy for data use terms — see lincolninst.edu
U.S. Census Bureau American Community Survey (ACS) 2022 Five-Year Estimates Census tract-level demographic data spatially joined to properties for environmental justice analysis. https://www.census.gov/programs-surveys/acs License: Public domain (U.S. Government work)
Federal Housing Finance Agency (FHFA) House Price Index Used to inflation-adjust property values where current market estimates were unavailable. https://www.fhfa.gov/data/hpi License: Public domain (U.S. Government work)
Excluded source data: Raw Redfin MLS property records were used in the analysis but are not included in this deposit due to commercial redistribution restrictions. Researchers seeking to replicate the analysis with current property data should obtain records directly from Redfin or an equivalent MLS data provider.
Associated code repository: https://github.com/willdean7/Managed-Shores License: MIT