Brief communication: Comparing top-down and bottom-up paradigms for global flood hazard mapping

Global floodplain mapping has rapidly progressed over the past few years. Different methods have been proposed to identify areas prone to flooding, resulting into a plethora of freely available products. Here we assess the potential and limitations of two main paradigms, and provide guidance on the use of these global products in assessing flood risk in datapoor regions. 15


Introduction
As economic losses and fatalities caused by floods have dramatically increased over the past decades (Winsemius et al., 2016), there has been much progress in the development of analytical tools for the identification of the areas that can be potentially flooded (Ward et al., 2015;Dottori et al., 2018;Nardi et al., 2019). This progress has also been accelerated by the adoption of the Sendai Framework for Disaster Risk Reduction and the Warsaw International Mechanism for Loss and Damage Associated 20 with Climate Change Impacts (Ward et al., 2015). As such, more and more scientists, experts and practitioners use global floodplain maps in data-poor regions for the identification of flood risk hotspots or the mapping of flood-prone areas (Ward et al., 2015;Winsemius et al., 2016;Dottori et al., 2018;Nardi et al., 2019).

The top-down paradigm
There are two main paradigms to map flooding. The traditional paradigm is (implicitly or explicitly) based on a definition of 25 the floodplain as the area falling within the extent of a given flood event. In this paradigm, which can be seen as top-down, a synthetic event with a given probability of occurrence or return period (Pappenberger et al., 2013;Ward et al., 2015;Dottori et al., 2018), such as the 1-in-200 year flood event, is typically estimated via hydrological modelling or statistical analysis of flood data. This synthetic event is then propagated along the river with hydrodynamic models to estimate the corresponding https://doi.org/10.5194/nhess-2019-418 Preprint. Discussion started: 23 January 2020 c Author(s) 2020. CC BY 4.0 License. inundated areas. The top-down paradigm has been widely used across multiple places and scales (Ward et al., 2015), including 30 large-scale flood hazard modelling in data-poor regions in Africa (Figure 1). While hydrodynamic modelling of floods has been successful in simulating historical events (Horritt and Bates, 2002), large uncertainties come into play when used to simulate synthetic events (Di Baldassarre, 2012). The estimation of a flood hydrograph with a given return period, for example, is extremely uncertain as time series of flood data are hardly ever available, especially in data-poor areas (Blöschl et al., 2013).

The bottom-up paradigm
An alternative paradigm to map flooding is based on a definition of floodplains as distinguished landscape features that have been historically shaped by the accumulated effects of floods of varying magnitudes, and their associated hydrogeomorphic processes (Nardi et al., 2006;Dodov and Foufoula-Georgiou, 2006). In this paradigm, which can be seen as bottom -up, 45 floodplains are identified directly from the topography (Nobre et al., 2011;Samela et al., 2017;Nardi et al., 2019), which is assumed to have been shaped by past flooding events, and building on the concept of fractal river basins (Bras and Rodriguez-Iturbe, 1985;Rodríguez-Iturbe and Rinaldo, 2001) or hydrogeomorphic theories (Bhowmik, 1984;Tarboton et al., 1988). The bottom-up paradigm does not require the estimation of a synthetic flood hydrograph, and consistently identify flood-prone areas across diverse climatic regimes with varying parametrizations (Manfreda et al., 2014;Nardi et al., 2018;Annis et al., 50 2019) which can be seen as an advantage in data-poor regions. Also, with the recent development of global DTMs (Ward et al., 2015;Nardi et al., 2019) and EO-based cloud computing platforms (Pekel, et al., 2016), worldwide mapping of floodplain areas is a reality and these global maps can be derived in a standard PC with a single click and limited computation time.
Hence, it allows to easily detect floodplains, and it is a useful tool for a variety of environmental and socio-economic analyses at large or global scale. 55 International development banks, water sector organizations, national and international bodies mandated with disaster risk reduction, sustainable development and humanitarian response use these global maps in data-poor regions for mapping risk 60 hotspots and flood-prone areas (Ward et al., 2015). To provide guidance in using these global products, we list limitations and advantages of the products derived using the two main paradigms in Table 1.

Comparing top-down and bottom-up paradigms
https://doi.org/10.5194/nhess-2019-418 Preprint. Discussion started: 23 January 2020 c Author(s) 2020. CC BY 4.0 License. More sensitive to data scarcity (time series of flood data are only seldom available and often too short for a robust estimation of a design flood).
Variable over time, e.g. any interventions would require and updating of the hydrodynamic model.
Less sensitive to scales.
Floodplains are defined based on a specific probability of occurrence: this allows cost-benefit analyses for e.g. the design of risk reduction measures is not possible.
It can explicitly account for the role of hydraulic structures, e.g. flood gates.
It provides additional variables, such as maximum flow depth, velocity and volume useful for some applications. More sensitive to scales.
Do not provide a specific probability of occurrence: cost-benefit analyses for the design of e.g. risk reduction measures are not possible.
It cannot account for the role of hydraulic structures, e.g. flood gates.
Scaling laws have limitations in dry climates.
Less sensitive to data scarcity (it does not require any time series).
More consistent over time, e.g. floodplain is identified as if protection structures were not in place. This can be seen as an advantage as erring on the side of least consequences (and total protection is impossible anyway).

Conclusions
Both paradigms are based on consolidated theories, and they have opposite advantages and uncertainties (Table 1). Thus, we argue that these maps are complementary and they should be exploited following the precautionary principle (Foster et al., 2000), which is an important component of much of the environmental legislation in the western world. The principle calls for erring on the side of least consequences. In this context, this means the identification of flood risk hotspots in data-poor areas 70 should consider both flood inundation areas derived by the two paradigms as depicted in the insert of Figure 1.