Flood damage costs under the sea level rise with warming of 1.5 °C and 2 °C

The RCP scenarios were not designed to keep temperature below 1.5 °C, 2 °C (SED 2015) or other prescribed thresholds. However, of the models available from the Coupled Model Intercomparison Project Phase 5 (CMIP5, Taylor et al 2012) there are temperature pathways close to 1.5 °C and 2 °C. The 5%–95% range of CMIP5 models for RCP2.6 and RCP4.5 are 0.91 to 2.31 °C and 1.62 to 3.21 °C relative to pre-industrial levels respectively (IPCC 2013). We follow the approach of Jackson et al (2018) and identify models from both RCP2.6 and RCP4.5 whose temperature pathway lies within ±0.21 and ±0.28 °C of 1.5 °C and 2 °C respectively over the period 2080−2100 (figure 1 and table S1 available at stacks.iop.org/ERL/13/074014/mmedia).

Zoom In Zoom Out Reset image size

Using these 1.5 °C and 2 °C subsets of the CMIP5 models we calculate the contribution to global sea level for the following sea level components (table S1, figure 1): ocean steric expansion (T), and melting of glaciers (GIC). We then calculate an ensemble mean and uncertainties for each of these components for 1.5 °C and 2 °C temperature scenarios (Jackson and Jevrejeva 2016). In contrast with the approach of Jackson et al (2018) we utilized RCP2.6 projections for total contributions from Antarctica (AIS), Greenland (GrIS) (these are in line with recent ice sheet simulations for this RCP, DeConto and Pollard 2016) and scenario independent projections of land-water storage (LW) due to the combined effect of man-made reservoirs and ground water extraction. This allows us to apply a probabilistic approach (Jackson and Jevrejeva 2016) to the sum of projected sea-level components,

$$\begin{equation} {\rm GSL} = {\rm T} + {\rm GIC} + {\rm GrIS} + {\rm AIS} + {\rm LW}, \end{equation} \tag{ 1 }$$

to give global sea level (GSL) projections (figure 1).

To further explore the GSL response to temperature change, we create a set of idealised warming scenarios for 1.5 °C and 2 °C, where we prescribe the trajectory of the temperature to reach and then stabilize at the two targets in 2030, 2050, 2070 and 2090 (figure S1, table 1). We investigate the global sea level response for these idealised scenarios by 2100 and their associated uncertainties using the semi-empirical model by Grinsted et al (2010). We use our idealised temperature scenarios to understand the sea level response to temperature changes and do not discuss through which intervention (e.g. the reduction of emissions) our scenarios could be achieved.

To assess the impact of sea level change under these limited warming targets by 2100, we compare our projections to the RCP8.5_J14 sea level rise projections (figure 1), where the RCP8.5 scenario is supplemented by Greenland and Antarctic contributions elicited by Bamber and Aspinall (2013) (Jevrejeva et al 2014, henceforth RCP8.5_J14). In our study we refer to the RCP8.5 scenario for temperature projections and RCP8.5_J14 for sea level projections associated with warming under RCP8.5. Differences between 1.5 °C, 2 °C and RCP8.5_J14 are shown in figure 2 as probability density functions (PDFs) of global sea level and its components in 2100.

Zoom In Zoom Out Reset image size

Regional sea level rise displays complex spatial patterns due to the dynamic redistribution of ocean mass and the gravitational spatial patterns (so-called 'fingerprints') associated with specific geographical distributions of ice loss from mountain glaciers, Greenland and Antarctica ice sheets and changes to land-water storage by man-made reservoirs and ground water extraction. For regional sea level projections we follow the approach by Jackson and Jevrejeva (2016) and combine global projections of each sea level component with their associated normalised fingerprint using a probabilistic method:

$$\begin{equation} \begin{array}{l@{}l@{}l@{}} {\rm RSL} = F({\rm T}) + F({\rm GIC}) + F({\rm GrIS})\\ + F({\rm AIS}) + F({\rm LW}) + F({\rm GIA}) \end{array} \end{equation} \tag{ 2 }$$

where RSL is regional sea level, F(T), F(GIC), F(GrIS),F(AIS), F(LW) are the normalised fingerprints scaled by the global average projected sea level (F()) respectively of: T, ocean thermal expansion plus dynamical changes in sea surface height; GIC, ice loss from glaciers (surface mass balance); GrIS, ice loss from Greenland (surface mass balance and ice dynamics); AIS, ice loss from Antarctica ice sheet (surface mass balance and ice dynamics); LW, land water storage. GIA refers to glacial-isostatic adjustment, which is a non-climate related component. To generate regional sea level projections we randomly sample the PDF of each sea level component (similar to figure 2). We produce sea level projections for each component (excluding GIA) by scaling the normalised fingerprint of each component by their associated random samples. We then sum the fingerprints of the sea level components making, in total, 5000 realisations of sea level. This allows us to create a probability density function for each grid point in our map of regional sea level projections. To account for GIA, we add the time-integrated spatial field of GIA induced sea level change from the ICE 6G model (Peltier et al 2015) to the sum of sea level components equation (2). We assume each of the sea level components is uncorrelated and that the spatial pattern of future land-based mass loss will be the same as at present (Jackson and Jevrejeva 2016).

The impacts of sea-level rise were computed using the DIVA modelling framework (version 2.1.0, database 32), an integrated bio-geophysical coastal systems model, which is driven by (climate change induced) sea level change and socioeconomic development (Vafeidis et al 2008, Hinkel 2005, Hinkel and Klein 2009, Hinkel et al 2014). Impacts are generated by dividing the world's coast into 12.148 linear segments (excluding Antarctica), each having similar bio-physical and socio-ecological characteristics. Global mean sea level rise is combined with estimates of vertical land movement due to GIA (Peltier 2004), plus subsidence or uplift in deltaic regions (39 locations with rates from Ericson et al (2006) and a further 78 where 2 mm yr⁻¹ of subsidence was assumed). This was used to determine changes of local extreme water levels for different return periods for each segment, based on a hydrodynamic modelling reanalysis of storm surges and extreme sea levels (GTSR dataset, Muis et al 2016). Extreme water level distributions are assumed to uniformly increase with regional mean sea-level rise, following 20th century observations (Menéndez and Woodworth 2010).

Land elevations were derived from the shuttle radar topographic mission high resolution digital elevation model (Jarvis et al 2008) and the GTOPO30 dataset (USGS 2015) for land areas poleward of 60°N and 60°S. Linear interpolation was used between grid points to provided discrete elevations at the required resolution. To calculate population exposure to potential flood events, the global rural urban mapping project (GRUMPv1) with a spatial resolution of 30 arc seconds was used (CIESIN 2011, Balk et al 2006). Shared Socioeconomic Pathway 2 (SSP2) was used to determine socio-economic development, in particular to project future coastal population and national gross domestic product (GDP) (Moss et al 2010, O'Neill et al 2014). SSP2 represents a future with a mix of adaptation and mitigation challenges. Global population increases until mid-century, reaching approximately 9 billion people globally before slightly declining. Global GDP increases throughout the century.

Existing protection by dikes was modelled through a generic rule based on income group and population density, as defined by Sadoff et al (2015) and complemented with protection standards for the 136 biggest coastal cities as defined by Hallegatte et al (2013). Following current and future guidelines in Kind (2014) and Deltacommissie (2008) the Netherlands was treated as a special case, with implemented protection equal to the 1-in-10 000 year water level for the whole country. Further details are available in table S2. We consider two adaptation option to explore future coastal damages. First, a scenario with no additional adaptation that keeps dike heights constant at the baseline level. Second, a scenario with business-as-usual adaptation where the standard of protection is updated every five years according to a generic rule used in the initialization phase. Thus, dike heights are raised to cope with rising sea levels and changes in population density.

Two kinds of costs were considered: annual sea flood cost and annual adaptation cost consisting of construction cost for raising existing dikes and the cost of maintenance. Reported total annual costs are the sum of annual sea flood cost and annual adaptation cost. Results are presented at the level of World Bank income groups (high, upper middle, lower middle and low income countries) (World Bank 2013, table S3).

The range of global mean sea level projections with warming of 1.5 °C and 2 °C is dependent on temperature trajectories (table 1, figure S1). Using idealised temperature pathways, the median sea level rise for 1.5 °C warming trajectories by 2100 is up to 52 cm (25−87 cm, 5th−95th percentile). The difference in projected global sea level rise by 2100 between warming of 1.5 °C and 2 °C is up to 11 cm at the median and up to 25 cm at the 95th percentile (table 1, figure S1). The rate of global sea level rise of 7.2 mm yr⁻¹ (median) and up to 12.7 mm yr⁻¹ at the 95th percentile by 2100 is projected for the trajectory of temperature reaching 2 °C in 2040 and kept below the 2 °C target after that. Over the 21st century, sea level rise rates are projected to exceed the highest rate of rise of the 20th century even if emissions are limited sufficiently to reach the 1.5 °C target.

Examining the results of our process-based approach, coastal sea level rise generally exceeds the global average (figure 3), with exceptions of coastline in the areas close to Greenland and Antarctic ice sheets. The largest differences between 1.5 °C and 2 °C scenarios in 2100 along coastlines are ~15 cm (median) and up to 20 cm (95th percentile) (differences in global averages of 6 cm (median) and 7 cm (95th percentile), Jackson et al 2018 and Goodwin et al 2018) and occur for the US east coast and small-island nations in the Pacific and Indian oceans. These low-lying island nations in the Tropics are particularly vulnerable to flooding from storms today. Potential changes in flooding frequency due in part to sea level rise will further challenge the sustainability of these coastal communities (Vitousek et al 2017, Woodruff et al 2013).

Zoom In Zoom Out Reset image size

In 2040, there is very little difference between RCP8.5_J14 and 1.5 °C scenarios for both global and coastal sea level projections at median or 95th percentiles (figure 1(b), figures S2(a), (c)). However, by 2100 (figure S2(b)) the difference between RCP8.5_J14 and 1.5 °C scenario for global sea level is around 39 cm (median), with large areas along the coastline of South and South East Asia, US east coast, Africa and Australia reaching differences up to 50 cm. For the small island states in the Pacific and Indian oceans the difference in median sea level projections with RCP8.5_J14 scenario and 1.5 °C temperature scenario would be more than double the total sea level rise occurring in the 20th century. The difference between these two scenarios for projected sea level rise in 2100 at the 95th percentile (figure S2(d)) is significantly higher: around 117 cm globally and up to 155 cm for small islands in the Western Pacific and 147 cm in the Indian Ocean.

Annual sea flood costs and total annual coasts are projected under global sea level of 0.52 m with warming of 1.5 °C, 0.63 m with warming of 2 °C (table 1), 0.86 m for RCP8.5_J14 (median) and 1.8 m RCP8.5_J14 (95th percentile) scenarios using the DIVA modelling framework. It is important to note that these are annual costs, which are dependent on adaptation assumptions in previous time steps.

The difference in 2040 between flood costs associated with 1.5 °C and 2 °C sea level rise (with a 2.8 cm difference in sea level rise in 2040) is projected to be US$ 0.3 trillion per year (0.1% of global GDP). By the end of the 21st century global annual flood costs are projected to be US$ 10.2 trillion per year (1.8% of GDP) under 1.5 °C and US$ 11.7 trillion per year (2.0% GDP) under 2 °C scenario, if no further adaptation is undertaken.

By 2100 the difference of 11 cm between global sea level rise with warming of 1.5 °C and 2 °C will result in additional costs of US$ 1.5 trillion per year (0.25% of global coastal GDP), assuming that there has been no additional adaptation (figure 4). If the 2 °C target is missed, and we follow the RCP8.5_J14 scenario (median sea level rise of 0.86 m and 95th percentile of 1.8 m in 2100), global annual flood costs without additional adaptation are projected to be US$ 14.3 trillion per year (2.5% of GDP) for the median scenario and up to US$ 27.0 trillion per year for the 95th percentile (figure 4(a)), accounting for 4.7% of global GDP (table S4).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Under the assumption of no additional adaptation the greatest annual flood costs by 2100, as a proportion of GDP for all scenarios, are projected for the upper middle income countries (table S5) ranging from 2.8% with warming of 1.5 °C to 7.3% with the RCP8.5_J14 scenario (table S5). A large proportion of this cost is attributed to China (table S6) as it has a long coastline, a large coastal population and a rapidly growing GDP. For the high income group, future annual flood costs tend to be lower as their population falls significantly under SSP2. Large cities in all income groups tend to be well protected because significant levels of infrastructure and assets tend to be (though are not exclusively) located there (Hallegatte et al 2013, IPCC 2014). Global sea level rise under the 2 °C warming scenario shows that all income groups are projected to experience increased annual flood costs compared to the sea level rise with 1.5 °C warming, up to 0.4% of GDP in 2100 (table S5).

Table S6 shows the top 10 countries with the largest values of annual sea flood costs (US$ per year) in 2100 with sea level rise associated with warming of 1.5 °C (52 cm) assuming no additional adaptation. The largest flood cost in 2100 is projected for China, which is an order of magnitude large than the USA and Japan. However, the countries affected most strongly in terms of percentage of country-level GDP (figure 5; table S7) are Kuwait (24%), Bahrain (11%), United Arab Emirates (9%) and Vietnam (7%). Changes in sea level rise by 2100 with warming from 1.5 °C–2 °C could result in an increasing annual flood cost for China of US$ 0.4 trillion per year and for Vietnam up to US$ 0.07 trillion per year (S6−S9).

While the annual sea flood cost projections without additional adaptation are mainly used for analytical purposes to explore risk and enable long term decision making, such losses are unlikely to be tolerated by society and adaptation is expected to be widespread (e.g. Hinkel et al 2014, IPCC 2014, Wong et al 2014, Diaz 2016). By 2100 global flood cost with additional adaptation is estimated to be 0.2% GDP for both 1.5 °C and 2 °C sea level projections, lower than 1.8% GDP (1.5 °C) and 2% GDP (2 °C) projected without additional adaptation. The difference in annual flood costs for each income group is illustrated in figure 6 and indicates that low income countries may experience greater flood cost as percentage of GDP compared to higher income groups because of their limited means to implement adaptation measures. Despite this, there is a large potential for coastal adaptation across all income groups.

Zoom In Zoom Out Reset image size