Introduction
Fluvial floods are destructive hazards that affect millions of people worldwide each year. Forecasting flood events and their potential impacts is crucial for effective disaster preparation and mitigation. Modeling flood inundation based on extreme value analysis of river discharges offers an alternative to computationally expensive physical models of flood dynamics.
This paper presents the implementation of a globally applicable, open-source fluvial flood model within a state-of-the-art risk modeling framework. By using openly available data, the model can rapidly compute flood inundation footprints of historic and forecasted events to estimate associated socio-economic impacts. The model is applied to the example of Pakistan, where it is used to assess flood depths, extents, and population displacement.
The United Nations Office for Disaster Risk Reduction’s (UNDRR) Sendai Framework for Disaster Risk Reduction emphasizes the need for a better understanding of disaster risk to increase resilience and employ adaptation and mitigation measures (UNDRR, 2015). To this end, decision-makers and humanitarian actors require accurate risk assessments, early warning systems, and impact forecasts of imminent events.
Disaster displacement, the temporary or permanent relocation of people from their homes due to natural hazards, causes severe humanitarian impacts. Displaced individuals often face challenges in accessing adequate shelter, food, clean water, healthcare, and long-term support. Children are particularly at risk of exploitation, abuse, malnutrition, and disease (UNICEF, 2023). The year 2022 saw a tragically high number of displacements, with more than 32 million internal displacements due to natural hazards, the majority of which were caused by floods (IDMC, 2023a).
Fluvial, pluvial, and coastal flooding often exhibit complex interactions, and merging these into a comprehensive modeling approach is an ongoing effort (Loveland et al., 2021; Eilander et al., 2023). Physical models for fluvial floods already require an elaborate model cascade, which may not be suitable for all applications due to computational demands (Winsemius et al., 2013).
Extreme value analysis of river discharge offers an alternative approach to relate predicted and past events. However, the limited historical time series of global climate models and river discharge reanalysis datasets can lead to uncertain estimates for extreme events with large return periods (Hirabayashi et al., 2013; Willner et al., 2018). Retaining inundation information in the analysis can help estimate economic damage and population displacement caused by future flood events (Sauer et al., 2021; Kam et al., 2021).
Flood Model
The proposed flood model computes a flood inundation footprint from gridded, geo-located river discharge data via an extreme value analysis. In a pre-processing step, the historical time series of discharges is analyzed for each grid point by fitting a Gumbel distribution. This distribution is then used to compute a return period, which relates the discharge input to the historical time series. Optionally, information on flood protection standards can be applied.
The return period is used to look up flood depths in flood hazard maps. The Global Flood Awareness System (GloFAS) provides global data on river discharge (Alfieri et al., 2013; Harrigan et al., 2023), while flood hazard maps displaying inundation extents for specific return periods are available from the European Commission’s Joint Research Centre (Dottori et al., 2016b).
The flood model can be applied to river discharge (ensemble) forecasts and reanalysis alike, computing flood inundation maps for entire countries in a few minutes. It is implemented as a Python module of the risk model CLIMADA, which serves as a platform for both climate risk assessment (Aznar-Siguan and Bresch, 2019) and impact-based forecasting (Rööosli et al., 2021).
Pre-processing: Extreme Value Analysis
In this pre-processing step, an extreme value analysis is applied to the historical discharge data. A right-handed Gumbel distribution is fitted to the yearly maximum of the discharge time series at every location independently, using the “SciPy” Python package (Virtanen et al., 2020).
The fitted parameters (μ and β) are then used to compute an equivalent return period for discharge input data at every location. The complementary probability of the cumulative distribution function of the fitted Gumbel distribution gives the exceedance frequency, the inverse of which is the return period of the event.
Regridding and Flood Protection
The spatial resolution of the GloFAS discharge data (0.1°) and the flood hazard maps (30”) differ significantly. The return period data are therefore regridded onto the grid of the flood hazard maps using bilinear interpolation.
The FLOPROS database contains data on return periods associated with modeled flood protection standards (Scussolini et al., 2016). This effect of protection measures can be considered by setting the return period to zero if it is lower than the protection standard at the same location.
Flood Depth Interpolation
The flood footprint is created by interpolating a flood depth value from the flood hazard maps. The flood hazard maps define a scalar field z with three-dimensional coordinates: the location x’ and the return period r. The flood depth at a particular return period rx’ at location x’ is interpolated linearly in the return period dimension.
Implementation
CLIMADA is an impact model that represents hazards, exposure, and vulnerability in a spatially explicit manner (Aznar-Siguan and Bresch, 2019). The presented flood model is implemented as a Python module of CLIMADA, producing one or multiple Hazard objects (the CLIMADA data structure of a geophysical hazard) from the discharge input data.
Users can state which GloFAS discharge data to download, and the module will compute a CLIMADA hazard from these data using default settings. Alternatively, users can adjust each step of the model pipeline individually.
The module benefits from parallel execution on multiple processors where applicable. Some tasks, such as the time series analysis, are performed during a one-time setup, while the rest of the computation pipeline uses Xarray multithreading to maximize performance.
Case Study: Pakistan
Pakistan is a flood-prone country and is particularly vulnerable to the effects of climate change (Eckstein et al., 2021). In summer 2022, Pakistan experienced its arguably most devastating floods to date, affecting over 30 million people and leaving more than 20 million in need of humanitarian assistance (OCHA, 2022b).
Flood Extent Comparison
We computed a flood footprint with our model and compared its spatial extent with satellite observation data. The Humanitarian Data Exchange (HDX) hosts a dataset published by the United Nations Satellite Centre (UNOSAT), which is based on observations of the Visible Infrared Imaging Radiometer Suite (VIIRS) instruments aboard the NOAA-20 satellites (UNOSAT, 2022).
We considered two separate model runs: one which applied flood protection standards as listed in the FLOPROS database and one which did not account for any flood protection. The model agreement is generally good, especially in the area around the city of Larkana. However, the flood model fails to capture observed floods in some regions and predicts extensive flooding where little was observed.
Binary classification metrics reveal that including FLOPROS protection levels does not clearly improve the estimated flood extent. While the specificity and precision increase, the recall decreases significantly. This indicates that the FLOPROS database tends to overestimate flood protection levels, causing the model to underestimate flood extents and severity.
Displacement Modeling and Calibration
We used data on displacement in Sindh Province during the 2022 floods to calibrate impact functions in our model. According to reports by the Provincial Disaster Management Authority (PDMA), the maximum number of about 7.3 million displaced people was reached by 30 September 2022 (PDMA, 2022).
We employed a “cross-calibration” method, using multiple subsets of the data to calibrate an ensemble of impact functions. This approach avoids overfitting and allows us to investigate the uncertainty in the calibrated vulnerability.
The calibrated impact function parameters indicate that the model without flood protection needs to estimate a lower impact per location to match the same displacement data. The median impact threshold is higher, and the median percentage of displaced population is lower than for the model with FLOPROS protection.
Historical Displacement Estimates
We used the calibrated impact models to estimate historical flood impacts in Pakistan. Comparing the model outputs to reports of flood disasters and associated displacement reveals that the No Protection model estimates a “baseline” of 1,000 to 10,000 displaced people each month, while the FLOPROS model predicts no displacement for most months.
High-impact flood events can be clearly distinguished from the baseline in both models, and the order of magnitude for displaced population matches the reported numbers. The timings of events estimated by the models also align well with the reported data.
Impact-Based Forecasting
Finally, we applied the flood impact model to a GloFAS river discharge forecast for July 2023, computing an impact-based forecast. The No Protection model predicts higher impacts overall, with a median displacement of 195,957 people compared to 32,142 for the FLOPROS model.
A sensitivity analysis reveals that the FLOPROS model is more sensitive to the variation in the hazard forecast and less sensitive to the vulnerability than the No Protection model. This is because the most critical information for the FLOPROS model is whether the protection level is exceeded or not.
Conclusion
We presented a model for mapping river flood inundation footprints to GloFAS river discharge data, offering a globally applicable and computationally efficient alternative to physical flood dynamics models. The model is readily implemented in the CLIMADA risk modeling framework and was applied to estimate population displacement due to river floods in Pakistan.
The results show that the model performs well in terms of countrywide numbers, matching disaster reports. However, significant differences were found on the district level between the calibrated model impact and reported displacements. The model’s strengths lie in estimating overall event impacts and identifying spatial hotspots, rather than in small-scale flood dynamics analysis.
Incorporating estimates of flood protection standards from the FLOPROS database changes flood footprints significantly, but its effects on overall model performance remain inconclusive. We suggest using both the No Protection and FLOPROS model versions to estimate “worst-case” and “best-case” scenarios, as we cannot state a general range of risk associated with the two.
A sensitivity analysis revealed that the statistical uncertainty within the model is negligible compared to the uncertainty represented in the GloFAS river discharge forecast and the cross-calibrated impact functions. Dissecting the overall uncertainty in the estimated impact into sensitivity coefficients for each input parameter provides crucial information for decision-makers, as major sources of uncertainty can be identified.
Further work on this flood model and its overall approach should focus on operationalizing early event detection and classification, supporting humanitarian organizations and stakeholders in anticipatory action and decision-making.
References
Alfieri, L., Burek, P., Dutra, E., Krzeminski, B., Muraro, D., Thielen, J., and Pappenberger, F.: GloFAS – global ensemble streamflow forecasting and flood early warning, Hydrol. Earth Syst. Sci., 17, 1161–1175, https://doi.org/10.5194/hess-17-1161-2013, 2013.
Aznar-Siguan, G. and Bresch, D. N.: CLIMADA v1: a global weather and climate risk assessment platform, Geosci. Model Dev., 12, 3085–3097, https://doi.org/10.5194/gmd-12-3085-2019, 2019.
Dottori, F., Salamon, P., Bianchi, A., Alfieri, L., Hirpa, F. A., and Feyen, L.: Development and evaluation of a framework for global flood hazard mapping, Adv. Water Resour., 94, 87–102, https://doi.org/10.1016/j.advwatres.2016.05.002, 2016b.
Eckstein, D., Künzel, V., and Schäfer, L.: Global Climate Risk Index 2021, Germanwatch e.V., https://reliefweb.int/report/world/global-climate-risk-index-2021, 2021.
Eilander, D., Couasnon, A., Leijnse, T., Ikeuchi, H., Yamazaki, D., Muis, S., Dullaart, J., Haag, A., Winsemius, H. C., and Ward, P. J.: A globally applicable framework for compound flood hazard modeling, Nat. Hazards Earth Syst. Sci., 23, 823–846, https://doi.org/10.5194/nhess-23-823-2023, 2023.
Harrigan, S., Zsoter, E., Cloke, H., Salamon, P., and Prudhomme, C.: Daily ensemble river discharge reforecasts and real-time forecasts from the operational Global Flood Awareness System, Hydrol. Earth Syst. Sci., 27, 1–19, https://doi.org/10.5194/hess-27-1-2023, 2023.
Hirabayashi, Y., Mahendran, R., Koirala, S., Konoshima, L., Yamazaki, D., Watanabe, S., Kim, H., and Kanae, S.: Global flood risk under climate change, Nat. Clim. Change, 3, 816–821, https://doi.org/10.1038/nclimate1911, 2013.
IDMC: 2023 Global Report on Internal Displacement, Internal Displacement Monitoring Centre (IDMC), https://www.internal-displacement.org/publications/2023-global-report-on-internal-displacement, 2023a.
Kam, P. M., Aznar-Siguan, G., Schewe, J., Milano, L., Ginnetti, J., Willner, S., McCaughey, J. W., and Bresch, D. N.: Global warming and population change both heighten future risk of human displacement due to river floods, Environ. Res. Lett., 16, 044026, https://doi.org/10.1088/1748-9326/abd26c, 2021.
Loveland, M., Kiaghadi, A., Dawson, C. N., Rifai, H. S., Misra, S., Mosser, H., and Parola, A.: Developing a Modeling Framework to Simulate Compound Flooding: When Storm Surge Interacts With Riverine Flow, Front. Climate, 2, 609610, https://doi.org/10.3389/fclim.2020.609610, 2021.
OCHA: Pakistan: 2022 Monsoon Floods, Situation Report 9, United Nations Office for the Coordination of Humanitarian Affairs (OCHA), https://reliefweb.int/report/pakistan/pakistan-2022-monsoon-floods-situation-report-no-9-14-october-2022, 2022b.
PDMA: Daily Situation Report, PDMA (SINDH)/(SITREP)/2022/1239, Provincial Disaster Management Authority (PDMA), Rehabilitation Department, Government of Sindh, https://pdma.gos.pk/Documents/Flood/Flood_2022, 2022.
Rööosli, T., Appenzeller, C., and Bresch, D. N.: Towards operational impact forecasting of building damage from winter windstorms in Switzerland, Meteorol. Appl., 28, e2035, https://doi.org/10.1002/met.2035, 2021.
Sauer, I. J., Reese, R., Otto, C., Geiger, T., Willner, S. N., Guillod, B. P., Bresch, D. N., and Frieler, K.: Climate signals in river flood damages emerge under sound regional disaggregation, Nat. Commun., 12, 2128, https://doi.org/10.1038/s41467-021-22153-9, 2021.
Scussolini, P., Aerts, J. C. J. H., Jongman, B., Bouwer, L. M., Winsemius, H. C., de Moel, H., and Ward, P. J.: FLOPROS: an evolving global database of flood protection standards, Nat. Hazards Earth Syst. Sci., 16, 1049–1061, https://doi.org/10.5194/nhess-16-1049-2016, 2016.
UNDRR: Sendai Framework for Disaster
