**1. Introduction**

Anthropogenic activities have changed the Earth's temperature by approximately 1 ◦C between the period of 1850–1900 and the year of 2017 [1]. This has led to an average global sea level rise (SLR) of 3.20 mm/year. It is expected with 67% confidence that by 2100, the global average sea level will rise by 0.28 m to 0.98 m relative to the mean sea level of the years 1986–2005 [2].

The effects of SLR include the salinization of coastal agricultural areas and water storages, the destruction of coastal eco-systems, the erosion of shorelines, and the destruction of buildings and infrastructure [1]. SLR affects coastal countries all over the world, such as China, the Netherlands, Nigeria and the United Kingdom [3]. Some of the most impacted regions are the islands located in the Caribbean Sea and in the South Pacific, the so-called Small Island Developing States, as well as southern and eastern parts of Asia [4]. The Intergovernmental Panel on Climate Change (IPCC) developed four Representative Concentration Pathways (RCPs). They refer to four different pathways

of Greenhouse Gas (GHG) emissions and atmospheric concentrations, air pollutant emissions and land use. While RCP 4.5 and RCP 6.0 refer to intermediate scenarios, RCP 2.6 refers to very low GHG emissions keeping global warming likely below 2 ◦C. RCP 8.5 refers to a very high GHG emissions pathway [5]. Kulp and Strauss [6] estimated that a global warming of 2 ◦C in relation to RCP 4.5 results in an increase of 40 million additional people to live permanently below the high tide line until 2050 and an increase of 90 million people until 2100.

In particular, Small Island Developing States are a ffected by this since their geomorphology is often characterized by low-elevation islands with population concentrated along their coasts [7]. In addition, Small Island Developing States heavily depend on the functioning of coastal ecosystems, and their economies are highly sensitive to slight changes [7,8]. Furthermore, they are more vulnerable to the e ffects of SLR because most of them lack institutional, financial and technical structures to adapt to it [7].

Vulnerability assessment studies for Small Island Developing States have been carried out within various states, such as islands in the Caribbean Sea [9,10], in South-East Asia [11,12] and in the Southern Pacific [13,14]. In 2008, Gravelle and Mimura [13] conducted a vulnerability assessment for the Republic of Fiji aiming to point out areas that are threatened by the SLR for di fferent SLR scenarios. This study illustrated that most urban centers will face partial inundation within this century and that a proportional increase between the SLR and the total inundated area can be observed. Other vulnerability assessments estimated the total flooded area of specific regions [15], the number of inundated households and buildings being inhabitable within a certain area [16], the length of a ffected roads by the SLR [17] as well as the financial value of lost built structure [18].

Based on such vulnerability assessments, adaptation measurements incorporating SLR are being developed on the city, region or country level [7,19–21]. These include direct protection actions, such as the construction of barriers in the sea, and preventive actions, such as the relocation of houses or entire villages [21]. Relocations are fairly drastic solutions and generally require large economic and human resources. Moreover, they tend to destroy social structures, cultural traditions, as well as causing emotional stress [22]. Forced demolitions of coastal buildings and infrastructure result in large amounts of construction and demolition waste, which cause environmental stress. On top of this environmental pressure, the new facilities being built to replace the demolished ones will require new materials for the reconstruction. Nevertheless, information on the demolition waste streams from inundated coastal buildings is crucial for Small Island Developing States in two aspects: first, waste managemen<sup>t</sup> is demanding due to limited land availability, remoteness, and high costs [23,24]; second, possibility of reuse of materials enables waste mitigation and contributes to the overcoming of resource shortages on Small Island Developing States [25,26].

However, while the e ffects of the SLR on the number of buildings inundated have already been studied by vulnerability assessment studies [16], these studies have not estimated demolition waste streams and materials required for reconstructions caused by SLR. The information on demolition waste streams and materials required for reconstructions can be extracted by the concept of material stock analysis (MSA). MSA is a method developed in the field of Industrial Ecology that allows the estimation of the amount of materials in use in the socio-economic sphere of our societies [27,28]. This tool has been used for estimating the mass of materials lost during the Great East Japan Earthquake [28]; waste flows coming from demolitions [29,30]; or the potential for urban mining [31]. To date, the concept of MSA has ye<sup>t</sup> to be combined with vulnerability assessments that focus on SLR.

This study develops a novel methodology for an estimation of construction material amounts through a hybrid combination of geospatial analysis and material stock analysis, applying and evaluating it to the case study of the Republic of Fiji. This novel methodology can be applied to small islands/coastal regions for the estimation of lost material stock (MS) caused by the SLR. The Republic of Fiji was chosen as a case study area because it represents a typical Small Island Developing State due to a relatively high population density living along the coast, as well as the country's limited climate adaptation capacity. In the following section (Materials and Methods), data sources and an introduction to the case study area, as well as a detailed explanation of the methodology are provided. In Section 3 (Results), the results of the spatial analysis are provided, along with a quantification of the construction materials (number of buildings and total mass) that will be permanently inundated. The results are separated by province in rural and urban dwellings. Additionally, high risk areas were identified. Section 4 (Discussion) provides insights and considerations on the findings. Section 5 (Conclusions) draws conclusions, discusses the limits of this methodology and points out future research steps.

#### **2. Materials and Methods**

This study's research approach consists of two parts: firstly, a Geographic Information System-Digital Inundation Analysis Model (GIS-DIAMs) to calculate the number of flooded buildings and secondly an MSA to estimate the construction material stocked in those. Figure 1 represents a conceptualization of how the models were combined, while data sources are listed in Table 1.

**Figure 1.** Schematic representation of the methodological approach of this study.

**Table 1.** Data sources used for the Geographic Information System-Digital Inundation Analysis Model (GIS-DIAM) and the material stock analysis (MSA).


The GIS-DIAM analysis enabled the identification of the number and location of buildings, subdivided per province and into rural and urban areas, subjected to inundation due to SLR. Using this result, the percentage of buildings inundated was estimated in comparison to the total number of buildings per province, also subdivided into rural and urban areas. The MSA was then conducted by assigning for each inundated building a construction typology, which carries information on the typical material intensities per m2. This information was then crossed with data on building size to estimate the amount of the materials concrete, steel and timber potentially lost due to SLR.

#### *2.1. Case Study Area*

The Republic of Fiji is a group of islands located in the South Pacific. Its population accounts for 881,000 people living on 18,000 km2, subdivided on 300 islands, of which 100 are inhabited [45]. Fiji's coastline measures around 1130 km [45] where 90% of the population lives and where the biggest urban regions, Lautoka, Nadi, Labasa and the capital Suva, are located [46]. This study focuses on the two main islands of Fiji, Viti Levu and Vanua Levu, where 96% of the population lives [33].

In 2017 and 2018, on average, 626 buildings valued US\$73.63 million were constructed each year [36,37]. In 2014, the village of Vunidogoloa, located in the Republic of Fiji, had to be relocated due to coastal erosion and storm surges caused by climate change induced SLR [22,47]. It cost approximately US\$500,000 to relocate the residents 2 km inwards from the coast [22]. It is predicted that until 2050, with reference of a SLR of 0.26 m relative to the years 1986–2005, 30,000 Fijians occupy land vulnerable to the SLR. Until 2100, with a predicted SLR of 0.59 m relative to the years 1986–2005, 80,000 Fijians occupy land vulnerable to SLR [6].

In 2017, the governmen<sup>t</sup> of Fiji began planning adaptation actions to counter the e ffects of SLR. Highly a ffected areas were identified and the relocation of several settlements was forecast. The governmen<sup>t</sup> additionally plans to secure funds, to focus on better managemen<sup>t</sup> of natural resources and to increase their human capital by investing in the education of engineers and by training existing technical stu ff [21]. Moreover, it plans to increase resilience in communities by identifying the most vulnerable villages [14] and by protecting urban coastlines from the e ffects of SLR [19]. According to The World Bank [14], the Fijian governmen<sup>t</sup> foresees the relocation of settlements where storms are happening in a frequency that makes the settlements unable to live in on a long-term view. The governmen<sup>t</sup> further plans on improving the current Digital Elevation Model of Fiji using LiDAR (Light Detection And Ranging) data, which will enable more detailed and accurate inundation analyses [14].

#### *2.2. GIS-Based Digital Inundation Analysis Model (GIS-DIAM)*

A digital elevation model of the two main islands of Fiji was taken as primary data. This model carries orographic information of the islands. Further data include a detailed 2D representation of the existing buildings. Moreover, metadata on the administrative boundaries of the region was used. Using ArcGIS, a popular software for GIS analyses, a simulation of the inundation areas due to SLR was conducted. Predictions on SLR are based on predictions by the IPCC [2]. Results are calculated for the years 2050 (scenario 1) and 2100 (scenario 2). This simulation generated two GIS-DIAMs which highlighted the number of inundated buildings per province, further discerned in rural and urban areas.

## 2.2.1. Elevation Data

Height information to was directly provided by the Geospatial Division of the Ministry of Lands & Mineral Resources Fiji [38]. The data's spatial resolution is not reported in the document, and, when inquired, the Ministry did not provide an answer. To overcome this limitation, we assumed that this data is generally based on satellite data displaying surface elevation rather than terrain height. This is due to a constantly displayed height di fference when comparing densely forested areas with open spaces located in close proximity. Additionally, the Fijian DEM was compared regarding its

accuracy to elevation data by the United States Geological Survey, Shuttle Radar Topography Mission (SRTM) data [48] and to data by the Japan Aerospace Exploration Agency, Advanced Land Observing Satellite (ALOS) data [49].

## 2.2.2. Inundation Data

Inundation is predicted for the years 2050 (scenario 1) and 2100 (scenario 2). The dates were chosen to provide an overview on two di fferent epochs in future time, one happening relatively soon and one relevant for long-term planning. Maximum tide inundation, including storm surges, were not incorporated in this study, because it is unable to accurately predict for which inundation interval a building is unusable. For the inundation data, only the permanent SLR is incorporated into the GIS-DIAM, which means that it does not take into consideration tidal e ffects.

This permanent SLR is based on the global SLR predictions by the IPCC's Fifth Assessment Report. SLR was chosen on a global scale, relative to the period 1986–2005. It was chosen in reference to RCP 2.6 and RCP 8.5. The likelihood of the SLR refers to a 'likely range' as referred to by the IPCC, meaning a probability of 66%–100% [2].

The IPCC [2] predicts, for the period 2046–2065, a minimum SLR of 0.17 m (RCP 2.6) and a maximum SLR of 0.38 m (RCP 8.5). As the report does not provide tabular data for the year 2050, the authors assumed an SLR of 0.22 m for the year 2050 (scenario 1). The value for 2100 was determined as the average of the lowest value predicted for RCP 2.6 (SLR of 0.28 m) and the highest value predicted for RCP 8.5 (SLR of 0.98 m). Thus, an SLR of 0.63 m is expected on average [2].

#### 2.2.3. Administrative Boundaries and the Spatial Localization of Buildings

The Fiji are divided into 15 provinces and 1602 enumeration areas. Data on administrative boundaries were downloaded from the PopGIS 2.0 platform which is managed by the Fiji Bureau of Statistics, based on the 2007 Census [32]. GIS data locating 89,628 Fijian buildings were taken from the Geofabrik platform, which retrieves data from Open Street Maps [34]. The data were then tested for their accuracy using satellite imagery by Esri et al. [35]. It was evident that Open Street Maps data do not cover all the buildings on Fiji's coastline. Therefore, an additional 6979 buildings were manually drawn as points in ArcGIS based on the satellite imagery by Esri et al. [35].

#### 2.2.4. Separation between Urban and Rural Areas

The number of inundated buildings was calculated according to the province a building is constructed, subdivided in urban and rural areas. In this study, an area was classified as urban when it is listed as 1st category urban area in the 2007 Census of Population and Housing [33]. The governmen<sup>t</sup> defines cities and towns by their urban attributes, their economic activity and their population size [50]. In the 2007 census, twelve areas are listed as 1st category urban area. The enumeration areas which are located within urban zones were subsequently manually assigned using the satellite imagery by Esri et al. [35].

#### *2.3. MSA-Based Construction Material Stocked Model*

To calculate the materials stocked in buildings subject to inundation, an MSA was conducted for characterizing the structural materials typically used for buildings on Fiji: concrete, steel, and timber. Using the 'Select By Location' tool by ArcGIS, a building was referred to as inundated if its polygon or point was within the features of the inundation layer. To proceed, an equation first described by Tanikawa et al. [28] was modified as in the following Equation (1):

$$MS\_K = \sum\_{j=1}^{4} \left( MI\_{K,J} \cdot \sum\_{i=1}^{n} \left( GFA\_{i,j} \right) \right) \tag{1}$$

where *MSk* is the stocked amount of a specific construction material *k*, *MIk,j* is the material intensity of the construction type *j* and material *k*, and *GFAj*,*<sup>n</sup>* is the gross floor area of the *i*-th building per construction type *j*. Note that the index *j* goes from 1 to 4 as there are 4 building typologies, while *i* goes up to *n* as there is a variable number of buildings for each typology.

## 2.3.1. Construction Types

Buildings in Fiji can be classified into 4 building typologies, depending on the material used for their walls: cement block masonry, timber frame cladded by timber panels, timber frame cladded by steel panels [39], and reinforced concrete [44]. Traditionally, Fijian buildings have one story, concrete foundations and steel based sheets as roof [39]. While cement block masonry, timber cladded and iron cladded buildings appear in both rural and urban areas, buildings based on reinforced concrete are only constructed in cities where houses typically have more than one story. This includes the city of Suva (three floors) as well as the cities Nadi, Labasa, Ba and Lautoka (all with two floors). See Supplementary Materials §1 for a list of the data sources used to determine the number of floors per city.

The Fiji Bureau of Statistics provides the average distribution of construction typologies used in each one of the 1602 enumeration areas [32]. This served as a basis for allocating to each province the share of each construction type, separated in rural and urban areas. As it is impossible to know exactly the actual construction type for a specific building from an aerial photo, we assumed the typologies of inundated buildings as proportional to the share of typologies in a certain enumerated area. Please see Supplementary Materials §2 for more information on the distribution of construction typologies per province.

## 2.3.2. Material Intensities

Table 2 shows the material intensities used for the MSA. To date, no typical material intensities for the construction types defined have been published, for Fiji nor for any other region in the world. Thus, material intensities were calculated by the authors manually. The material intensities of buildings with walls based on cement block masonry, timber sheets and steel based corrugated iron sheets were based on a housing construction manual provided by the Habitat for Humanity [40] in combination with baseline data on local building structure and materials published by Caimi et al. [39]. Details on calculations can be retrieved in Supplementary Materials §3.


**Table 2.** Material intensity per construction type (kg/m<sup>2</sup> of gross floor area) [28,31,39–44].

Information that helps with calculating the material intensity of multi-storied reinforced concrete buildings is not reported in Habitat for Humanity [40], nor in Caimi et al. [39]. It was assumed that multi-storied reinforced concrete buildings on Fiji are built in a similar way as they are in other regions

of the world. Therefore, data was retrieved from research previously conducted by Cheng et al. [31] for buildings in Taipei, Taiwan. Here, Cheng and colleagues cite research by Chang [41]. Subsequently, all material intensities were compared to the values published by Tanikawa et al. [28], which lists material intensities by Nagaoka et al. [42] and Tanikawa and Hashimoto [43].

#### 2.3.3. Gross Floor Area (GFA)

The total gross floor area of inundated buildings was not directly retrievable from the GIS data, as we identified over 7000 buildings that were not mapped. For this reason, a probabilistic approach had to be implemented.

The estimation of the total footprint of inundated buildings is reported in Equations (2) and (3):

$$FP\_{\text{total,urban},p} = FP\_{\text{avg,urban},p} \cdot n\_{\text{urban},p} \tag{2}$$

$$FP\_{total,rural,p} = FP\_{avg,rural,p} \cdot n\_{rural,p} \tag{3}$$

where *FPtotal*, *urban*, *p* indicates the total building footprint (expressed as m2) of inundated buildings located in urban areas of the province *p*; *FPavg*,*urban*,*<sup>p</sup>* reports the average footprint (expressed in m2) of an inundated building in urban areas of the province *p*; and *nurban*,*<sup>p</sup>* is the number of buildings that are inundated in urban areas of the province *p*. Equation (3) is identical to Equation (2), the only difference being that it refers to rural areas rather than urban ones.

The average area of a building was estimated based on the average area of those buildings that were withdrawn as polygons from Geofabrik GmbH and OpenStreetMap Contributors [34]. Please see Supplementary Materials §2 for a list of the average area of a building per province.

The calculation of the total gross floor area of inundated buildings in a certain province is shown in Equation (4):

$$GFA\_{total,p} = FP\_{total,urban,p} \cdot Flow\_{urban} + FP\_{total,rural,p} \tag{4}$$

where *GFAtotal*,*<sup>p</sup>* is the total floor area which is going to be inundated in the province *p*; and *Floorsurban* indicates the average number of floors present in buildings in the urban area of the province *p*. Note that there is not a term floors for the rural area, as buildings in rural areas are always limited to a single story.

The total gross area of a province is then discerned into typologies as per Equation (5):

$$GFA\_{j,p} = GFA\_{\text{total},p} \cdot \upsilon\_{j,p} \tag{5}$$

where *GFAj*,*<sup>p</sup>* is the gross floor area of the construction type *j* in province *p*; and <sup>ν</sup>*j*,*<sup>p</sup>* is the ratio of building having type *j* in province *p*, as calculated in Equation (6).

$$w\_{j,p} = \frac{n\_{j,p}}{n\_{\text{total},p}} \tag{6}$$

where <sup>ν</sup>*j*,*<sup>p</sup>* is calculated as the fraction between the number of buildings having type *j* in the province *p* (*nj*,*<sup>p</sup>*) over the total number of buildings in the province *p* (*ntotal*,*<sup>p</sup>*).

The overall gross floor area is the calculated as in Equation (7):

$$GFA\_{\bar{\jmath}} = \sum\_{p=1}^{15} GFA\_{\bar{\jmath},p} \tag{7}$$

where *GFAj* is the overall gross floor area for the construction type *j*. Note that *p* goes from 1 to 15, as there are 15 provinces in this case study.
