Trading-Off Fish Biodiversity, Food Security, and Hydropower in The

Trading-off ﬁsh biodiversity, food security, and hydropower in the Mekong River Basin

Guy Ziva,1,2, Eric Baranb, So Namc,3, Ignacio Rodríguez-Iturbed,1, and Simon A. Levina aDepartment of Ecology and Evolutionary Biology, and dDepartment of Civil and Environmental Engineering, Princeton University, Princeton, NJ 08544; bWorldFish Center*, Phnom Penh, Cambodia; and cInland Fisheries Research and Development Institute, Fisheries Administration, Phnom Penh, Cambodia

Contributed by Ignacio Rodríguez-Iturbe, January 28, 2012 (sent for review November 10, 2011) The Mekong River Basin, site of the biggest inland fishery in the season carrying capacity (BCC, Fig. 1B) of each species, which world, is undergoing massive hydropower development. Planned we assume to be proportional to fish population in either season. dams will block critical fish migration routes between the river’s downstream floodplains and upstream tributaries. Here we esti- Results mate fish biomass and biodiversity losses in numerous damming The MRC has formally adopted six dam-development scenarios scenarios using a simple ecological model of fish migration. Our for 2015 and 2030, known as Basin Development Plan 2 scenarios framework allows detailing trade-offs between dam locations, (BDP2) (15). These scenarios include between 16 and 78 tributary power production, and impacts on fish resources. We find that dams and up to 11 main-stem dams (Fig. 1 and Table 1). Using the completion of 78 dams on tributaries, which have not previ- our model, we estimate the impact of each scenario on the bio- ously been subject to strategic analysis, would have catastrophic mass of migratory species in the floodplains, as well as the number impacts on fish productivity and biodiversity. Our results argue for of species at risk for extinction because of habitat loss. We find reassessment of several dams planned, and call for a new regional that construction of all 78 tributary dams, excluding main-stem agreement on tributary development of the Mekong River Basin. dams, would produce less energy and pose greater environmental risk than the construction of just the upper six main-stem dams on sustainable development | hydropower planning | fish species richness the lower Mekong River (Table 1). Furthermore, although the

1995 Mekong Agreement requires international consultations SCIENCES

rom its origins in the Tibetan Himalayas in China, the Mekong before constructing main-stem dams, tributary dams are within ENVIRONMENTAL FRiver flows along a 4,600-km course toward the South China national jurisdiction, and necessitate only a “notification” to the Sea, passing through Myanmar, Lao People’s Democratic Re- MRC Joint Committee; this is despite their potentially significant public (Lao PDR), Thailand, Cambodia, and Vietnam. The transboundary impacts, as we demonstrate in this work. For Mekong River Basin (MRB) 800,000 km2 watershed is home to brevity, we choose to concentrate on a limited set of trade-offs, about 65 million people (1, 2), with about two-thirds of them and exclude main-stem dams from our analysis. fi We focus hereafter only on dams for which the fate is still

living in rural areas and relying on subsistence sheries for their SCIENCE diet (1, 3). Over 1 million tons of freshwater fish are caught an- undetermined, namely the 27 dams that have construction plan- SUSTAINABILITY nually in the Cambodian and Vietnamese floodplains alone (3–5), ned between 2015 and 2030 (Table 1). To evaluate how much fi making the sustainability of floodplains fisheries a major food each of these tributary dams may contribute to the loss of sh fi security concern in both nations. Apart from its important role as production, we compute the difference in average migratory sh a food source, the MRB is also a biodiversity hotspot for many biomass for all scenarios with and without each dam. Individually, fi species, in particular fish (6), and is characterized by large-scale the dams with the largest impact on sh biomass are Lower Se San fi fish migrations (7, 8). Both fish production and species richness 2 (LSS2, 9.3% drop in sh biomass basin-wide), Se Kong 3d are now threatened by imminent construction of hydropower (2.3%), Se Kong 3u (0.9%), and Se Kong 4 (0.75%). See Table S2 plants on tributaries as well as on the Mekong River itself (3, 9– for the complete list, including a similar ranking based on bio- 13). Although main-stem dams have been the subject of a thor- diversity loss. This analysis, however, does not provide a complete ough strategic assessment (14), transboundary cumulative impact picture of the combined impact of multiple dams. Therefore, we assessment for tributary dams is still lacking (2, 10). Here we analyze all possible permutations of these 27 dams in Fig. 2. Fig. A 27 fi develop a framework capable of filling this gap, and demonstrate 2 shows the impact of all 2 scenarios on migratory sh bio- the trade-offs between hydropower, food security (namely fish mass, plotted against their respective hydropower production. production), and biodiversity (endangered fish species). The scenario space is divided into eight subspaces by LSS2 and The Mekong River is the second most biodiverse river in the the four dams on Se Kong River. For the remaining 22 dams, all world (after the Amazon River). We identified 877 fish species in the MRB (not including estuary marine species). Regional species richness (Table S1) ranges from 484 species in the Mekong Author contributions: G.Z., E.B., I.R.-I., and S.A.L. designed research; G.Z., E.B., and S.N. collected data; G.Z. and E.B. performed research; G.Z., E.B., I.R.-I., and S.A.L. analyzed Delta to only 24 species in the Chinese headwaters far north. data; and G.Z., E.B., I.R.-I., and S.A.L. wrote the paper. As many as 103 of these species migrate upstream of Kratie in The authors declare no conflict of interest. Cambodia, and would be potentially impacted by hydropower 1To whom correspondence may be addressed. E-mail: [email protected] or irodrigu@ development. Our modeling approach, described in Methods, princeton.edu. computes the dry-season population and the contribution of each 2Present address: Natural Capital Project, Woods Institute for the Environment, Stanford locality to the wet-season productivity in the floodplains. Within University, Stanford, CA 94305. our theoretical framework, each upstream habitat has a local 3The conclusions of this research paper do not represent an official opinion or policy of carrying capacity of migratory fish (LCCMF) (Fig. 1A) that is IFReDI, the Fisheries Administration, or the Royal Government of Cambodia. determined by the local carrying capacity (LCC) in the dry-sea- *The WorldFish Center is an international, nonprofit research organization dedicated to son and the fraction of that LCC occupied by migratory (vs. reducing poverty and hunger by improving fisheries and aquaculture. WorldFish is a member of the CGIAR Consortium, a global partnership that unites organizations en- nonmigratory) species. Migratory species fraction at each site gaged in research for sustainable development. Its funders include governments, foun- is a function of distance-dependent migration cost, habitat suit- dations, and international and regional organizations. fi ability for different sh guilds, and interguild competition. This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. Summing over accessible LCCMF, we get the basin-wide dry- 1073/pnas.1201423109/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1201423109 PNAS Early Edition | 1of6 Fig. 1. Assessing impacts of hydropower development on migratory fish species. (A) The Mekong River (blue) flows from China, through Myanmar, Lao PDR, Thailand, Cambodia, and Vietnam (bottom left map). The relative dry-season migratory fish habitat capacity (stationary solution of Eq. 1) highlights the importance of the Se Kong, Sre Pok, and Se San subbasins for maintaining the migratory fish population (Inset shows expanded view and the location of five key planned dams). Kratie (star) in Cambodia is the border between the wet-season floodplains and the upstream habitats. (B) The BCC of each migratory fish species for different basin development scenarios described in the text. (C) Predicted population decline of each species based on lost habitat in each scenario, relative to an estimated historical population size in a pristine river. subspaces contain some of the 227 scenarios (Fig. S1). This figure scenarios with less than 5% biomass loss, biodiversity is relatively clearly demonstrates that at a regional scale a target hydropower unaffected, whereas above this threshold about six species will production level can be achieved in many ways, with markedly become endangered for every 1% of lost biomass. This finding is different environmental consequences. extremely important because it relates fish abundance, a pro- Given any target energy-production level, we highlight the visioning ecosystem service with significant value in the eyes of scenarios with the least environmental impact in Fig. 2A. These local stakeholders, to biodiversity, a supporting ecosystem ser- “Pareto-efficient” scenarios (PESN) show a clear nonlinear vice whose value is still vague or underestimated (3, 16, 17). Our trade-off between hydropower and fish biomass. Furthermore, findings that the trade-off illustrated in Fig. 2A is highly non- we observe three distinct regimes in the PESN (Fig. 2A, colored linear and exhibits three distinct regimes are robust to many of backgrounds). We find that fish biomass first decreases by about the model parameters (Fig. 2D). These are the main results of 0.3% (∼1,700 tons/y) for each additional teraWatt hour per year this article, and have far-reaching consequences for policy and (TWh/y) up to ∼14 TWh/y (70% of total span, value relative to decision-making, discussed below. BDP2 “Definite Future” scenario). Beyond that, the construction of dams on the Se Kong River causes 1.3% productivity loss Discussion (∼8,200 tons/y) per TWh/y up to 88% hydropower production, Although our work demonstrates the significant transboundary and the LSS2 dam amounts to 4% of fish loss (∼25,300 tons/y) impact of tributary dam development, the focus of scientific re- per TWh/y produced. The number of endangered species ex- search, media, and international efforts has been on avoiding hibits a similar trade-off with hydropower production (Fig. 2B). main-stem dam construction, in particular of the controversial The PESN in this case are those with the least biodiversity risks Xayaburi dam. Our analysis shows that construction of all plan- at each level of energy production. We see moderate (with LSS2) ned tributary dams, nearly all within Lao PDR national borders, or negligible (without LSS2) increase in the number of endan- would have graver impacts on fish biodiversity basin-wide and on gered species up to 12–14 TWh/y, followed by a steep increase in the Cambodian and Vietnamese floodplain’s fish productivity, biodiversity loss for each extra TWh/y. We also observe a signif- than the combined impact of the six upper main-stem dams on icant negative correlation (Pearson R = 0.9875, P < 0.001 based the lower Mekong River, including Xayaburi. on bootstrapping, n = 134,217,728) between migratory fish bio- Which tributary dams should be built and which should be mass, calculated based on the 48 dominant species, and the avoided? This decision could be made by assigning monetary number of endangered species, calculated based on predicted value to hydropower and fish biomass. However, the monetary population declines for all species. In fact, Fig. 2C shows that for valuation of subsistence fisheries is still a subject of debate and is

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1201423109 Ziv et al. Table 1. Impact assessment of different dam scenarios on fish biomass and fish species richness Biodiversity risks: Number of species... Δ(Hydropower Δ(Migratory No. of tributary No. of main- generation) fish biomass) Critically BDP2 scenario Year dams in LMB stem dams in LMB† (TWh/y) (%)‡ Vulnerable Endangered Endangered

Baseline 2000 16* 0 −82.1 +4.9 38 9 1 Deﬁnite Future 2015 41 0 ——78 11 1 No main-stem dams 2030 78§ 024−19.1 9 85 6 Six main-stem dams 2030 78 6 56.5 −23.6 2 90 8 on upper LMB|| No Cambodian 2030 78 9 70.1 −44.8 0 0 100 main-stem dams All 11 main-stem 2030 78 11 89.8 −51.3 0 1 100 dams Deﬁnite Future + six 2030 41 6 35.8 −6.9 64 30 2 main-stem dams on upper LMB||

Most of the biodiversity and fish production losses are because of tributary dams, in the absence of main-stem dams. Changes in hydropower production and biomass are shown relative to BDP2 Definite Future scenario (97.4 TWh/y and 52.0%, respectively). LMB, Lower Mekong Basin. *We include several decommissioned dams that are no longer included in the latest BDP2 but still pose a migration barrier for fish. These dams include Nam Ko, Nam Ngay, Huai Kum, Xelabam, and Lam Ta Khong. †All scenarios include seven main-stem dams in the Chinese Upper Mekong Basin, except for the “Baseline” scenario, which includes only the Manwan dam completed in 1996. We added the regulatory dam Ganlanba to the BDP2 count because it poses a migration barrier. ‡ A pristine river with no dams would support by definition 100% migratory fish biomass. One-percent biomass corresponds to annual catch of ∼6,400 tons of fish per year. §Our analysis considers 27 dams. We excluded dams planned upstream of BDP2 Definite Future dams (Duc Xuyen, Upper Kontum, Xeset 3, Nam Ngum 3,

Xekaman 4A, and Xekaman 4B) because our model predicts these will not impact ﬂoodplains production. Four dams on tributaries draining to the Tonle Sap SCIENCES

below Kratie (Battambang 1, Battambang 2, Pursat 1 and Pursat 2) were also excluded. ENVIRONMENTAL ||The six main-stem dams are the Pakbeng, Luang Prabang, Xayaburi, Paklay, Sanakham, and Pakchom dams. beyond the scope of this work. Nevertheless, the highly nonlinear socioeconomic progress is desirable, sustainable development shape of the PESN, and especially the separation into three dis- requires that unnecessary risks to ecosystems and environmental tinct regimes, can significantly help decision-making for this services, such as fish production and biodiversity, be avoided. In question. Specifically, it is clear that under a large range of sub- the MRB we find that some hydropower projects can and should SCIENCE stitutability between energy production and fish productivity, be avoided. Although the Mekong River and its tributaries are SUSTAINABILITY scenarios within the first regime are preferable to those in the expected to be influenced by climate change and anthropogenic second regime (with Se Kong cascade), and those scenarios in stressors (such as water extraction for domestic use of an in- the third regime (with LSS2) should probably be avoided. Hence, creasing population), damming will continue to dominate the the construction of the LSS2 dam is highly detrimental according impacts on migratory fish (see SI Discussion). Our analysis has to this analysis, and the benefits of the Se Kong cascade are shown that hydropower generation, nearly all planned to be questionable. These conclusions are based only on two aspects of exported from Lao PDR to Thailand and Vietnam, will have a much more complex socioecological and economic debate, and strong and nonlinear trade-offs with floodplains fisheries pro- a more definite answer necessitates the integration of other as- duction. This finding highlights the need for new regional pects, such as social costs of damming, and costs of nutrient and agreements addressing transboundary impacts of tributary dam sediment loss. Still, we can use the analysis of the PESN to design development in the Mekong River Basin. a simple decision-supporting tool (Fig. 3). For each energy level we color-code the percentage of scenarios within that range that Methods include a particular tributary dam. Dams are sorted by the energy- Data Collection and Curation. Our dataset was compiled from an extensive level “threshold” above which >50% of scenarios include that review of published and gray literature from 1936 to 2010 (including nu- project. As expected, LSS2 is last, which implies all PESN sce- merous projects Environmental Impact Assessments), of which 70 detailed ∼ fi fish species lists in 21 Mekong subbasins or main-stem sections were iden- narios exclude LSS2 until the demand is over 20 TWh/y. We nd fi fi that the proposed Se Kong and Nam Ou cascades are ordered ti ed. We corrected for taxonomic name changes using www. shbase.org. This dataset provides a comprehensive description of the fish biodiversity according to their catchment area, with the most upstream dams fi pattern in the Mekong River at a subbasin level (Table S1). Migratory species included rst and downstream dams appearing in PESN at a were identified using life-history information in four databases and pub- higher energy level. The Se Kong 3d dam is not included in the lished resources [Mekong Fish Database (20), www.fishbase.org, Halls and majority of any target energy level, suggesting this dam can always Kshatriya (21), and fishermen local knowledge (22)]. Finalization of the list be avoided. Finally, one can simplify this analysis further by was based on the expertise of one of the authors (E.B) as well as two ad- adopting a simple threshold rule, namely assume that for a given ditional Mekong fish experts. energy production target, all dams for which the energy level threshold is exceeded should be built. Compared with the PESN Fish Migration Model. Mekong fish fauna can be roughly classified into two scenarios within each energy level, both energy production (Fig. guilds: long-distance migratory and nonmigratory species. We include in the fi S2A) and fish biomass (Fig. S2B) are within 2.5% of the best latter guild sh that perform limited lateral migration. We developed a simple model of fish migration for understanding the relative abundance of model scenario in nearly all cases (Fig. S2C). We also find this D the two guilds within each locality. Although animal migrations in seasonally threshold rule works well for conservation (Fig. S2 ). varying environments have been studied in several models [see e.g., Holt and In recent years construction of hydropower dams has occurred Fryxell (23) and references therein], our model differs from these by two key at an unprecedented rate. It is now estimated that 50% of all large assumptions; (i) we assume that the upstream habitats are at their limiting rivers have been affected by dam construction (18, 19). Although (carrying) capacity at the end of each season, and (ii) we describe migration

Ziv et al. PNAS Early Edition | 3of6 Fig. 2. Trade-off analysis between hydropower generation, fish production, and biodiversity risks. (A) Change in migratory fish biomass versus change in total hydropower production in the MRB, both relative to BDP2 Definite Future scenario. Each point represents a different dam construction scenario. All 227 scenarios fall within eight areas (I–VIII), determined by the LSS2 and the cascade of Se Kong 3d (SK3d), Se Kong 3u (SK3u), Se Kong 4 (SK4), and Se Kong 5 (SK5) (Inset table in C). The PESN (black) show a highly nonlinear trend, which roughly divide into three regimes (green, yellow, and red; see main text). (B) Trade-off analysis between hydropower production and number of endangered species. PESN including LSS2 are marked black, and PESN with LSS2 excluded are marked green. (C) Correlation between change in migratory fish biomass and number of endangered species. (D) Sensitivity analysis. The slopes of the PESN three regimes are robust to uncertainty in model parameters. In particular, neither assuming uniform rank-abundance (uniform rank abundance), nor equal relative fraction of migratory species within each habitat (uniform migratory species fraction, using just total discharge as proxy of habitat), changes significantly the slopes of A. Ignoring spatial heterogeneity in runoff (uniform migratory species fraction and runoff, using only accessible area to measure fish habitat) still shows that impact of the LSS2 dam is much greater than other dams planned for 2030.

κi K2 by a simple “migration effectiveness” e(d) that measures the number of This model has a unique globally stable solution fi ¼ 1 − , where κi ≡ eðdi Þ K1 returning offspring as function of distance d. We shall call the floodplains is the seasonality of habitat B (hereafter labeled by subscript i), and di is the “habitat A,” and a particular upstream site “habitat B.” We formulate the distance of habitat B from the floodplains. We assume the population of model as a dynamical process for f(t), the fraction of migratory guild pop- each migratory species is proportional to sum over all accessible LCCMF ’ ulation in the dry-season in year t. Let K and K denote habitat B LCC in within the species natural range (Table S3). Because nearly all planned dams 1 2 fi fi fi season 1 (the dry season) and season 2 (the wet season), respectively. After do not consider sh passes, and the ef ciency of sh passes in the Mekong, a river characterized by very high biodiversity, large biomass, and several each annual cycle of migration, e(d)·f(t)·K1 migratory individuals return to migration pulses a year, has been strongly questioned, only habitats habitat B. These compete with K nonmigratory individuals that grew there 2 downstream of built dams are summed in calculating BCC. We note that in during the wet-season. This description holds regardless of the ecological this model, all species found within one locality must have the same value reason for downstream migration: that is, feeding migration (spawning of e(di), which justifies the use of a species-independent function. To un- occurs upstream and young migrate to feed in floodplains) or spawning derstand this, consider two species with migration effectiveness ev and eu. fi migration (adults nd refuge upstream, but go downstream to spawn); both For the former species to invade into a resident population of the latter, fi types are found in sh species of the Mekong River (24, 25). Assuming when the invading species is in scarcity ðfv ðtÞ ≈ 0Þ and the resident has the neutrality between the guilds, the number of migratory individuals surviving equilibrium abundance found before ðf ðtÞ¼1 − κ Þ, one should satisfy u eu the dry-season (LCCMF) is the result of the following dynamic relation: ev ·fv ðtÞ fv ðt þ 1Þ¼ > fv ðtÞ, which to first order in f (t) gives e > e . κþeu ·fu ðtÞþev ·fv ðtÞ v v u À Á ð Þ· ð Þ· Thus, invasion is possible if the migration effectiveness of the invading þ · ¼ e d f t K1 · ; f t 1 K1 K1 [1] species is larger than that of the resident species. This process naturally leads K2 þ eðdÞ·fðtÞ·K1 to a successional series of invasions until some biological reason limits e. This where for simplicity we ignored year-to-year variability in all parameters. limit is just the hallmark of the trade-off between fecundity and juvenile

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1201423109 Ziv et al. for the amount of resources available for fish (20, 27) and its variability shows clear correlation to variability of local population size (28). The same approach was used in other studies of fish biodiversity in the Mississippi-Missouri (29) and Indian Peninsula (30). The MRC Hydropower Database (15) was used to determine the geographical position and mean annual energy production of each dam listed in the BDP2. We used the United States Geological Survey (USGS) HydroSHEDS database to define the catchment area upstream of each

dam. The migration effectiveness e(di), which depends on distance from floodplains (we choose Phnom Penh as a convenient origin for distance measurement because it is between Cambodian Tonle Sap and Vietnamese floodplains; see Fig. 1A), is determined as explained in the legend of Fig. S4. Wetland habitats are less suitable than streams for migratory species (31, 32). We estimated that migration effectiveness is ∼50% lower in wetlands, based on the two surveys in upstream wetland villages (Table S4). The USGS global land-cover dataset was used to determine wetlands in each area. We defined wetlands as parcels the classification of which is irrigated grassland, wooded wet swamp, inland water, rice paddy/field, marsh wetland, or mangrove.

Estimating Biodiversity and Productivity Losses. To determine biodiversity risks we adopted the definitions of the International Union for Conservation of Nature (IUCN), also used in the IUCN Red List (33). The criterion we use is the expected population decline after 10 y IUCN defines species whose population is predicted to decline by over 30%, 50%, and 80% in 10 y as vul- Fig. 3. Decision-supporting tool for tributary hydropower development. At each hydropower production level (upper abscissa, discretized at 1 TWh/y) nerable, endangered, and critically endangered, respectively. We use as a “ ” we estimate the maximal remaining fish biomass (lower abscissa), and show baseline the pristine river (i.e., a basin without any dams; a situation what percentage of scenarios (within a range of 1% biomass of maxima at which actually existed before the year 1965). For brevity, we combine the that energy range; for example, blue rectangle of Fig. 2A) include each dam. endangered and critically endangered categories in Fig. 2. fl Stars denote the energy level threshold above which >50% of scenarios To estimate oodplains migratory biomass, we used species-level catch- “ ” fi include each project. We suggest that a simple threshold rule can be used to time series available from the bagnet ( dai ) shery in the Tonle Sap River – SCIENCES design the best dam scenario for a given energy production target. Dams (collected every year from 1998 2009). We sum the total catch per species (m) whose threshold is to the right of a vertical line at the target hydropower in all years to estimate relative abundances c of 48 dominant migratory ENVIRONMENTAL level should be avoided. This analysis improves on the ranking based only on species (Table S5). The remaining 55 species are rare, or were otherwise not (m) a single dimension (Fig. S3). caught in the surveys. Total biomass loss was calculated by multiplying c by each species population loss, and summing over species. survival. Species can either have many young, with little care and therefore Sensitivity Analysis. To check how our main result depends on modeling fi small survivorship (r selection), or fewer offspring with better chances of assumptions, we consider in Fig. 2D several simpli cations of our model. (m) reaching adulthood (K selection). Ignoring catch data [i.e., using c =1/48 for all dominant species (uniform SCIENCE rank-abundance)] does not change our main results. Furthermore, ignoring SUSTAINABILITY Geospatial Analysis. Seasonality (κ) was calculated based on hydrological seasonality and distance-dependent migration effectiveness by setting all fi modeling of the entire Mekong River at 5 arc-minute resolution by Costa- to the mean value (0.6) retains the nonlinear trade-off we observe (uniform Cabral et al. (26). We defined κ as the ratio of mean runoff production in the migratory species fraction). Finally, the disproportional deleterious impact of wet season (July–October) and dry season (December–May). Seasonality can, the LSS2 dam in Cambodia shows up even if we set all fi = 0.6 and ignore the in principle, be affected by other abiotic factors, in particular water temper- spatial heterogeneity of dry-season runoff, namely set K1 = 1 for all habitats ature. Unlike temperate rivers, however, the water temperature in tropical (uniform migratory species fraction and runoff). rivers such as the Mekong River does not change appreciably between the two seasons. For example, from monthly monitoring data between 1985 and ACKNOWLEDGMENTS. We thank David Tilman, Pamela Matson, and Jianguo 1997 (courtesy of the MRC), water temperature in Pakse is around 25.8 8Cin Wu for providing valuable comments on previous drafts of this manuscript. Fisheries data were kindly provided by the Inland Fisheries Research and the dry season and 27.3 8C in the wet season. All spatially explicit calculations Development Institute. We thank Ashley Halls for assistance in fish species were performed with ArcGIS 9.3 (ESRI), and the average quantities over in- classification. This work was supported in part by James S. McDonnell dividual catchments were used throughout this study. LCC in each catchment Foundation Grant 220020138 (to G.Z. and I.R.-I.); the AXA Research Fund area is proportional to the total discharge, namely runoff (precipitation minus (to G.Z.); and Mitsui-Bussan and the Government of Japan (Japan Mekong evapotranspiration) multiplied by surface area. This quantity is an indicator Fund) (E.B. and the WorldFish Center).

1. Mekong River Commission (2010) State of the Basin Report 2010 (Mekong River 11. Dugan PJ (2008) Mainstream dams as barriers to fish migration: International learning Commission, Vientiane, Lao PDR). and implications for the Mekong. Catch and Culture 14(3):9–15. 2. Mekong River Commission (2011) IWRM -Based Basin Development Strategy for the 12. Ferguson JW, Healey M, Dugan P, Barlow C (2011) Potential effects of dams on mi- Lower Mekong Basin (Mekong River Commission, Vientiane, Lao PDR). gratory fish in the Mekong River: Lessons from salmon in the Fraser and Columbia 3. Dugan PJ, et al. (2010) Fish migration, dams, and loss of ecosystem services in the Rivers. Environ Manage 47:141–159. Mekong basin. Ambio 39:344–348. 13. Marmulla G, ed (2001) Dams, Fish and Fisheries. Opportunities, Challenges and 4. UNEP (2010) Blue Harvest: Inland Fisheries as an Ecosystem Service (WorldFish Center, Conflict Resolution (FAO, Rome, Italy). Penang, Malaysia). 14. ICEM (2010) MRC Strategic Environmental Assessment of Hydropower on the Mekong 5. Barlow C, Baran E, Halls AS, Kshatriya M (2008) How much of the Mekong fish catch is Mainstream (ICEM, Hanoi, Vietnam). at risk from mainstream dam development? Catch and Culture 14(3):16–21. 15. Mekong River Commission (2009) Economic, Environmental and Social Impact As- 6. WWF (2008) First Contact in the Greater Mekong—New Species Discoveries (World sessment of Basin-Wide Water Resources Development Scenarios (Mekong River Wildlife Fund, Hanoi, Vietnam). Commission, Vientiane, Lao PDR). 7. Poulsen AF, Poeu O, Viravong S, Suntornratana U, Tung NT (2002) Fish Migrations of 16. Daily GC (1997) Nature’s Services: Societal Dependence on Natural Ecosystems (Island the Lower Mekong Basin: Implications for Development, Planning and Environmental Press, Washington, DC). Management (Mekong River Commission, Phnom Penh). 17. Tilman D (1999) The ecological consequences of changes in biodiversity: A search for 8. Baran E (2006) Fish Migration Triggers in the Lower Mekong Basin and Other Tropical general principles. Ecology 80:1455–1474. Freshwater Systems (Mekong River Commission, Ventiane, Lao PDR). 18. Lehner B, et al. (2011) High-resolution mapping of the world’s reservoirs and dams for 9. Baran E, Myschowoda C (2009) Dams and fisheries in the Mekong Basin. Aquat Ecosyst sustainable river-flow management. Frontiers in Ecology and the Environment 9: Health Manage 12:227–234. 494–502. 10. Grumbine RE, Xu JC (2011) Environment and development. Mekong hydropower 19. Nilsson C, Reidy CA, Dynesius M, Revenga C (2005) Fragmentation and flow regula- development. Science 332:178–179. tion of the world’s large river systems. Science 308:405–408.

Ziv et al. PNAS Early Edition | 5of6 20. Guegan JF, Lek S, Oberdorff T (1998) Energy availability and habitat heterogeneity 26. Costa-Cabral MC, et al. (2008) Landscape structure and use, climate, and water predict global riverine fish diversity. Nature 391:382–384. movement in the Mekong River basin. Hydrol Process 22:1731–1746. 21. Halls AS, Kshatriya M (2009) Modelling the Cumulative Barrier and Passage Effects of 27. Oberdorff T, Guegan JF, Hugueny B (1995) Global scale patterns of fish species rich- Mainstream Hydropower Dams on Migratory Fish Populations in the Lower Mekong ness in rivers. Ecography 18:345–352. Basin (Mekong River Commission, Vientiane, Lao PDR). 28. Bayley PB, Li HW (1996) Riverine Fishes. River Biota: Diversity and Dynamics, eds 22. Baran E, Starr P, Kura Y (2007) Influence of Built Structures on Tonle Sap Fisheries Petts GE, Calow P (Blackwell Science, Oxford, UK; Cambridge, MA), pp 92–122. (Cambodia National Mekong Committee and the WorldFish Center, Phnom Penh, 29. Muneepeerakul R, et al. (2008) Neutral metacommunity models predict fish diversity Cambodia). patterns in Mississippi-Missouri basin. Nature 453:220–222. 23. Holt RD, Fryxell JM (2011) Theoretical reflections on the evolution of migration. in 30. Lynch HJ, et al. (2011) How restructuring river connectivity changes freshwater fish Animal Migration: A Synthesis, eds Milner-Gulland EJ, Fryxell JM, Sinclair ARE (Oxford biodiversity and biogeography. Water Resour Res, 10.1029/2010WR010330. University Press, New York), pp 17–31. 31. Welcomme RL (2001) Inland Fisheries: Ecology and Management (Wiley-Blackwell, 24. Poulsen AF, Ouch P, Sintavong V, Ubolratana S, Tung NT (2002) Fish migrations of the Hoboken, NJ). Lower Mekong River Basin: implications for development, planning and environ- 32. Welcomme RL, Winemiller KO, Cowx IG (2006) Fish environmental guilds as a tool for mental management (Mekong River Commission, Phnom Penh, Cambodia). assessment of ecological condition of rivers. River Res Appl 22:377–396. 25. Baird IG, Flaherty MS, Phylavanh B (2003) Rhythms of the river: Lunar phases and 33. International Union for Conservation of Nature (IUCN) (2001) IUCN Red List Catego- migrations of small carps (Cyprinidae) in the Mekong River. Nat Hist Bulletin of the ries and Criteria: Version 3.1 (IUCN Species Survival Commission, Gland, Switzerland Siam Society 51(1):5–36. and Cambridge, UK).

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1201423109 Ziv et al. Supporting Information

Ziv et al. 10.1073/pnas.1201423109 SI Discussion is mainly because of wet-season runoff, with dry-season runoff Impacts of Dams on Upstream Fisheries and Downstream Habitat nearly unchanged (or slightly decreased) across most of the Quality. Beyond blocking migration routes, hydropower dams MRB. In our model, these changes would reflect in a small can have numerous other upstream and downstream impacts (1– (∼10% in most catchments) decrease in fi. For the Basin De- 3). Dams alter the habitat of nonmigratory fish species, risk velopment Plan 2 (BDP2) Definite Future scenario, the decrease extinction of local endemic populations, reduce sediment and in migratory species’ relative abundance would roughly translate nutrient flows toward downstream habitats, river deltas, and into a 18% decrease in floodplain fish productivity. Nevertheless, estuaries, impact water quality and temperature downstream, the additional biomass loss because of the 27 tributary dams cause eutrophication and deoxygenation by decomposing organic planned by 2030 would not change (∼39% decrease instead of matter, and even emit greenhouse gas from their reservoirs. ∼36% decrease). Thus, although climate change would have Damming might also change the timing of hydrological cues that fi fl fi signi cant impacts on ood risk and food scarcity (4), the po- set the onset of sh migration. Many of these impacts depend on tential deleterious impacts of tributary dams would still be the dam’s design and operation. In this work we choose to a major concern. concentrate on the unique aspect of the Mekong River Basin Net runoff, evapotranspiration from nonagricultural land, and (MRB) fisheries, namely their reliance on migratory fish species. rain-fed agriculture constitute ∼82% of the water use in the Thus, our findings are conservative because migratory—as well MRB. Of the remaining 18%, most of the water is used for ir- as nonmigratory—fish will have to struggle with additional dam- related issues not covered by our model. rigation (16%), and the remaining 2% is for domestic (0.8%) and industrial (1.2%) use (4). Although the population is ex- ∼ Climate Change and Other Anthropogenic Drivers. The impact of pected to grow to 111 million by 2030, this growth would not climate change and demographic growth by the year 2030 has change these numbers significantly (4). Furthermore, because recently been the focus of a detailed study (4). From the projected almost all arable land in the MRB is already cultivated, we do rainfall and evapotranspiration patterns, it is estimated that the not foresee net runoff to decrease much because of further land Mekong River runoff will increase by roughly 21%. This increase use change.

1. Rosenberg DM, et al. (1997) Large-scale impacts of hydroelectric development. Environ 3. Marmulla G (2001) FAO Fisheries Technical Paper. no. 419,edMarmullaG(FAO,Rome,Italy). Rev 5:27–54. 4. Eastham J, et al. (2008) Mekong River Basin Water Resources Assessment: Impacts of 2. Rosenberg DM, McCully P, Pringle CM (2000) Global-scale environmental effects of Climate Change (Commonwealth Scientiﬁc and Industrial Research Organisation, hydrological alterations: Introduction. Bioscience 50:746–751. Clayton South, Australia).

Ziv et al. www.pnas.org/cgi/content/short/1201423109 1of11 Fig. S1. (Continued)

Ziv et al. www.pnas.org/cgi/content/short/1201423109 2of11 Fig. S1. Trade-off analysis between hydropower generation and ﬁsh biomass for all 27 dams on Mekong tributaries. Scenarios including each particular dam are marked black. Scenarios without the dam are drawn in gray. All dams, except Lower Se San 2 (LSS2) and the dams on the Se Kong River (Se Kong 3d, Se Kong 3u, Se Kong 4, and Se Kong 5) have some scenarios in all areas I to VIII, deﬁned in Fig. 2A.

Ziv et al. www.pnas.org/cgi/content/short/1201423109 3of11 Fig. S2. Testing decision-tool for dam design. Comparison of “target” and “realized” results of using the threshold rule. A and B show the realized vs. target energy production (A)orﬁsh biomass (B), which are within 2.5% (C) deviation for all but three scenarios (black symbol, corresponding to target energy level below 5 TWh/y and above 20 TWh/y. D shows that those scenarios, which are optimized based on trade-off between biomass loss and hydropower production, also achieves good conservation goals, with all scenarios falling close to the “Pareto-Efﬁcient” scenarios (PESN) of Fig. 2B. Fig. S3 compares this optimal trade- off analysis to other dams ranking options.

Fig. S3. Comparison of different dam projects ranking. As an alternative to our optimal trade-off analysis, decision-makers may use individual the dam’s impacts to decide which dams to avoid. Here we show three other rankings and compare them to the optimal trade-off results. First, each of the 27 dams planned by 2030 was ranked according to four different perspectives: (i) Impact on floodplain fish production (Biomass, third column in Table S2); (ii) Impact on fish species richness (Conservation, fifth column in Table S2); (iii) Impact based on amount of hydropower produced (Hydropower, assuming larger plants have larger impacts); and (iv) Our optimized trade-off analysis (Fig. 3: ranking starts at LSS2 and continues upwards). We plot all pair-comparisons, as well as Pearson’s R coefficient and P value (calculated by bootstrapping, n = 27). We find that none of the other ranking methods would give the same recom- mendations as our optimal trade-off decision-tool, although conservation and biomass do correlate with the optimal trade-off ranking.

Ziv et al. www.pnas.org/cgi/content/short/1201423109 4of11 Fig. S4. Distance dependence of migration effectiveness. The migration effectiveness e(d) is the number of successfully returning offspring per individual migrant, and is a function of the distance d (measured for convenience in kilometers from Phnom Penh). For two subbasins (China upper reach and China headwaters, red squares) we ﬁnd no migratory species, which by our theoretical model requires that e(d) < κ for these habitats. For the other subbasins we demand that e(d) > κ so migration is maintained (black squares). Each square symbol shows the average (and SD) of one subbasin’s seasonality κ. The fraction of migratory species (f) is also known at several villages along the Mekong main stem and some tributaries (Table S4). From these, and the respective sea- sonalities at each village location, the local migration effectiveness is estimated in main-stem villages (ﬁlled circles) and tributary villages (empty circles). The = function eðdÞ ≡ 1:4·ð3400 − dÞ1 3 (green line) was adjusted to obey the abovementioned inequalities and pass through the average value of all points. (The exact functional form for e(d) is unimportant, as it does not change any of the conclusions in this article.)

Ziv et al. www.pnas.org/cgi/content/short/1201423109 5of11 Fig. S5. Subbasins of the lower and upper Mekong. CH, China headwaters; CL, China lower reach; CM, China middle reach; CSV, main-stem Chiang Saen to Vientiane; CU, China upper reach; KFST, main-stem Khone Falls to Stung Treng; KPP, main-stem Kratie to Phnom Penh; MC, Mun/Chi; MD, Mekong Delta; MY, Myanmar Mekong; NK, Nam Kadinh; NM, Nam Mang; NN, Nam Ngum; NO, Nam Ou; SG, Songkhram; SK, Se Kong; SP, Sre Pok SS, Se San; STK, main-stem Stung Treng to Kratie; TS, Tonle Sap; VKF, main-stem Vientiane to Khone Falls; XBF, Xe Bang Fai; XBH, Xe Bang Hiang. Kratie (star) marks the upper limit of wet- season ﬂoodplains.

Table S1. Number of Mekong fish species in basin and subbasin level Number of fish species Number of fish species Subbasin (no. migratory species) Subbasin (no. migratory species)

Mekong Delta 484 (76) Sre Pok 240 (82) Tonle Sap 328 (96) Se San 133 (54) Main-stem, Stung Treng–Kratie 204 (77) Se Kong 213 (65) Main-stem, Khone Falls–Stung Treng 190 (76) Mun/Chi 266 (89) Main-stem, Vientiane–Khone Falls 191 (79) Xe Bang Fai 157 (38) Main-stem, Chiang Saen–Vientiane 140 (45) Xe Bang Hiang 160 (46) China lower reach 122 (20) Songkhram 214 (69) China middle reach 48 (4) Nam Theun/Nam Kadinh 97 (18) China upper reach 34 (0) Nam Mang 57 (10) China headwaters 24 (0) Nam Ngum 155 (38) Nam Ou 136 (38)

We compiled 70 publications and references into a list of species in 21 Mekong subbasins and main-stem sections. We identiﬁed 877 unique species in the entire MRB, from which 103 are long-distance migratory species potentially impacted by dam construction. This table shows the total number of speciesand the number of migratory species (in parenthesis) in each subbasin and in sections of the main stem. The complete list of sources and species is available upon request. See Fig. S5 for geographic location of each subbasin.

Ziv et al. www.pnas.org/cgi/content/short/1201423109 6of11 Table S2. Impact of individual dams on fish productivity and biodiversity Average Δ(migratory Rank (impact on Average Δ (number of Rank (impact on fish Dam biomass) (%) fish biomass) newly endangered species) species richness)

Lower Se San 2 9.29 1 56.29 1 Se Kong 3d 2.29 2 9.42 2 Se Kong 3up 0.90 3 3.47 3 Se Kong 4 0.75 4 3.02 4 Nam Ou 1 0.49 5 1.99 5 Nam Kong 1 0.35 6 1.77 6 Nam Ngiep-regulating dam 0.28 7 1.76 7 Nam Ngiep 1 0.28 8 1.70 8 Nam Theun 1 0.26 9 1.43 9 Nam Ou 2 0.26 10 0.86 12 Se Kong 5 0.25 11 0.93 11 Nam Tha 1 0.22 12 1.33 13 Nam Lik 1 0.22 13 0.89 10 Nam Ou 3 0.16 14 0.46 15 Nam Suang 1 0.13 15 0.76 17 Xepian-Xenamnoy 0.11 16 0.36 18 Nam Suang 2 0.10 17 0.49 14 Nam Beng 0.07 18 0.49 20 Xe Katam 0.06 19 0.19 16 Nam Pha 0.06 20 0.40 22 Nam Ou 4 0.05 21 0.15 19 Nam Phak 0.03 22 0.23 21 Houay Lamphan 0.03 23 0.10 25 Nam Ou 5 0.03 24 0.06 23 Nam San 3 0.02 25 0.13 24 Nam Ou 6 0.01 26 0.02 26 Nam Ou 7 0.01 27 0.01 27

For each of the 27 dams we calculate the difference between mean ﬁsh biomass and number of endangered species in all scenarios with and without that dam. Dams are ranked starting at the worst dam (i.e., the one having the largest impact). The correlation between average ﬁsh biomass loss and average number of endangered species is nearly 1 (Pearson R = 0.996, P < 0.001, n = 27).

Ziv et al. www.pnas.org/cgi/content/short/1201423109 7of11 Table S3. Presence and absence data of migratory species Latin name STK KFST VKF CSV CL CM CU CH SP SS SK MC XBF XBH SG NK NM NN NO

Aaptosyax grypus ● ● ● ○○○○○○●○● ○ ○ ●○ ○ ○ ● Acanthopsoides delphax ○ ○ ● ●○○○○●●○● ● ○ ●○ ○ ○ ○ Amblyrhynchichthys truncatus ○ ● ● ●○○○○●○●● ○ ○ ●○ ● ● ○ Anguilla marmorata ● ● ● ●○○○○●○●● ○ ○ ●● ○ ○ ● Bagarius yarrelli ● ● ● ●●●○○●●●● ● ● ●● ○ ● ● Bangana behri ● ● ● ○○○○○●●●● ○ ○ ○○ ○ ● ● Bangana pierrei ● ○ ● ○○○○○○○●● ● ○ ○● ○ ○ ● Barbichthys laevis ● ● ○ ○○○○○●○○● ○ ○ ●○ ○ ○ ○ Brachirus harmandi ● ● ● ○○○○○●○●● ● ● ●○ ○ ● ○ Catlocarpio siamensis ● ● ● ●○○○○○○●● ○ ○ ●○ ○ ○ ● Chitala blanci ● ● ● ●○○○○●●●● ● ● ●○ ○ ● ● Cirrhinus caudimaculatus ● ○ ○ ○○○○○○○○● ○ ○ ●○ ○ ○ ○ Cirrhinus jullieni ● ● ● ●○○○○●●●● ○ ○ ○○ ○ ○ ○ Cirrhinus microlepis ● ● ● ○○○○○●●●● ● ○ ●○ ○ ○ ○ Cirrhinus molitorella ● ● ● ●●○○○●●●● ● ● ●● ● ● ● Clupisoma sinense ○ ● ● ○●●○○●●○● ○ ○ ○● ○ ● ○ Cosmochilus harmandi ● ● ● ●○○○○●●●● ○ ● ●○ ○ ● ● Crossocheilus atrilimes ● ○ ● ●○○○○●●●● ● ● ○○ ● ● ○ Crossocheilus reticulatus ● ● ● ○●○○○●●●● ● ● ●○ ● ● ● Cyclocheilichthys apogon ● ● ● ●○○○○●○●● ● ● ●○ ○ ● ○ Cyclocheilichthys armatus ● ● ● ●○○○○●●●● ● ○ ●○ ○ ● ○ Cyclocheilichthys enoplus ● ● ● ○○○○○●●●● ○ ● ●○ ○ ● ○ Cyclocheilichthys furcatus ● ● ● ●○○○○●○○● ○ ○ ●○ ○ ○ ○ Cyclocheilichthys heteronema ● ○ ○ ○○○○○●○○● ○ ○ ○○ ○ ○ ○ Cynoglossus microlepis ● ● ● ○○○○○●○○○ ○ ○ ○○ ○ ○ ○ Dasyatis laosensis ○ ● ● ●○○○○●●●● ○ ○ ●○ ○ ○ ● Datnioides undecimradiatus ● ● ● ○○○○○●●●● ● ● ●○ ○ ○ ○ Epalzeorhynchos frenatus ● ● ● ○○○○○○○●● ● ● ●○ ○ ○ ○ Epalzeorhynchos munense ● ○ ○ ●○○○○○○○● ● ○ ●○ ○ ● ○ Garra fasciacauda ○ ● ● ●○○○○●○●● ● ● ○● ○ ○ ○ Gyrinocheilus pennocki ● ● ● ○○○○○●●●● ○ ● ●● ○ ○ ○ Helicophagus leptorhynchus ● ○ ○ ●○○○○○○○● ○ ● ○○ ○ ● ● Helicophagus waandersii ○ ● ● ●○○○○●●●● ○ ● ●○ ○ ○ ○ Hemibagrus ﬁlamentus ● ● ○ ●○○○○●●●● ● ● ●○ ○ ● ○ Hemibagrus wyckii ● ● ● ●○○○○●●●● ○ ○ ●○ ○ ● ● Hemibagrus wyckioides ● ● ● ●●○○○●●●● ○ ● ●● ○ ● ● Hemisilurus mekongensis ● ● ● ○○○○○●○○● ○ ○ ●○ ○ ○ ● Henicorhynchus lobatus ● ● ● ○○○○○●●●● ● ○ ●○ ○ ● ● Henicorhynchus siamensis ● ● ● ○○○○○●●●● ● ● ●● ● ● ● Himantura krempﬁ ○ ○ ○ ○○○○○●○○○ ○ ○ ○○ ○ ○ ○ Hypsibarbus lagleri ● ● ● ●○○○○●●●● ○ ● ●○ ○ ○ ○ Hypsibarbus malcolmi ● ● ● ○○○○○●●●● ○ ● ●○ ○ ○ ● Hypsibarbus pierrei ● ● ● ●●○○○●○○● ○ ○ ○○ ○ ○ ○ Hypsibarbus vernayi ○ ○ ○ ○●○○○○○○● ● ○ ●● ○ ○ ○ Hypsibarbus wetmorei ● ● ● ○○○○○●●●● ○ ● ●○ ○ ○ ● Labiobarbus leptocheilus ○ ● ● ○○○○○○●●● ● ○ ●○ ○ ● ● Labiobarbus lineatus ○ ○ ○ ●●○○○○●○● ● ● ●○ ● ○ ○ Labiobarbus siamensis ● ● ● ●○○○○●○○● ○ ○ ●○ ○ ○ ○ Leptobarbus hoevenii ● ● ● ○○○○○●●●● ○ ● ○○ ○ ○ ○ Lobocheilos cryptopogon ○ ● ○ ○○○○○○○○● ○ ○ ○○ ○ ○ ○ Lobocheilos melanotaenia ● ● ● ○●○○○●●●● ● ○ ○○ ○ ● ○ Luciocyprinus striolatus ○ ○ ○ ●●○○○○○●○ ○ ○ ○● ○ ● ● Luciosoma bleekeri ● ● ● ○○○○○●●●● ○ ○ ●○ ○ ● ○ Mekongina erythrospila ● ● ● ○○○○○●●●● ● ● ○● ○ ○ ● Osteochilus enneaporos* ○ ○ ○ ○○○○○○○○○ ○ ○ ○○ ○ ○ ○ Osteochilus microcephalus ● ● ● ●○○○○●●●● ○ ● ●○ ○ ● ○ Osteochilus schlegelii ● ○ ● ○○○○○●○○○ ○ ○ ○○ ○ ● ○ Osteochilus waandersii ● ● ● ○○○○○●●●● ● ○ ●○ ● ● ○ Pangasianodon gigas ● ● ● ●○○○○●○○● ○ ○ ●○ ○ ○ ● Pangasianodon hypophthalmus ● ● ● ●○○○○●●●● ○ ○ ●○ ○ ● ● Pangasius bocourti ● ● ● ○○○○○●●●● ○ ○ ●○ ○ ○ ○ Pangasius conchophilus ● ● ● ○○○○○●●●● ○ ● ●○ ○ ● ○ Pangasius djambal ○ ● ● ○○○○○●○○● ○ ○ ○○ ○ ○ ○

Ziv et al. www.pnas.org/cgi/content/short/1201423109 8of11 Table S3. Cont. Latin name STK KFST VKF CSV CL CM CU CH SP SS SK MC XBF XBH SG NK NM NN NO

Pangasius elongatus ● ○ ○ ○○○○○○○○○ ○ ○ ○○ ○ ○ ○ Pangasius krempﬁ ● ● ● ●○○○○●●●● ○ ○ ○○ ○ ○ ○ Pangasius kunyit ○ ○ ● ○○○○○●○○● ○ ○ ○○ ○ ○ ○ Pangasius larnaudii ● ● ● ●○○○○●●●● ● ○ ●○ ○ ○ ● Pangasius macronema ● ● ● ○○○○○●●●● ○ ● ●○ ○ ● ● Pangasius mekongensis ● ○ ○ ○○○○○●○○○ ○ ○ ○○ ○ ○ ○ Pangasius nasutus ○ ○ ○ ○●○○○○○○○ ○ ○ ○○ ○ ○ ○ Pangasius pangasius* ○ ○ ○ ○○○○○○○○○ ○ ○ ○○ ○ ○ ○ Pangasius polyuranodon ○ ● ● ○○○○○●○○○ ○ ○ ○○ ○ ○ ○ Pangasius sanitwongsei ● ● ● ●●○○○○○○● ○ ○ ●○ ○ ○ ● Paralaubuca harmandi ○ ○ ○ ○○○○○●○○● ○ ○ ●○ ○ ○ ○ Paralaubuca riveroi ● ○ ● ○○○○○●○●● ○ ○ ●○ ○ ○ ○ Paralaubuca typus ● ● ● ●○○○○●●●● ● ● ●○ ○ ○ ○ Phalacronotus apogon ● ● ● ●○○○○●○●● ● ● ●○ ● ○ ● Phalacronotus bleekeri ● ● ● ○●○○○●●●● ○ ● ●○ ○ ● ○ Probarbus jullieni ● ● ● ○○○○○●●●● ○ ● ●○ ○ ○ ● Probarbus labeamajor ○ ● ● ○○○○○●●●● ○ ○ ○○ ○ ● ○ Probarbus labeaminor ● ● ○ ○○○○○●○○● ● ○ ●○ ○ ○ ○ Pseudolais micronemus ● ○ ● ○●○○○●○○● ○ ○ ○○ ○ ○ ○ Pseudolais pleurotaenia ● ● ● ●○○○○●●●● ○ ● ●○ ○ ● ○ Puntioplites bulu ○ ○ ○ ○○○○○●○○○ ○ ○ ○○ ○ ○ ○ Puntioplites falcifer ● ● ● ●○○○○●●●● ● ● ●● ○ ● ● Puntioplites proctozystron ● ● ● ○●○○○●○●● ○ ● ●● ○ ○ ● Puntioplites waandersi ○ ○ ○ ○●○○○○○○● ○ ○ ●○ ○ ○ ○ Raiamas guttatus ● ● ● ●●●○○●●●● ● ● ●● ○ ○ ● Rasbora aurotaenia ● ● ○ ○○○○○●○○● ○ ○ ●○ ○ ○ ○ Scaphognathops bandanensis ● ● ● ●○○○○●●●● ● ● ●○ ○ ○ ○ Scaphognathops stejnegeri ● ● ● ○○○○○●○●● ● ● ○● ○ ○ ● Setipinna melanochir ● ● ○ ○○○○○○○○● ○ ○ ●○ ○ ○ ○ Sikukia gudgeri ● ● ● ○○○○○●○●● ● ● ●○ ○ ○ ● Sikukia stejnegeri ● ○ ○ ●●○○○○○○● ○ ● ○○ ○ ○ ○ Syncrossus beauforti ● ○ ○ ●●○○○●●●● ● ● ●○ ● ● ○ Syncrossus helodes ● ● ● ●○○○○●●○● ● ● ●○ ○ ○ ○ Tenualosa thibaudeaui ● ● ● ○○○○○○●●● ● ● ●○ ○ ○ ● Tenualosa toli ○ ○ ● ○○○○○●○○○ ○ ○ ○○ ○ ○ ○ Thynnichthys thynnoides ● ● ● ●○○○○●●●● ○ ● ●○ ○ ○ ○ Tor sinensis ○ ○ ● ●●●○○●○●○ ● ○ ○● ○ ● ○ Tor tambroides ○ ● ● ●○○○○●○●○ ○ ● ○● ● ● ● Wallago leerii ○ ● ● ●○○○○●●●● ○ ● ●○ ○ ● ● Yasuhikotakia modesta ● ● ● ○○○○○●●●● ● ● ●○ ○ ○ ●

The presence (●) or absence (○) of each of the 103 species we identiﬁed as migrating upstream of Kratie toward 19 subbasins and main-stem sections. CH, China headwaters; CL, China lower reach; CM, China middle reach; CSV, main-stem Chiang Saen to Vientiane; CU, China upper reach; KFST, main-stem Khone Falls to Stung Treng; MC, Mun/Chi; NK, Nam Kadinh; NM, Nam Mang; NN, Nam Ngum; NO, Nam Ou; SG, Songkhram; SK, Se Kong; SP, Sre Pok SS, Se San; STK, main-stem Stung Treng to Kratie; VKF, main-stem Vientiane to Khone Falls; XBF, Xe Bang Fai; XBH, Xe Bang Hiang. *These species have no speciﬁc subbasin data. Hence we could not evaluate their extinction risk because of damming, and they were removed from the analysis.

Ziv et al. www.pnas.org/cgi/content/short/1201423109 9of11 Table S4. Fraction of migratory species in different habitats Village name Distance to Phnom Penh (km) Habitat type* Percentage migratory ﬁsh biomass f† Source‡ Seasonality κ

Sre Sronok 396 Tributary 43 AMCF 6.8 Pres Bang 428 Tributary 21 AMCF 7.7 Banfang 478 Tributary 44 AMCF 6.3 Day Lo 504 Tributary 61 AMCF 5.1 Khongpat 1008 Tributary 42 Warren 10.5 Ban Nam Ngieb 1056 Tributary 20 AMCF 7.6 Naxaeng§ 1407 Tributary 35 Vattenfall 6.2 Xiengdet§ 1436 Tributary 13 Vattenfall 6.8 Phonsavat§ 1381 Tributary 57 Vattenfall 5.0 Songphatong§ 1319 Tributary 40 Vattenfall 6.6 Pram 140 Main stem 57 AMCF 9.2 Sandan 229 Main stem 78 AMCF 5.7 Koh Khne 286 Main stem 29 AMCF 5.0 Kang Memai 328 Main stem 54 AMCF 5.0 Ou Run 375 Main stem 51 AMCF 5.4 Nalair 764 Main stem 58 AMCF 6.8 Song-khon 803 Main stem 58 AMCF 8.6 Ban Nam Kum 820 Main stem 37 AMCF 12.1 Ban Mouang Sum 877 Main stem 54 AMCF 8.2 Ban Xinh Xay 1038 Main stem 65 AMCF 8.0 Ban Thamuang 1187 Main stem 73 AMCF 8.2 Huasai 1213 Main stem 59 AMCF 8.9 Peam chumnik 1220 Main stem 80 AMCF 8.9 Pa-sak 1220 Main stem 55 AMCF 8.9 Ban Pha O 1649 Main stem 54 AMCF 6.6 Ban Done 1904 Main stem 56 AMCF 7.8 Phaphang 940 Wetlands 28 AMCF 6.1 Nongbueng 919 Wetlands 17 AMCF 8.5

Literature review values for percentage of migratory fish species along the Mekong River. *AMCF listed habitat as main stem, tributary, or floodplain. To avoid confusion, we renamed the last category “wetland,” to differentiate these from the wet- season habitat below Kratie. † Calculation of f for AMCF data used total catch of guilds 2, 3, 8, and 9 divided by total catch. In Warren and Vattenfall, the total catch of species classified as “migratory” (Table S2) was divided by the overall catch. ‡Assessment of Mekong Capture Fisheries (AMCF): guild-level catch, monitored between December 2003 and November 2004 (1). Warren: Ecological Impact Assessment study in Nam Hinboum (2). Vattenfall: Ecological Impact Assessment in Nam Ngum subbasin (3). §Distance measured along streamlines from Nam Lik 2 site, Nam Ngum 3 site, Nam Ngum 2 site, and Nam Ngum 1 site to Phnom Penh in Cambodia.

1. Halls A (2010) Estimation of annual yield of fish by guild in the Lower Mekong Basin (WorldFish Center, Phnom Penh, Cambodia). 2. Warren TJ (1999) A monitoring study to assess the localized impacts created by the Nam Theun-Hinboun hydro-scheme on fisheries and fish populations (Theun-Hinboun Power Company, Vientiane, Lao PDR). 3. Rydgren B, et al. (2009) Preparing the Cumulative Impact Assessment for the Nam Ngum 3 Hydropower Project (Asian Development Bank, Manila, Philippines).

Ziv et al. www.pnas.org/cgi/content/short/1201423109 10 of 11 Table S5. Relative abundance of migratory ﬁsh species in the Cambodian ﬂoodplains Latin name c(m) Latin name c(m)

Henicorhynchus lobatus 0.234025 Cyclocheilichthys armatus 0.0019 Lobocheilos cryptopogon 0.192042 Luciosoma bleekeri 0.000928 Henicorhynchus siamensis 0.144327 Pangasius bocourti 0.000734 Labiobarbus lineatus 0.095536 Probarbus labeamajor 0.000724 Puntioplites proctozystron 0.037989 Hemibagrus wyckii 0.000672 Thynnichthys thynnoides 0.036847 Hemisilurus mekongensis 0.000661 Cirrhinus microlepis 0.032626 Catlocarpio siamensis 0.000626 Labiobarbus siamensis 0.030607 Clupisoma sinense 0.000597 Cyclocheilichthys enoplus 0.029903 Osteochilus schlegelii 0.000468 Pangasius larnaudii 0.025448 Puntioplites bulu 0.000244 Yasuhikotakia modesta 0.021483 Hemibagrus ﬁlamentus 0.000178 Pseudolais pleurotaenia 0.018927 Hypsibarbus vernayi 0.000173 Syncrossus helodes 0.014678 Cirrhinus jullieni 0.000159 Amblyrhynchichthys truncatus 0.014298 Pangasius sanitwongsei 0.0001 Hypsibarbus malcolmi 0.012884 Raiamas guttatus 8.12E-05 Cosmochilus harmandi 0.009584 Epalzeorhynchos frenatus 7.49E-05 Pangasianodon hypophthalmus 0.009159 Pangasius krempﬁ 6.32E-05 Pangasius conchophilus 0.007945 Epalzeorhynchos munense 4.13E-05 Probarbus jullieni 0.007105 Cyclocheilichthys furcatus 3.55E-05 Helicophagus waandersii 0.004939 Bagarius yarrelli 2.81E-05 Barbichthys laevis 0.003411 Pangasianodon gigas 2.53E-05 Tenualosa thibaudeaui 0.002671 Hypsibarbus wetmorei 1.71E-05 Leptobarbus hoevenii 0.002562 Bangana behri 5.76E-06 Setipinna melanochir 0.002469 Tenualosa toli 4.48E-07

Species not listed in table are rare, or were otherwise not caught in the years and lots surveyed. Our analysis assumes the contribution of these to floodplains biomass is negligible. Values are based on total catch between 1998 and 2009 in stratified survey of fishing lots (“dais”) along the Tonle Sap River, conducted as part of the Fisheries, Ecology Valuation and Mitigation component of the MRC Fisheries Program in cooperation with the Inland Fisheries Research and Development Institute.

Ziv et al. www.pnas.org/cgi/content/short/1201423109 11 of 11