1. Geocentric sea-level (GSL) data
We obtained GSL data from coastal satellite altimetry, a product that is freely available from the SEANOE repository (https://doi.org/10.17882/74354). This includes monthly GSL time series (with annual and semi-annual cycles removed) from January 2002 to December 2019 at 756 satellite tracks (satellite altimetry stations) from 20 km offshore to the shoreline. This dataset was originally provided by the Jason 1, 2, and 3 missions and reprocessed by computing high-resolution along-track altimetry ranges, applying an adaptive leading-edge subwaveform retracking method, plus geophysical and environmental corrections and re-estimating the inter-bias missions at the regional scale 22. Four satellite altimetry virtual stations (Chenier, Marsh Island, Breton Sound, and Mississippi Sound) were used in this study, providing rates of GSL change at each data point within the tracks (Extended Data Fig. 1). GSL change at the Grand Isle tide gauge was derived by using the nonlinear vertical land motion (VLM) correction from 17 and subsequently compared with the data points at the four satellite altimetry virtual stations (Extended Data Fig. 2).
2. Relative water-level (RWL) data
RWL data (2009-2021) are available from water-level gauges at 382 monitoring sites throughout coastal Louisiana by means of the Coastwide Reference Monitoring System (CRMS) (https://cims.coastal.louisiana.gov/monitoring-data/). Hourly measurements were converted to daily mean values. The data are provided with respect to the NAVD 88 (Geoid 12A) reference system (geodetic surveys of the water-level gauges were performed in 2014) and collected from open water bodies like bays, bayous, or ponds that are hydrologically connected to the adjacent wetland where surface-elevation data are collected (see below and Extended Data Fig. 3). The water-level gauges are attached to wooden posts, typically with a 4-5 m installation depth below the nearby wetland surface (Extended Data Fig. 4). This setup indicates that besides the geocentric water level, VLM (i.e., subsidence) below the post's installation depth is also captured by these measurements. Hence, they provide RWL measurements. Monthly mean values of RWL were calculated based on daily means, considering only the months with a daily data completeness exceeding 70%. Following this, sites with monthly mean data completeness below 70% were excluded, resulting in a final selection of 325 sites. To remove seasonality from the monthly RWL data, we calculated the long-term average over each calendar month (after detrending the raw RWL data), and then subtracted the long-term average from the corresponding month over the entire study period.
3. Surface-elevation change (SEC) data
SEC is monitored by the aforementioned CRMS network by means of the rod surface-elevation table (RSET). SEC is measured biannually with vertical pins that slide through a horizontal steel arm that is attached to a vertical rod driven to ~ 20 m depth (Extended Data Fig. 4). Geodetic surveys of the RSETs (top of the rod) were performed in 2014, enabling the wetland surface elevation to be converted to the NAVD 88 (Geoid 12A) reference system with the known distance between the top of the rod and the horizontal steel arm. Again, we excluded sites with <70% of data completeness, resulting in 253 sites with both SEC and RWL data.
4. Statistical analysis
Linear trends of monthly GSL, RWL, and biannual SEC data were estimated using ordinary least squares regression. Pairwise correlation analyses were conducted between the residuals (detrended data) of GSL change at Grand Isle and the satellite altimetry-derived GSL changes from 2009 to 2019, as well as between the residuals of GSL change at Grand Isle and the RWL change at the water-level gauges from 2009 to 2021.
To test the statistical significance of the linear trends in monthly GSL and RWL data, and to obtain correlation coefficients, we applied Monte Carlo simulations by assuming the residuals of GSL and RWL data can be explained by an autoregressive process of the order 1 (AR1), and 10,000 red noise time series were generated to simulate the residuals 45. We used the 10,000 synthetic red noise series to calculate the significance of observed linear trends and correlation coefficients. Due to the limited length of SEC records for estimating the AR1 parameters, we generated 10,000 white noise time series instead by assuming the residuals are temporally uncorrelated for all SEC records.
5. Vertical land motion (VLM) correction
All RWL and SEC measurements are affected by VLM. However, the installation depths of the RSETs and water-level gauges are substantially different, resulting in different behaviors between the two instruments. If both instruments are installed in the largely compaction-free Pleistocene basement, no differential VLM correction is required as no shallow sediment compaction is captured by the data, and deep subsidence affects both instruments equally (Extended Data Fig. 4). However, when one or both instruments do not penetrate into the Pleistocene basement (i.e., they are “floating” in the compaction-prone Holocene strata; Extended Data Fig. 4) correction for differential VLM is necessary. Therefore, we extracted installation depths of the RSETs (https://www.lacoast.gov/crms_viewer/Map/CRMSViewer) and used the depth of the Holocene-Pleistocene (HP) interface 46 to determine whether VLM correction is required. For RSETs with unknown installation depth (n=17), we used a value of 20 m which approximates the mean installation depth for all RSETs in this study. For RSETs at locations with unknown depths of the HP interface (n=14, all in the western Chenier Plain) we used the nearest neighbor values.
The mean shallow subsidence rate for the top ~20 m in coastal Louisiana has been reported as 6.8±7.9 mm yr-1 16. Since most compaction occurs in the top 1-3 m 47 and the installation depth of the water-level gauges is approximately 4-5 m, the differential VLM correction between the water-level gauges and the RSETs is likely smaller than the shallow subsidence rate. Here, we assume that the VLM correction between RSETs and nearby water-level gauges follows a normal distribution, where 1 and 4 mm yr-1 correspond to the 5th and 95th percentiles, respectively. We also assumed that these corrections remain constant throughout the study period. We then randomly selected 10,000 VLM corrections from this distribution and subtracted them from the 10,000 linear RWL trends generated previously for each site. aaa
In addition to the linear trends, the RWL elevation, measured with respect to the installation depth of the water-level gauges, is also influenced by VLM. To facilitate further analyses, we applied a VLM correction to the monthly RWL elevations (raw data without seasonal correction) from 2009 to 2021. Since the water-level gauges were surveyed in 2014 (requiring no VLM correction for that year), we adjusted the RWL elevations after 2014 by subtracting the cumulative elevation changes (rate of VLM multiplied by the number of years from 2014). Conversely, we adjusted the RWL elevations before 2014 by adding the cumulative elevation changes. Ultimately, both wetland and water-surface elevations are referenced to the installation depths of the RSETs under the NAVD 88 datum (with deep subsidence being ignored).
6. Wetland response classification
The wetland response to RWL change is classified by examining the relationship between the RWL (after VLM correction) and SEC data, as well as the amount of flooding (Fig. 3). At each site, we determined the monthly RWL as the mean tide, while calculating low tide and high tide by subtracting or adding half the annual tidal amplitude from the mean tide, respectively. Then, for each site, we compared the rates of RWL change and SEC using the 10,000 Monte Carlo simulations. Subsequently, we assessed the flooding condition by comparing the wetland surface elevation with the elevation of low tide (Fig. 4A), mean tide (Extended Data Fig. 5A), and high tide (Extended Data Fig. 5B) during the same month when the SEC data was collected. The classification details are as follows:
-
Give-up (drowning complete): the RWL change rate is likely (³66%) to be higher than the SEC rate, and the wetland is very likely (≥90%) to be flooded during the study period.
-
Give-up (drowning in progress): the RWL change rate is likely (³66%) to be higher than the SEC rate, and the probability of wetland flooding during the study period ranges from 10% to 90%.
-
Give-up (drowning projected): the RWL change rate is likely (³66%) to be higher than the SEC rate, but the wetland is very unlikely (≤10%) to be flooded during the study period.
-
Keep-up (dynamic equilibrium): the RWL change rate is about as likely as not (33-66%) to be higher than the SEC rate, and the probability of wetland flooding during the study period is >10 %.
-
Keep-up (stable equilibrium): the RWL change rate is about as likely as not (33-66%) to be higher than the SEC rate, and the wetland is very unlikely (≤10%) to be flooded during the study period.
-
Catch-up: the RWL change rate is unlikely (≤33%) to be higher than the SEC rate, and the wetland is more likely than not (>50%) to be flooded during the study period.
-
Speed-up: the RWL change rate is unlikely (≤33%) to be higher than the SEC rate, and the probability of wetland flooding is ≤50% during the study period.
In addition to categorizing the wetland response, we calculated the surface elevation deficit (the rate difference between RWL change and SEC) of the wetlands over 10,000 simulations for each site.
7. GSL projections
Future rates of GSL change are available from the IPCC-AR6 Sea Level Projections (http://zenodo.org/record/6382554). Along the Louisiana coast (Lat: 28.8° to 30.6°; Lon: -94° to -88.8°), there are 16 locations with projected rates of GSL rise until 2150 with 10-year increments. We used the mean value by averaging the data from all 16 locations at each 10-year increment until 2150. The 1s confidence intervals of the projected rates are also provided in the dataset.
8. Organic-matter content
Soil properties, including organic-matter content (%) for the top 24 cm, are available from CRMS (https://cims.coastal.louisiana.gov/monitoring-data/) based on shallow cores that were collected at 239 sites in 2018. The mean organic-matter content was calculated by averaging the values for the six 4-cm--increments analyzed in each core. Of the 253 monitoring sites used in the analysis of the SEC-RWL change relationship, only 223 have soil core data, and 122 of these sites are organic-rich (organic-matter content >30%).