There were several factors that contributed to the selection of the indicators that comprised each of the hazard, exposure, and vulnerability inputs that are used to calculate the overall FRI. Firstly, relevant prior literature was thoroughly consulted and strongly informed the inputs that were desired to be used. Secondly, characteristics of this study area and the Australian environment were also taken into account. For example, SM was employed as one of the indicators for the hazard index, as it was deemed to be an important representative of antecedent conditions across various hydrological literature. Data availability issues became an additional factor which limited some inputs from being used in this research. Given that a key element of novelty in this proof-of-concept FRA is the simplicity of the indices, risk components were limited to only 3–4 inputs.
Flood Hazard Indicators and Data Collection
This research defined flood hazard as an adaptation of the IPCC’s definition, being: “the potential occurrence of a natural or human-induced physical flood event or trend that may cause loss of life, injury, or other health impacts, as well as damage and loss to property, infrastructure, livelihoods, service provision, ecosystems and environmental resources” [
22].
One of three flood hazard indicators was the Maximum 3-Day Precipitation input. This indicator is quantified as the maximum amount of rainfall received in a given 3-day period. Satellite-based rainfall estimates were used to provide data for this indicator. Rain gauge estimates would also be applicable; however, these are limited to the locations of gauge installations and do not always have a homogenous spatial distribution, which means that the use of satellite data makes this indicator more replicable to other study areas. FRAs using index methods frequently use a form of rainfall indicator, as rainfall is logically a key modulator of pluvial and fluvial flooding events. Some earlier studies, such as [
44,
45] quantify rainfall in their FRAs using annual rainfall; however, in this study, a 3-day period was chosen given how much more important rainfall over a shorter time frame is in a flooding context. This 3-day precipitation concept is observed sparsely in other literature, such as by [
46]; however, it has not been applied in an Australian context. Ultimately, this 3-day period was chosen as a compromise to cover both single-day pluvial floods and multi-day fluvial flooding.
It is acknowledged that radar data are considered the most accurate precipitation estimates over large areas. However, in absence of radar data, precipitation estimates from satellite remote sensing provide an alternative to gauge- or radar-based measurements with greater spatial coverage and improving accuracy. Although meteorological services keep rain gauge records usually dating back decades, satellite precipitation data can complement and potentially improve conventional precipitation records, leading to an improved ability to place extreme precipitation events within a climatological context. This leads to better heavy precipitation and drought monitoring, amongst numerous other applications.
For this study, satellite precipitation data were obtained from the World Meteorological Organization’s (WMO) Space-based Weather and Climate Extremes Monitoring (SWCEM) [
47]. For the SWCEM, satellite precipitation datasets were provided by the Japan Aerospace Exploration Agency (Global Satellite Mapping of Precipitation, or GSMaP) and the U.S. National Oceanographic and Atmospheric Administration (Climate Prediction Center morphing technique, or CMORPH). The SWCEM datasets for the East Asia and Western Pacific region (50° E–120° W; 40° N–45° S) are available from the WMO SWCEM website (
https://public.wmo.int/en/programmes/wmo-space-programme/swcem accessed on 13 January 2023). A comprehensive analysis of the SWCEM satellite precipitation data performed over Australia showed that blended satellite-gauge products had higher correlations and smaller errors than gauge analysis [
48,
49,
50,
51,
52,
53]. Similarly, earlier studies demonstrated the usefulness of satellite precipitation data not only for Australia but also for countries in the Pacific region [
54,
55], which means that the SWCEM data as well as the developed in this study flood risk assessment methodology could potentially be used in other countries in the Asia-Pacific region.
Evaluating satellite precipitation estimates, earlier studies found that the Global Satellite Mapping of Precipitation (GSMaP) dataset had high performance over Australia [
48,
49,
50,
51]. GSMaP uses the Global Precipitation Mission (GPM) constellation and NOAA Climate Prediction Center 4 km infrared product [
56,
57]. The version used in this study is the Gauge Near-Real-Time (GNRT) version where the Climate Prediction Center daily global dataset of rain gauges is used to calibrate the GSMaP estimates by roughly matching their values across the past 30 days. Based on extensive validation of GSMaP data over Australia conducted in earlier studies [
48,
49,
50,
51] and conclusions that is it a high performing dataset, in this study, using GSMaP data, 3-day precipitation estimation periods from the 17th to the 26th of March 2021 (the length of the flood event; see
Appendix D for detail) were calculated, and the highest 3-day precipitation total at each specific grid cell (0.1 degrees resolution) was used for the indicator. The resultant input was a gridded dataset of 3-day accumulation values from a combination of date ranges, depending on which was the highest at each point.
The second of three flood hazard indicators used was Distance to River—Elevation-Weighted (DREW). This input quantifies the distance of any point in the study area to the nearest river. This is an important metric in an FRA context because of how strongly the river locations modulate the locations of flooding in fluvial and pluvial scenarios. Particularly in the HNC, the proximity of a given point to the lower-elevated, downstream Hawkesbury River areas corresponds strongly to flooding outputs. This is the motivation for combining a typical Distance to River input (seen commonly in FRAs) (e.g., [
58,
59]) with an elevation layer, to capture that lower-elevated areas are more likely to flood after an extreme rainfall event, as opposed to weighting all river areas at different elevation levels equally.
This input was created by calculating the distance of each grid point to the nearest river line using an ‘Euclidean Distance’ function. Then, this layer was multiplied with an elevation layer to account for the influence of elevation. The resulting input described the distance of each point to the nearest river, whilst being weighted as more hazardous if the elevation were lower.
Finally, Soil Moisture was the third input used for the flood hazard quantification. SM is considered an important modulator and representative of the antecedent conditions of an environment. SM has been found to have a direct influence on Australian flood timing, particularly in southern Australia [
60]. Other recent literature notes the consensus among Australian research that consideration of changes in the antecedent conditions are crucial to predictive flood modelling [
61,
62,
63]. As thus, SM prior to the case study flooding event, being representative of the antecedent conditions, was quantified in this FRA.
The SM data were collected from the BoM’s Australian Water Resources Assessment Landscape model (AWRA-L), which is a daily gridded SM dataset measuring absolute moisture content in the root zone (0–1 m) [
64]. SM content is measured as the amount of water in the soil as a volumetric quantity, for example, a SM percentage of 50% would mean that half of the soil’s total water carrying capacity has been filled. To represent the antecedent conditions of this study area, a 7-day SM average was taken prior to the beginning of the flood event (11 March 2021–17 March 2021).
Flood Exposure Indicators and Data Collection
This research utilised the IPCC’s flood exposure definition as “the presence of people; livelihoods; services and resources; infrastructure; or economic, social, or cultural assets in places and settings that could be adversely affected by a flood event” [
22]. Three indicators were chosen to capture flood exposure in the HNC for this study: population density, land use type, and critical infrastructure density.
Population density was selected as an indicator for flood exposure because it is commonly used in FRAs to capture population exposure to floods (e.g., [
65,
66]). Population density is positively correlated with flood exposure in that an increase in population density results in a proportional increase in people directly exposed to flood exposure. This indicator was available at the Statistical Area 2 (SA2) level and was sourced from the Australian Bureau of Statistics (ABS) via a 2021 regional population estimate dataset. It was then rasterised using QGIS 3.24 software, being cut to boundaries applicable to that of the HNC. Importantly, density was calculated in particular as this standardises between differently sized SA2s.
Land use type is an important indicator for a Flood Exposure Index (FEI) because it ranks different land use types based on the extent to which they are flood-exposed, indicating which land use types are associated with the greatest cost (e.g., [
65,
67]). This indicator can be correlated to flood exposure by assigning values to land types (e.g., [
58]), where higher values indicate greater flood exposure. The land use type analysis for this study prioritised the built environment, such as infrastructure and roads (note that it remains different to the critical infrastructure density indicator described below). Capturing these elements is critical to a flood exposure analysis as inland flooding is, as mentioned, prevalent in urban environments, which may present poorer flood-conscious urban planning. For this study, this indicator was sourced as a 50 m vector from the NSW Government’s 2017 Land use v1.2 dataset. It was then rasterised and cut to HNC boundaries. Importantly, this clipped dataset had thirty-one land use types, which was too complex for this study’s scope. Therefore, the dataset was reclassified into a subset of eight: water bodies, nature conservation, forestry, cropping, grazing, horticulture, infrastructure, and other. Such reclassification is common in the literature (e.g., [
68,
69]). Having done this, a flood exposure value with a respective rating was then assigned to each reclassified land use type, as described in
Table A1,
Appendix A.
Critical infrastructure (CI) density, which describes the amount of CI over a given land area, is critical to an FRA because it again captures the flood-exposed built environment (e.g., [
70,
71]). This study assumes eight CIs: roads (including State Emergency Service (SES) evacuation routes), power stations, power substations, electricity transmission lines, hospitals, police stations, SES offices, and broadcast transmission towers. Importantly, these CIs capture important services and utilities such as transport, emergency services, and communication. CI density is positively correlated with flood exposure in that an increased CI density results in greater flood exposure. All CI datasets were downloaded as vectors, rasterised, and cut to HNC boundaries. It should be noted that roads were defined as dual carriageways, principal roads, secondary roads, and SES-recommended evacuation routes. CI density was calculated in particular because SA2s in the HNC vary significantly in size and CI density is a way of standardising between such SA2s.
Flood Vulnerability Indicators and Data Collection
In accordance with the IPCC definition, flood vulnerability was defined as “the propensity or predisposition to be adversely affected. It encompasses a variety of concepts and elements, including sensitivity or susceptibility to harm and lack of capacity to cope and adapt” [
22]. In this study, three categories of flood vulnerability were addressed: environmental, social, and economic. The indicators that were chosen to measure these categories were elevation, degree of slope, Index of Relative Socio-economic Disadvantage (IRSD), and Hydrologic Soil Groups (HSG).
The elevation of the area was investigated as a component of the topography of the HNC. The height at which land lies impacts the movement and drainage of water [
72,
73]. As a physical indicator, elevation shows the highest vulnerability where the lowest-lying areas occur, as water flows from higher- to lower-elevated areas. It was important to investigate this indicator in order to understand the influence of topography on the vulnerability of communities. The elevation at which communities place their residencies and infrastructure could contribute to the overall risk of the area.
Often found coupled with elevation in many datasets, the degree of slope is the percentage of change in elevation over a certain distance [
74]. It describes the shape and relief of the land as opposed to the height of the land. Slope is calculated using Equation (2):
This indicator was chosen as it complements elevation data and has been frequently used in past international flood vulnerability assessments [
75,
76,
77]. Slope influences the speed at which water travels, meaning that in areas with a higher degree of slope, water will run off more readily. The result can be highly sloping areas receiving less pooling of rainwater and thus a lower flood risk. The runoff from sloping areas can also result in a high vulnerability region if directed into a low-lying, flat area [
76,
78]. In this scenario, low slope is associated with higher vulnerability.
The Index of Relative Socio-economic Disadvantage (IRSD) is an index designed by the ABS as a component of the Socio-Economic Indexes for Areas (SEIFA) product. It comprises sixteen indicators that describe the relative disadvantage of areas. The IRSD is beneficial as a chosen indicator as it encompasses both social and economic aspects of the HNC. IRSD scores are mapped using quintiles. The lowest scoring 20% of areas are given a quintile number of 1, the second-lowest 20% of areas are given a quintile number of 2 and so on, up to the highest 20% of areas which are given a quintile number of 5. Low IRSD scores indicate that an area has a relatively greater disadvantage in general. For example, areas that have low IRSD scores may have (i) many households with low incomes, and/or (ii) many people with long-term health conditions or disabilities, and/or (iii) many dwellings that are overcrowded. Conversely, areas with a high IRSD score would present a lack of disadvantage, meaning, for example, (i) few households with low incomes, and/or (ii) few people with long-term health conditions or disabilities, and/or (iii) few overcrowded dwellings. Low IRSD scores indicate higher vulnerability to loss during a flood event.
Hydrologic soil groups (HSG) were used as an additional environmental indicator. HSG measures the ability of different soil types to absorb water. Soil properties play an important role in flood water and runoff behaviour [
79,
80]. HSG infiltration behaviour ranges from group D (very low) to group A (high). Lower infiltration rates mean that water is not readily absorbed into the soil. High infiltration means that water is very readily absorbed into the soil. The variation of infiltration behaviour is dependent on soil type, texture, grain size, and aggregate size. Low infiltration rates and their subsequent high runoff potential result in greater vulnerability.
Table 1 contains a description of each HSG and its relationship with runoff.
Table 2 outlines the Data Collection process for each of the flood risk inputs across the three risk elements.