Reconciling biodiversity conservation and flood risk reduction: The new strategy for freshwater protected areas

Natural disaster risk reduction (DRR) is becoming a more important function of protected area (PAs) for current and future global warming. However, biodiversity conservation and DRR have been handled separately and their interrelationship has not been explicitly addressed. This is mainly because, due of prevailing strategies and criteria for PA placement, a large proportion of PAs are currently located far from human‐occupied areas, and habitats in human‐occupied areas have been largely ignored as potential sites for conservation despite their high biodiversity. If intensely developed lowland areas with high flooding risk overlap with important sites for biodiversity conservation, it would be reasonable to try to harmonize biodiversity conservation and human development in human‐inhabited lowland areas. Here, we examined whether extant PAs can conserve macroinvertebrate and freshwater fish biodiversity and whether human‐inhabited lowland flood risk management sites might be suitable to designate as freshwater protected areas (FPAs).


| INTRODUC TI ON
Societal expectations for protected areas (PAs) have changed intermittently over the past 150 years (Watson et al., 2014). proposed PAs (Murti & Buyck, 2014). Traditionally, the strategy for establishing stringent PAs was to separate a portion of the natural environment from human activities. Therefore, PAs were preferentially established in areas unsuitable for commercial activity, rather than in those with high biodiversity conservation value (Margules & Pressey, 2000). Thus, most extant PAs were established at higher elevation or in isolated locations -these are often described as "rock and ice" habitats (Joppa & Pfaff, 2009). Protected areas, however, may play a role in ecosystem-based DRR (eco-DRR: the sustainable management, conservation and restoration of ecosystems to reduce disaster risk) and can be established in human-dominated areas, such as urban and agricultural landscapes. Under the encouragement of the Aichi Biodiversity Targets, both the number and area of PAs are rapidly increasing worldwide (Watson et al., 2014); with the PAs estate increasing by 2.3% on land (i.e. covering 15% of land) and 5.4% in the oceans (i.e. 7% of the oceans) from 2010 to 2020 (IUCN, 2020). Given global efforts to achieve the Post-Aichi Biodiversity Targets, a new strategy for identifying candidate sites for PAs is required to enhance the effectiveness of PAs in meeting societal needs (Convention on Biological Diversity, 2020).
Freshwater ecosystems are degrading at a faster rate than any other biome in the world (Dudgeon et al., 2006;Reid et al., 2019).
However, extant PAs do not sufficiently cover the distribution of endangered freshwater species, because freshwater ecosystem conservation was largely not a priority when they were established (Nel et al., 2007). In Japan, the main PA system is composed of national parks. The principle purpose of this national park system is to conserve biodiversity and maintain the scenic and recreational value of the designated areas. As is the case internationally, the existing national parks (34 in total, with an average area ±SD of 632.8 km 2 ± 669) tend to be located mainly in mountainous areas.
Although all national parks in Japan enclose some freshwater habitats, none were established purely to conserve freshwater ecosystems. Therefore, there is an urgent need to establish additional freshwater protected areas (FPAs). In order to conserve as many species as possible, new FPAs should be placed in lowland areas where the majority of riverine animals are distributed (Tockner & Stanford, 2002). In general, because lowland areas are highly productive but sparsely distributed in mountainous regions, they tend to be heavily populated. Studies have shown that areas of high productivity overlap with areas of high biodiversity in many regions of the world (Chase & Leibold, 2002).
Flood disasters are some of the worst natural disasters in terms of damage incurred and loss of life (Perry, ). In the period 2006-2015, floods accounted for more than 45% (1719 floods) of the total natural disasters around the world (Chau, 2017;Sanderson & Sharma, 2016). At the global scale, more than US$56 billion were lost (i.e. infrastructure damage and lost commercial activity) due to floods in 2016, representing the largest economic losses among all natural disasters. Most of the 10 largest floods in history, such as the flood disaster in Thailand in 2013, have occurred in Asian monsoon countries (Guha-Sapir et al., 2016). The magnitude of damage caused by hydrological disasters has increased in recent years in densely populated lowland floodplains (James & Cutter, 2008). Furthermore, many climate models predict an increase in heavy precipitation as temperatures rise, which may increase flooding risk (Kollat et al., 2012;Wuebbles et al., 2008). Thus, effective DRR strategies for floods are crucial for maintaining human well-being.
Lowland FPAs are increasingly important due to the need for both DRR and biodiversity conservation. If intensely developed lowland areas with high flooding risk overlap with important sites for biodiversity conservation, this would indicate an urgent need to start a discussion on how to achieve biodiversity conservation alongside human development in human-inhabited lowland areas. Doing so would potentially achieve a win-win solution by reducing flooding risk while simultaneously conserving biodiversity. However, the relationship between biodiversity and disaster risk has not yet been explicitly defined. Therefore, in this study, we (1) examined whether extant PAs, which serve as FPAs, can conserve macroinvertebrate and freshwater fish biodiversity, and (2) analysed the relationship between candidate sites for new FPAs and flood disaster risk and land use intensity at a national scale across Japan.

Editor: April Blakeslee
Main conclusions: Floodplain conservation provides suitable habitat for many freshwater organisms and helps control floodwaters, so establishing new FPAs in areas with high flood risk could be a win-win strategy for conserving freshwater biodiversity and enhancing ecosystem-based DRR (eco-DRR).

K E Y W O R D S
aquatic animals, climate change, complementarity, eco-DRR, river, win-win strategy 2 | ME THODS

| Distribution records of freshwater fishes and macroinvertebrates
We compiled national-scale distribution data for freshwater fishes and macroinvertebrates from several surveys contained in the Japanese government's River Environmental Database (Ministry of Land, Infrastructure, Transport and Tourism, http://www.nilim.go.jp/ lab/fbg/ksnka nkyo/index.html, accessed 12 October 2021), which includes data from 2000 to 2010, and the Natural Environmental Survey Database (Ministry of the Environment, http://www.biodic. go.jp/, accessed 12 October 2021), which includes data from the same period. We identified both intra-and international exotic species based on the expert opinions of several freshwater scientists in Japan and removed them from further analysis. Those records not identified to the species level were also removed. We aggregated the species distribution records into 1 km × 1 km grid cells, because point locations for several records were not available.

| Status of Japanese river and environmental data
The total length of all rivers in Japan is 350 158 km, and the national river density (total river length/total land surface area; 0.93 km/ km 2 ) is higher than for other countries such as the United States (0.13 km/km 2 ) and some European countries such as Spain (0.19 km/ km 2 ) and Germany (0.21 km/km 2 ) (Andreadis et al., 2013). The freshwater fauna of Japan has declined due to human activities such as river channelization, damming and land-use change (Nakamura et al., 2017). We focused on national parks as a focal PA system because they constitute the core of Japanese PA systems (The Nature Conservation Society Japan, 2013). The mean annual cost of flood damage in Japan was approximately 1.2 billion USD from 1990 to 2017, but has rapidly increased to 15.5 billion USD during the most recent 3-year period (2018-2020) (EM-DAT, 2021).
PA planning is always influenced by local human activities in the form of land cost. In this study, we adopted the human influence (HI) index calculated using global human footprint data (NASA Socioeconomic Data and Applications Center, Palisades, NY; http:// dx.doi.org/10.7927/H4BP00QC, accessed 12 June 2015) to represent the land cost of each planning unit (1 km × 1 km grid cell). We defined special protection zones and special zones (classes Ӏ, II and III) of national park as national PAs in this study. The distribution of the focal PAs was obtained from polygon data (GIS data) downloaded from the National Land Information Division (https://nlftp. mlit.go.jp/ksj/index.html, accessed 10 May 2010) and was refined using a 1/25000 geographical map to improve accuracy before use.

| Target set, Marxan complementarity and irreplaceability analysis
Using the distribution records of freshwater fishes and macroinvertebrates, we performed a complementarity analysis using Marxan (Ball et al., 2009). Marxan is a practical tool that allows users to delineate conservation candidate sites by referencing "conservation area goodness" to maximize protection coverage at minimal cost. Here, we used an index of human activity as a cost layer in the Marxan analysis (see below). The conservation features we input into Marxan were the distribution grids of each freshwater fish (6303 grid cells for all species and 3183 grid cells for red-listed species) and macroinvertebrate species (1484 grid cells for all species and 737 grid cells for red-listed species). We assumed two protection scenarios by considering feasibility: in scenario 1, conservation feasibility is determined by human activities, and in scenario 2, conservation feasibility is equal throughout the planning units, meaning that no land is considered to be more or less desirable for human use (i.e. even cost . We used two kinds of conservation targets, reflecting two common objectives of conservation planning. First, a representation target was considered achieved if any portion of the species' distribution was included in conservation-priority grid cells. Second, a sliding target was considered achieved for a species where 100% of its occupied grid cells were included in conservation-priority grid cells when the species' area of occupancy is <50 grid cells (for fishes) or <5 grid cells (for macroinvertebrates), and if 10% of its occupied grid cells were included in conservation-priority grid cells when its area of occupancy is >1000 grid cells (for fishes) and >200 grid cells (for macroinvertebrates). For species with ranges that fall between these two thresholds, the representation target was interpolated using a linear regression on the log-transformed area of occupancy at the national scale (Maiorano et al., 2006;Rodrigues et al., 2004). The thresholds were intended to ensure species persistence nationally, and were determined based on the expert opinions of several freshwater scientists in Japan. The main reason for the use of separate thresholds for fishes and macroinvertebrates was the over 10-fold difference in average area of occupancy between the two taxa.
We used actual distribution records for the analysis because species distribution models of riverine animals that were built in previous studies, especially in Japan, did not show sufficient explanatory power at finer grain sizes such as 1 km × 1 km (Fukushima et al., 2007), and because errors of commission (false positives) in predicted occurrence can introduce strong biases in the results of complementarity analysis.
To identify conservation-priority areas, we ran Marxan 100 times under each scenario to obtain each solution and the Marxan selection frequency, which is used as a measure of irreplaceability for each planning unit. We classified the range of HI index values into five categories by aggregating the data into intervals of 20%.

| Status of freshwater organisms in Japan
We collected data for 131 freshwater fish species (

| Extant PAs and protection of fishes and aquatic macroinvertebrates
The number of conservation-priority grid cells obtained by the Marxan did not show clear differences between the two scenarios for either species group (Table 1)

| The features of conservation candidate sites
For all species of both groups, conservation-priority grid cells were unevenly distributed across the gradient of flood risk and were more heavily distributed at low or high flood risk (Figure 4; chi-squared test, p < .01; Appendix S6, S7). In particular, >40% of candidate conservation-priority grid cells were located in flood risk areas, and most of them were at high flood risk (i.e. Risk 3-5; Figure 4). The proportion of conservation-priority grid cells located in high flood risk areas was higher for red-listed species of both groups, and 80% of the candidate conservation-priority grid cells for macroinvertebrate species were located in flood risk areas. The number of conservation-priority grid cells was higher in areas of high HI index as well in those of low HI index (including extant PAs) in all cases (chi-squared test, p < .01; Appendix S6, S7).

| DISCUSS ION
This study suggests there is a great discrepancy with regard to biodiversity conservation in extant PAs in freshwater ecosystems.
In particular, only a few red-listed species in each group were represented within extant PAs. This is not surprising because many PAs had globally been established in locations that are disproportionately unimportant for biodiversity (Venter et al., 2018), but our results were some of the lowest rates among previous studies focusing on gaps between extant PAs and biodiversity (Lawrence et al., 2011).

HI index HI index HI index HI index
animals generally tends to increase in lowland areas (Tockner & Stanford, 2002). Nevertheless, the distributions of floodplaindependent species are restricted to remnant habitats in lowland areas because most historic floodplains worldwide have been developed for human use and are impacted by flood control measures such as artificial levees (Tockner & Stanford, 2002). Therefore, our model tends to select areas with high flooding risk and human influence as sites for additional FPAs. Floodplain conservation provides suitable habitats for many freshwater organisms and may help to control floodwaters, suggesting that establishing FPAs in high flood risk areas would be a win-win strategy both for conserving freshwater biodiversity and enhancing eco-DRR.
Establishment and management costs are crucial considerations for the designation of additional PAs, including FPAs. However, in our study system, the irreplaceability of each planning unit remained similar across the two scenarios for each species group, and the majority of conservation-priority grid cells for both taxa and species groups had high to medium HI-index value. These findings suggest that important sites for freshwater biodiversity under our conservation targets are already limited, and the selection of priority grid cells within more developed areas cannot be avoided. The species richness of many types of animals (including macroinvertebrates) is positively correlated with human population density at coarser grain size (e.g. 10 km ×10 km), whereas it is negatively correlated at finer grain size (e.g. 1 km ×1 km) (Pautasso, 2007), suggesting that small-scale habitats with high biodiversity are vulnerable in densely populated areas. These studies support our suggestions that additional PAs are essential in human-dominated areas (i.e. those with a high HI index) as well as in "high and far" areas (Joppa & Pfaff, 2009 (Sloey et al., 2015;Warner et al., 2018).
Flood disasters occurred at these places in 1424 and 1997, respectively, and the inhabitants decided to leave the regions because they could not afford the cost and resources needed to recover. Part of De Biesbosch has become a national park and part of Liberty Island a conservation reserve, representing some of the few FPAs in the world (Sloey et al., 2015;Warner et al., 2018).
According to several climate change models, the frequency of catastrophic flooding will increase (Hirabayashi et al., 2013), and many countries are likely to need considerable adaptation (Willner et al., 2018). In fact, more than half of the United States by area will need to at least double their protection against flood events within the next two decades (Willner et al., 2018). Under these scenarios, the establishment and management costs for additional FPAs could be less expensive than those for flood recovery. In addition, many countries will face serious population declines over the next century, which will likely lead to land abandonment (Queiroz et al., 2014) and could help to secure lands for additional FPAs.
The spatial resolutions of recent studies that focused on FPAs were mainly at the watershed scale (e.g. Saunders et al., 2002). This grain size is useful because riverine conservation planning often requires spatial configurations such as river networks. However, particularly in mountainous countries including Japan, it is not realistic to consider additional FPAs that are watershed size, because human-occupied areas are concentrated in a limited number of lowelevation areas (Marini et al., 2009). In this situation, a finer grain size, such as 1 km ×1 km, seems practical for considering sites for additional FPAs. Nevertheless, many aquatic animals, especially migratory fishes, do not necessarily complete their life cycle within a 1-km 2 area. Additionally, flood risk often influences by upstream river and riparian conditions (Capon & Pettit, 2018). Therefore, land managers should consider the condition of the entire river network when they select FPA sites, even if the proposed FPA itself is small.
Although the majority of identified conservation-priority areas had high flood risk, some of the stakeholders in these areas, particularly in developed areas, may not support the establishment of FPAs on their land (e.g. Kamal & Grodzinska-Jurczak, 2014), because displacement from an area often entails huge and varied costs (e.g. Maghfiroh & Hanaoka, 2019;Maldonado et al., 2013). Indeed, the resilience of individual livelihoods and other socioeconomic impacts often influence relocation decisions (Cong et al., 2018). Additionally, social capital, including community network and place identities, are also integral to relocation decisions (Chamlee-Wright & Storr, 2009;Cong et al., 2018). Thus, resolving trade-offs among livelihoods, attachments to place and community, and flood exposure should be fundamental for establishing FPAs in developed areas. In recent years, however, many landowners have personally experienced the socioeconomic risks that accompany environmental degradation or have become more aware of the prospect of increased socioeconomic risks and associated declines in human well-being in the future (Rieb et al., 2017). Moreover, experiences of natural disasters have been shown to increase awareness of the need for relocation (Bukvic et al., 2015). Additionally, stakeholders tend to support land-modification projects when the projects provide them with ecosystem services (Howe et al., 2014). Thus, one important way to achieve biodiversity conservation is not to pursue biodiversity conservation by itself, but to achieve it through floodplain management. For example, although the Kissimmee River restoration projects in Florida, USA, provide suitable habitats for many aquatic and terrestrial organisms, the main purpose of the projects is water management (Koebel, 1995). Similarly, flood control basins constructed in the Chitose River, Japan are used as farmland, but also serve as semi-natural wetlands that provide summer habitat for wetland species (e.g. fishes, aquatic insects, birds and plants), especially those that inhabit environments with hydrological variation Yamanaka et al., 2020). Also, floodplain ponds along the Ishikari River, Japan are slated for future use as floodwater retention ponds (unpublished information); the fragmented shapes of these ponds will serve as refugia from biological invasions for an endangered minnow . The designers of Bishan-Ang Mo Kio Park in Singapore, a well-known urban park, used the blue-green infrastructure approach to prevent flooding (by increasing retention capacity and reducing flow velocity) while also improving recreational function (Lim & Xenarios, 2021); as a result, the diversity of various taxa such as freshwater fishes, aquatic insects and birds have also increased within the park (Lim & Xenarios, 2021;Wilkinson et al., 2021). Thus, it is F I G U R E 4 Number of candidate conservation grid cells that belong to each flood risk and HI index class

Macroinvertebrates
All species Red-listed species

Freshwater fishes
All species Red-listed species

Number of candidate conservation grids
Number of candidate conservation grids

Number of candidate conservation grids
possible that a variety of river management plans could incidentally help biodiversity conservation. Proposed FPAs with ecosystem service functions such as flood DRR through water retention (Schober et al., 2015) may foster agreement among stakeholders and increase the likelihood of establishing additional FPAs.

ACK N OWLED G EM ENTS
We appreciate Dr. N. Takamura in National Institute for Environmental Studies for giving helpful advice. We also thank to Professor Monica Turner in Wisconsin University and two anonymous reviewers for giving us significant and helpful comments to enhancing the paper. This study was supported by the Environment Research and Technology Development fund (S9-4, and 4-1504) of the Ministry of Environment of Japan, funds provided by Grants in Aid for Scientific Research (no. 15K20842) from the Ministry of Education, Science and Culture, Japan, and Environment Research and Technology Development.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13517.

DATA AVA I L A B I L I T Y S TAT E M E N T
The species distributions of freshwater fishes and macroinvertebrates are available from http://www.nilim.go.jp/lab/fbg/ksnkankyo and http://www.biodic.go.jp. The distribution data of red-listed species are available upon request for research or application purposes (requests should be submitted through the above websites). Requests will be considered by the data owners (the Ministry of Land, Infrastructure, Transport and Tourism or the Ministry of the Environment, Japan), and data will be released after receiving approval from the respective custodians. Any requests, especially on behalf of non-Japanese speakers, may be supported by the corresponding author.