Research ArticleEARTH SCIENCES

Tracking the weight of Hurricane Harvey’s stormwater using GPS data

See allHide authors and affiliations

Science Advances  19 Sep 2018:
Vol. 4, no. 9, eaau2477
DOI: 10.1126/sciadv.aau2477
  • Fig. 1 GPS motions during migration of Harvey.

    GPS stations (black triangles; top), with their corresponding stacked time series illustrating their motions (bottom) from different areas. Top: Path of hurricane Harvey is plotted as yellow line with position of eye as blue dots [at noon in universal time conversion (UTC)]. Bottom: Yellow shaded region bordered by vertical red line and dashed line demarks timing of Harvey’s initial and second landfall, respectively. Stations around Houston (blue box; top) show an average initial subsidence of 8 mm, followed by a rapid and then prolonged uplift over ~5-week period (blue line; bottom). Stations around west Louisiana (red box; top) show a delayed and subdued subsidence corresponding to second landfall of Harvey (red line; bottom).

  • Fig. 2 Comparison of filtered vertical GPS data with model predictions over time.

    First and third rows (A to E and K to O) show GPS data after ICA filtering of select days, and second and bottom rows (F to J and P to T) are model prediction (negative is subsidence). GPS motions show clear subsidence in southwest Texas and migration toward the Texas-Louisiana border coincident with position of Harvey (path and eye plotted as red line and dot, respectively). After Harvey dissipated on 1 September (L), subsidence gradually decreases over a ~5-week period.

  • Fig. 3 Comparison of accumulation and dissipation of TWS with observed precipitation across the Gulf Coast over time.

    First and third rows (A to E and K to O) show TWS as estimated from inverting the GPS motions each day (Fig. 2), depicted here as equivalent water thickness (in meters) inferred for each grid node. Total water storage volume and 1σ uncertainty are labeled in the lower left of each plot in blue, with upper value derived from the NLDAS model and GPS inversion estimate underneath. Precipitation in second and fourth rows (F to J and P to T) estimated from satellite and rain gauges (20). The lower left value in red shows total precipitation volume deposited for each day (not cumulative over time), which is a flux, as opposed to the TWS values that are an integrative quantity in time. Harvey path and position are plotted as red line and dot, respectively. TWSs for other days are found in fig. S21. Corpus Christi and Houston are shown by green and red squares, respectively.

  • Fig. 4 Water volume changes from Hurricane Harvey.

    (A) Hydrograph illustrating total water volume changes (positive is accumulation, and negative is dissipation) with time for different components of the hydrologic system. Total TWS across the entire domain estimated from GPS (black line), simulated TWS from NLDAS hydrologic model (green line) (18), NOAA stage IV cumulative precipitation (dark blue line) (20), cumulative river discharge (orange line; see Methods and fig. S13), contribution of cumulative ET after initial landfall (dashed magenta line) (21, 24), cumulative NLDAS surface runoff and groundwater flow (SG; dashed cyan line), and our estimated cumulative surface runoff and groundwater flow SG from closing the water budget (cyan line), with the separated cumulative groundwater flow in red. Green vertical solid line in all figures marks first landfall of Harvey on 26 August, and dashed line marks second landfall on 30 August. (B) GPS vertical motions (solid blue line; with station ID in bottom left) shows close agreement with changes of water as measured from a nearby wellhead levels located northeast of Houston (orange line; scale on right), suggesting that GPS motions reflect effects of water loading and not calculated poroelastic effects (dashed blue line; scale to left; see fig. S6 and Methods). (C) Comparison of water volume at the Barker and Addicks reservoirs estimated from GPS (black) and daily U.S. Geological Survey (USGS) reservoir provisional water volume data from water level gauges (red).

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/4/9/eaau2477/DC1

    Materials and Methods

    Text S1. Analysis of CME

    Text S2. Smoothing parameters

    Text S3. Synthetic test

    Fig. S1. Estimating CME from subset of stations.

    Fig. S2. Comparison of CME estimate from two regions of GPS network illustrating difference in degree of mixing of the hydrological signal.

    Fig. S3. ICA analysis of vertical positions after CME removal.

    Fig. S4. ICA analysis of horizontal positions (north component) after CME removal.

    Fig. S5. ICA analysis of horizontal positions (east component) after CME removal.

    Fig. S6. Comparison of vertical motions from GPS and expected poroelastic rebound.

    Fig. S7. Comparison of vertical motions from GPS and predicted from hydrologic (NLDAS) model.

    Fig. S8. Comparison of north motions from GPS and predicted from hydrologic (NLDAS) model.

    Fig. S9. Comparison of east motions from GPS and predicted from hydrologic (NLDAS) model.

    Fig. S10. Percent of variance reduction (POVR) for different temporal and spatial smoothing values.

    Fig. S11. Accumulation and dissipation of a synthetic water volume source as a function of time in days.

    Fig. S12. Inversion result of synthetic test.

    Fig. S13. River discharge and cumulative water profiles.

    Fig. S14. GPS vertical positions processed using rapid orbits to assess use for near–real-time applications.

    Fig. S15. ICA filtering of GPS solutions processed using rapid orbits, which were available with 1-day latency.

    Fig. S16. Raster matrix highlighting space-time distribution of missing data points for all GPS time series.

    Fig. S17. Components of spatial and temporal variations of GPS data separated using ICA.

    Fig. S18. Model uncertainty (1σ) of each grid node in water volume (km3).

    Fig. S19. Daily simulated TWS from NLDAS hydrologic model.

    Fig. S20. Soil moisture measurements from NASA’s SMAP satellite.

    Fig. S21. TWS inversion estimates for rest of days not shown in Fig. 3.

    Data file S1. GPS raw and filtered time series positions shown in Figs. 1 and 2.

  • Supplementary Materials

    The PDF file includes:

    • Materials and Methods
    • Text S1. Analysis of CME
    • Text S2. Smoothing parameters
    • Text S3. Synthetic test
    • Fig. S1. Estimating CME from subset of stations.
    • Fig. S2. Comparison of CME estimate from two regions of GPS network illustrating difference in degree of mixing of the hydrological signal.
    • Fig. S3. ICA analysis of vertical positions after CME removal.
    • Fig. S4. ICA analysis of horizontal positions (north component) after CME removal.
    • Fig. S5. ICA analysis of horizontal positions (east component) after CME removal.
    • Fig. S6. Comparison of vertical motions from GPS and expected poroelastic rebound.
    • Fig. S7. Comparison of vertical motions from GPS and predicted from hydrologic (NLDAS) model.
    • Fig. S8. Comparison of north motions from GPS and predicted from hydrologic (NLDAS) model.
    • Fig. S9. Comparison of east motions from GPS and predicted from hydrologic (NLDAS) model.
    • Fig. S10. Percent of variance reduction (POVR) for different temporal and spatial smoothing values.
    • Fig. S11. Accumulation and dissipation of a synthetic water volume source as a function of time in days.
    • Fig. S12. Inversion result of synthetic test.
    • Fig. S13. River discharge and cumulative water profiles.
    • Fig. S14. GPS vertical positions processed using rapid orbits to assess use for near–real-time applications.
    • Fig. S15. ICA filtering of GPS solutions processed using rapid orbits, which were available with 1-day latency.
    • Fig. S16. Raster matrix highlighting space-time distribution of missing data points for all GPS time series.
    • Fig. S17. Components of spatial and temporal variations of GPS data separated using ICA.
    • Fig. S18. Model uncertainty (1σ) of each grid node in water volume (km3).
    • Fig. S19. Daily simulated TWS from NLDAS hydrologic model.
    • Fig. S20. Soil moisture measurements from NASA’s SMAP satellite.
    • Fig. S21. TWS inversion estimates for rest of days not shown in Fig. 3.

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • Data file S1 (.zip format). GPS raw and filtered time series positions shown in Figs. 1 and 2.

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article