Nearshore total water level (TWL) proxies (2018-2100) for Northern California

Same methods as laid out in Erikson and others (2018) with changes for geography of this study region as noted in following steps. Deep-water waves are translated to estimates of nearshore values at designated CSTs using the look-up table (LUT) approach laid out in Erikson and others (2018). Rather than using a spectral wave approach, the LUT was populated by relating nearshore conditions from CDIP with deep-water conditions from wave model data and wave observations at buoys in the region over the years 2000-2017. In the manner described in Erikson and others (2018) observations from four buoys were employed: observations at buoy NDBC 46214 were used to build LUTs for CSTs #8067-9172 (within Marin and Sonoma counties), NDBC 46014 for CSTs #9173-10225 (Mendocino County), NDBC 46213 for CSTs #10226-1121 (Humboldt County), and NDBC 46027 for CSTs #11222-11594 (Del Norte County). Significant wave heights were binned from 0.5-10.25 m at 0.25 m intervals; peak wave periods from 3-24 s at 3 s intervals; peak wave directions from 5-360 degrees at 5 degree intervals; and wind speeds from 0-24 m/s at 6 m/s intervals. Interval sizes for H_s and T_p were based on the average RMSE for each variable. For each combination of deep-water H_s, T_p, D_p, and U, time indices falling into each bin were identified. For each nearshore location, median H_s, T_p, T_m, D_p, and D_m corresponding to all time indices of a given set of deep-water binned conditions were computed to complete the lookup table. Because swell travel time from offshore to nearshore is on the order of 1.5 hours (assuming an average depth of 100 m and T_p = 15 s over a distance of approximately 120 km) and the model outputs are at three-hourly intervals, no time lag is assumed between deep-water and nearshore conditions. The LUTs, developed from the Jan 2000 through 2017 wave observation and CDIP data, were then used to generate bulk wave statistic time-series at each of the CSTs for the 2018-2100 time period, and an additional hindcast period (1979-1999) for a longer historical period. Wave model data for projected waves and winds from GCM data, representing the GFDL-ESM2M RCP4.5 climate scenario (Dunne and others, 2012), were used in conjunction with the LUT as described by Erikson and others (2018). Wave statistic time series for the historical period were also generated using information from historical periods of GCM data.

An estimation of the storm surge (SS) term for TWL (equation 3 in Erikson and others, 2018) was generated using wind and pressure data from GCM data and previously generated wave statistic time series. Coefficients were found by obtaining least-square residual best fits for data spanning 1979 through Dec 2017 of overlapping nearshore hindcast data and tide gauge water level measurements at regional tide stations (San Francisco, Point Reyes, Arena Cove, North Spit, and Crescent City). The coefficients for each tide station ranged as follows: -0.0139 - 0.1321 for C0, 0.1019 - 0.1865 for C1, and 1.3 - 1.9 for C2. An SS time series was generated for each tide station, and nearshore SS at each CST was interpolated between the stations (at each time step), owing to the variation in shelf width and coastal exposure at the tide stations compared to study region in Erikson and others (2018).

Sea level anomaly (SLA) time-series were derived from observations for the hindcast period and from a simple linear model that used sea-surface temperature anomalies (SSTA) as the independent variable for the future period (2017-2100). For the hindcast time-period, monthly mean sea levels from NOAA were prepared with the average seasonal cycle and linear trends removed, for all tide stations listed in the previous step. The coefficients (equation 4, in Erikson and others, 2018) were developed from mean SSTA observations at the San Francisco buoy (NDBC 46237); SSTAs were computed by subtracting out the seasonal signal and long-term means (1971-2000; Reynolds and others, 2002) from the satellite-derived SST time-series for 1981-2014. Coefficients C0 and C1 were found to equal 0.033758 and 0.095694, respectively, by a least squares linear fit through the upper envelope of the mean monthly SSTA and SLA. The upper envelope SLA was defined by the maximum SLA within 0.25-degree SSTA bins from -3.0 deg C to +3.0 deg C. A fit through the upper envelope, rather than all the data, errs conservatively high by assuring a positive SLA for higher SSTAs. Due to scatter in the data and relatively small SLAs, a fit through all the data would yield only a slight positive SLA for the maximum observed SSTA, which is well below observed extremes. As in the previous step, SLA time series were generated at each tide station using the determined coefficients and GCM data, and values were interpolated alongshore at CSTs at each time step.

An estimation of runup (R2%; equation 2 in Erikson and others, 2018) was made for the study region using the nearshore wave statistics at each CST. A representative slope of 0.066 was used for Northern California. TWL proxies were generated for each CST by summing all SS, SLA and R2% contributions, per equation 1 in Erikson and others (2018). These proxy values can functionally be referenced as elevations above Mean Sea Level (MSL). Data were prepared for packaging into NetCDF format. Proxy values are identified and organized by CST location. Time units are in decimal days since January 01, 1900. See NetCDF attributes for variables for more information.