Rapid mapping of global flood precursors and impacts using novel five-day GRACE solutions

Abstract

Floods affect communities and ecosystems worldwide, emphasizing the importance of identifying their precursors and enhancing resilience to these events. Here, we calculated Antecedent Total Water Storage (ATWS) anomalies from the new 5-day (5D) Gravity Recovery and Climate Experiment (GRACE) and its Follow-On (GRACE-FO) satellite solutions to enhance the detection of pre-flood and active flood conditions and to map post-flood storage anomalies. The GRACE data were compared with ~ 3300 flood events reported by the Dartmouth Flood Observatory (2002-2021), revealing distinct ATWS precursor signals in 5D solutions, in contrast to the monthly solutions. Specifically, floods caused by saturation-excess runoff-triggered by persistent rainfall, monsoonal patterns, snowmelt, or rain-on-snow events-show detectable ATWS increases 15 to 50 days before and during floods, providing a valuable opportunity to improve flood monitoring. These 5D solutions also facilitate a more rapid mapping of post-flood storage changes to assess flood recovery from tropical cyclones and sub-monthly weather extremes. Our findings show the promising potential of 5D GRACE solutions, which are still in the development phase, for future integration into operational frameworks to enhance flood detection and recovery, facilitating the rapid analysis of storage changes relative to monthly solutions.

Figure 1

Pre-event Antecedent Total Water Storage (ATWS) signatures in 5D solutions. The base map shows the ATWS variability from April 17, 2002, to July 1st, 2022, with the climatological mean in Fig. S13. Subplots display ATWS wetness storage as a percentage over a 3-degree area surrounding flood locations for 5D and monthly mascon solutions. The monthly ATWS sums the current month and 50% of the previous month, whereas 5D solutions consider the current 5D period and the weighted prior six 5D intervals (6 × 5 days = 30 days, ~ 1 month equivalent). The equivalent precipitation (P) is shown. Vertical lines mark the start and end of a flood event, determined independently from the DFO catalog. The time series data for ATWS between 2002 and 2022 are shown in Fig. S14. Flood events are annotated with their location, year, flood mechanisms, and impact on livelihoods, including mortality and the number of displaced individuals. Plots were generated using MATLAB.

Figure 2

Utility of 5D in mapping atmospheric moisture transport events (i.e. Hurricane Karina). (a,b) Snapshots of the TWS change for both the monthly and 5D solutions after Hurricane Katrina, which made landfall between August 29th and 30th, 2005. (c) Originating as a low-pressure system in southeastern Bahamas on August 23, 2005, it quickly intensified, impacting southeast Florida, and later escalating to a category 5 hurricane, devastating the Louisiana and Mississippi Gulf Coast on August 29th and 30th, 2005, and dissipating on August 31st eastern Canada. The displayed TWSA is obtained by detrending and removing seasonal variations from the time series, with the August–September TWSA subtracted from that of July 2005. Only positive anomalies are depicted, highlighting the significant influence of hurricanes on water storage. Notably, monthly data exhibit no anomalies, vividly illustrating the initiation, evolution, and dissipation of anomalies along the hurricane trajectory in the 5D solution. All plots were generated using MATLAB.

Figure 3

The precursor coincidence rate (PrCR) of the ATWS preceded or coinciding with 3272 flood events between 2002 and 2021 from the DFO catalog. Four maps illustrate the PrCR at temporal lags (τ) of 5 and 50 days and a tolerance window ($\mathrm{\Delta }$T) prior to and concurrent with the event duration. The PrCR metric represents the proportion of flood days preceded by at least one 5D wet period with an ATWS > 0. A PrCR approaching zero denotes no overlap between the ATWS and flood days, whereas values equivalent to unity (100%) suggest complete overlap. The lower panel displays only those events for the PrCR rate with a 95% statistical significance. An empirical cumulative density function of PrCR, considering various τ and $\mathrm{\Delta }$T, is depicted in Fig. S16. A snapshot of the PrCR concept is shown in the lower right of the figure. Plots were generated using R software.

Figure 4

Relationship between flood attributes (i.e. duration, severity, magnitude, area, mechanism) and precursor coincidence rates: (a) Summary of fixed effects with mean and 95% credible interval (CI). The x-axis represents the slope of the PrCR, reflecting changes in the sign and magnitude of the PrCR relative to a unit change in flood attributes. Positive values indicate increases in PrCR (flood|ATWS) with each unit increase in the predictor, whereas negative values indicate decreases. Further discussion of flood attributes can be found in SI, sect. S4. (b) Interaction of lag or lead time (τ) and tolerance window (ΔT): Darker hues in the heatmap highlight positive PrCR impacts, whereas gray hues suggest limited or negative effects. The bar plot, marked with a 95% CI, reveals log-transformed effects at specific lags, emphasizing the values of shorter ΔT and τ of PrCR. (c) Spatial dependency: Portrays PrCR (flood|ATWS) spatial variance. Each event, marked by its mean and 95% CI signifies locales with an elevated PrCR, indicating the role of ATWS in projecting flood vulnerability. Events with positive values denote higher ATWS associations with flood locations (SI, S3, and S4). We used the same classification of flood mechanisms as reported by the DFO but categorized events such as breaches or glacial melts under ‘Other' because of their infrequent occurrence. Plots were generated using R software .

Figure 5

The response coincidence rate (ReCR) of 3272 flood events associated with detrended deseasonalized positive TWS levels between 2002 and 2021. Four maps display the ReCR at temporal lags (τ) of 5 and 50 days during and after flood events, with a tolerance window ($\mathrm{\Delta }$T). The depicted percentages represent the fraction of positive detrended TWS 5D period that were preceded by at least one day of flooding. The lower panel displays only those events for the PrCR rate with a 95% statistical significance. An empirical cumulative density function of ReCR, considering various τ and $\mathrm{\Delta }$ T is depicted in Fig. S19. A snapshot of the concept of ReCR is shown in the lower right of the figure. Plots were generated using R software.

Figure 6

Response coincidence rate of intense rainfall (85th percentile) and an increase in synoptic TWS from 2004 to 2009 with no gaps in GRACE data. The four maps illustrate the ReCR at lags (τ) of 5 days for the four climate seasons. These percentages represent the fraction of synoptic TWS 5D period adjusted for the linear trend and both annual and semi-annual components, preceded by at least 5D period of intense rainfall. The varying patterns and percentages of the ReCR rate reflect the monthly seasonal extremes in climatology, as illustrated in Fig. S20. A sensitivity analysis using different lags and tolerance windows is presented in the results section. For further explanation, please refer to the methods section. Areas where the ReCR values are significant at 95% confidence interval are marked with black dots. SON September to November; DJF December to February; MAM March to May; JJA June to August. Plots were generated using MATLAB.