On this part, we current an outline of the information assortment course of and reveal an utility the place GDAT is used to calculate dam-induced world floor water modifications. The primary subsection, “International Dam Tracker (GDAT),” presents the GDAT database and describes the information sources. The second subsection,“GDAT Use Case,” particulars our utility of GDAT and descriptions an algorithm for calculating floor water modifications inside dam catchment areas.
International dam tracker (GDAT)
The International Dam Tracker (GDAT), a database with 35,140 dams in all continents besides Antarctica, is an authentic knowledge compilation effort24. We construct upon present world and regional dam databases, together with AQUASTAT (www.fao.org/aquastat), the International Reservoir and Dam Database (GRanD – globaldamwatch.org/grand)25, and the World Assets Institute (WRI) database (datasets.wri.org/dataset/globalpowerplantdatabase)26. For every nation, we consolidate these world and area databases with accessible knowledge from authorities businesses, lecturers, and nonprofits. Greater than 90% of dams in GDAT are geocoded (n = 31,780), and the coordinates are extensively verified utilizing Google Earth and different geospatial software program. This course of creates an in depth cross-validated database that can be utilized for evaluation at varied geographical scales.
Description of regional knowledge sources
This part supplies an outline of main regional or nationwide knowledge sources that contributed considerably to GDAT. Detailed info on the information sources for every nation is offered within the Area Highlights part. Current dam databases have restricted protection of creating and low-income international locations. As such, our knowledge assortment effort focuses on international locations in Africa, Asia, and South America, the place we collected major knowledge from administrative and different sources. Information for international locations in Europe, North America, and Oceania are primarily obtained from present world dam databases.
Dams in India have been collected from the India Water Assets Data System (WRIS), a database that’s maintained and funded by the federal government of India (indiawris.gov.in/wris). Reservoir quantity knowledge are usually not offered by the WRIS and, therefore, are usually not presently accessible in GDAT for dams in India.
Dams in China have been compiled from a number of nongovernmental sources: AQUASTAT, GRanD, Worldwide Rivers, and the Consultative Group for Worldwide Agricultural Analysis (CGIAR). Whereas the Ministry of Water Assets of the Folks’s Republic of China maintains a database on dams in China, it isn’t launched to the general public. Therefore, regardless that the GDAT database has compiled greater than 1,000 dams in China, the precise dam depend is anticipated to be a lot larger, on condition that many smaller dams could be lacking from the database.
For Continental Southeast Asia, GDAT database builds upon three essential knowledge sources: AQUASTAT, GRanD, and CGIAR Larger Mekong (wle-mekong.cgiar.org). The CGIAR database focuses on the key river basins of continental Southeast Asia (Mekong, Purple, Salween, Irrawaddy) and comprises info for greater than 800 dams that span China, Thailand, Vietnam, Laos, Cambodia, and Myanmar.
Dams within the Center East got here from AQUASTAT and GRanD, besides Turkey and Iran, the place we obtained knowledge from authorities web sites. The Turkish Common Directorate of State Hydraulic Works (www.dsi.gov.tr)publishes info for greater than 600 dams, and the Iran Water Assets Administration Firm (www.wrm.ir) publishes info for greater than 1,000. Neither database, nevertheless, is geolocated. Therefore, the geographic coordinates for these two international locations are manually verified by way of GeoNames and Google searches.
Dams in Brazil have been compiled from 4 sources: AQUASTAT, GRanD, the Brazilian Nationwide Dam Security Data System (SNISB), and Dams in Amazonia (archive.internationalrivers.org). The SNISB for 2016, the newest model accessible, contains virtually 23,000 dams. We excluded all dams and reservoirs used completely for tailings and unsafe waste storage. The SNISB contains constructions with a peak better than or equal to fifteen meters and a complete reservoir capability better than 3hm3. In consequence, the GDAT database is probably going lacking a substantial variety of small dams in Brazil that aren’t monitored by the SNISB.
Whereas a lot of the new knowledge in GDAT got here from the creating world, particularly Africa, Asia, and South America, we additionally cleaned and consolidated knowledge for Europe, Oceania, and North America from present world databases and authorities sources. Information for Oceania principally got here from AQUASTAT and GRanD, apart from Australia, which was collected from the Register of Giant Dams maintained by the Australian Nationwide Committee on Giant Dams. The GDAT database for Europe comprises info from AQUASTAT, GRanD, the WRI International Energy Plant Database, and stories for greater than 3,000 dams in Europe. For North America, knowledge have been compiled from AQUASTAT and GRanD, apart from america, which included knowledge from america Geological Survey (USGS). USGS lists greater than 8,000 dams, dikes, levees, and different water-engineering constructions. We filtered the USGS knowledge to take away non-dam constructions.
Description of dam attributes
Along with better spatial protection, the GDAT database comprises detailed attribute info for every dam. These attributes embrace primary figuring out info such because the dam title and alternate names; geographic attributes equivalent to nation, province, river basin, latitude, and longitude; dam traits equivalent to completion and building years, peak, and size; reservoir capability and space; functions; and energy-generating capability for hydroelectric dams. These attributes served as the premise for knowledge assortment, and any info present in regional and country-specific sources was matched to those attributes to be included in GDAT in a constant and standardized format.
Dams are largely concentrated alongside main river basins in Asia, Africa, Europe, and the Americas (Fig. 1 and Desk 1). Asia has the best variety of dams accomplished thus far, with 9,526 dams, or 27 % of worldwide dam building. North America and South America even have vital dam infrastructure, representing 23 % and 21 % of the worldwide dam depend, respectively (Desk 1). By way of put in capability, Asia and South America account for 50 % and 20 % of the worldwide complete put in capability, respectively, whereas Europe and North America account for 18 % and 9 %, respectively (Desk 1).
Map of dams within the International Dam Tracker database. (a) Places of all dams within the GDAT database, with every dam represented by a blue level. Giant concentrations of dams may be present in america, Brazil, India, South Africa, Europe, and East and Southeast Asia. (b) The catchment space of dams delineated utilizing drainage movement instructions.
There are notable variations within the distribution of dam completion years throughout continents. Whereas most developed international locations in North America, Europe, and Oceania have been witnessing a decline in dam building for the reason that Seventies, creating international locations in Africa, Asia, and South America have been experiencing a continued improve in dam building (Fig. 2). Most notably, the Yangtze basin in China, the Ganges-Brahmaputra basin in South Asia, and the Amazon basin in South America have many dams presently deliberate or underneath building4. A time collection of completion 12 months (Fig. 3) demonstrates a major acceleration in dam building from 1970 to 2000, adopted by a slight deceleration over the previous decade. International investments in hydropower peaked within the second half of the 20 th century in response to a rising want to diversify vitality sources and cut back dependency on fossil fuels5.
Distribution of dam completion 12 months by continent.
Cumulative distribution of dam completion 12 months by continent.
The commonest essential functions of dams in GDAT are irrigation and hydroelectricity, which characterize 25 % and 20 % of the information (Fig. 4). Extra essential functions embrace water provide, flood management, recreation, livestock, and navigation. 32 % of dams had the principle objective labeled as “different” because of restricted or lacking info within the major sources.
Distribution of essential objective of dams in GDAT – Total.
In Asia and South America, irrigation is the most typical essential objective of dams (Fig. 5). As compared, the most typical essential functions in North America, Europe, and Oceania are flood management, hydroelectricity, and water provide/storage, respectively. The variety of dams proven underneath every essential objective displays the first and secondary sources used within the knowledge assortment. For instance, Europe’s excessive share of dams used for hydroelectricity is partly as a result of excessive focus of hydroelectric energy crops within the WRI knowledge. Extra sources will must be consulted to offer a extra full illustration of the dam essential functions by area, significantly for Europe and Oceania.
Distribution of essential objective of dams in GDAT – By continent.
GDAT Use Case: Calculating Dam-Induced International Floor Water Adjustments
To reveal how GDAT would enhance our understanding of the worldwide impression of dam building, we use GDAT to calculate dam-induced world floor water modifications over the previous three a long time. Following prior work22,27, we apply an algorithm (Fig. 6a) that makes use of GDAT and varied state-of-the-art satellite tv for pc knowledge merchandise to research temporal floor water dynamics round dam catchment areas. The algorithm has 5 essential steps.
Conceptual diagram of the algorithm. We develop an algorithm to estimate the modifications in world floor water from dam building. (a) To calculate the distinction in floor water protection earlier than and after dams are constructed, we snap every dam to a river community to right for its location, calculate the drainage route, and delineate its catchment space. (b) Procedures for imputing the completion years of dams when the knowledge is lacking. This enables the evaluation of surface-water modifications to be expanded past dams with recognized completion years.
First, to make sure that dams fall exactly on a river community, we snap all 31,780 geocoded dams in GDAT onto the closest river strains utilizing a fine-scale world river community dataset at 15 arc-second decision, HydroRivers. HydroRivers supplies vectorized line community of all world rivers which have a catchment space of no less than 10 km2 or a mean river movement of 0.1 cubic meters per second, or each. It encompasses 8.5 million particular person river reaches with a mean size of 4.2 km, representing 35.9 million km of rivers globally27. This step ensures the dam coordinates are precisely aligned with the river community.
Second, we use the USGS International Multi-Decision Terrain Elevation 2010 knowledge (GMTED 2010) at 7.5 arc-second decision to generate the drainage movement route at every dam web site. GMTED 2010, developed by the USGS and the Nationwide Geospatial-Intelligence Company (NGA), is an enhanced world elevation knowledge of alternative that comes with the very best accessible knowledge sources28. The GMTED 2010 knowledge characterize a major enchancment in consistency and vertical accuracy over present world elevation knowledge, equivalent to GTOPO30.
Third, with the PCRaster library in Python, we delineate the catchment space representing the buffer zone upstream of the dam location containing its reservoir. The PCRaster code assigns a movement route worth to every pixel within the GMTED elevation knowledge, which it does by analyzing the values of the eight surrounding pixels. The code then retraces the movement of water ranging from every dam by transferring upstream till there are not any extra upstream pixels or till the algorithm hits one other dam29. The world coated by the retraced pixels is the catchment space for the dam, which is the world we use to calculate floor water modifications attributable to the dam. In complete, we get hold of 28,460 catchment areas (Fig. 1b) akin to dams in GDAT. A small subset of GDAT dams doesn’t have catchment areas because of limitations within the elevation and river community knowledge.
Calculating floor water change requires info on the dam completion 12 months. Since greater than 70% of GDAT dams have recognized completion years, we use these as coaching knowledge for an imputation process on dams with lacking info (Fig. 6b). Across the 12 months of dam completion, we anticipate to see a major change within the upstream catchment space, equivalent to in Fig. 7. Due to this fact, because the fourth step, we impute the completion 12 months for 8,342 dams by detecting development breaks within the catchment space time collection. The satellite-derived floor water protection between 1984 and 2019 comes from the International Floor Water (GSW) V1.1 dataset accessible by way of Google Earth Engine. Developed by the European Fee Joint Analysis Centre, GSW is a globally constant, validated dataset that exhibits the spatial and temporal distribution of worldwide floor water over the previous three a long time utilizing three million Landsat satellite tv for pc photos30.
Change in floor water protection for 3 exemplary dams. Dam completion 12 months in parenthesis. (a) Development for Ataturk Dam in Turkey was accomplished in 1990 and the reservoir was stuffed in 1992. A comparability of the pre-dam (left) and post-dam (proper) photos present a transparent improve in water pixels after the completion, a complete improve of 983,482 km2. The pre-dam picture displays {a partially} stuffed reservoir, demonstrating that giant dams could take an prolonged time to fill. (b) Luis Eduardo Magalhaes Dam in Brazil was accomplished in 2001. The previously free-flowing part of the river is seen within the pre-dam picture, whereas the dam flooded an space of 660,130.5 km2. (c) Mohale Dam in Lesotho, with an imputed completion 12 months of 2004, illustrates the logic behind our imputation process. The precise 12 months of building is 2002, and the dam was formally commissioned in 2003–4. After completion, the dam flooded an space of 15,884.5 km2.
Within the fifth and remaining step, we calculate world floor water modifications induced by dam building. From GSW, we get hold of pixel counts for the “no water,” “seasonal water,” and “everlasting water” classes as annual time collection. We then use the completion 12 months info to calculate the median pre- and post-dam pixel counts. We select the median because the abstract measure to reduce idiosyncratic measurement errors from climatic or technical elements which will have an effect on satellite tv for pc picture high quality. We quantify the temporal floor water dynamics for 11,710 dams accomplished between 1985 and 2019. We drop dams with out pre- or post-completion statement, i.e., these constructed earlier than 1984 or after 2019. Beneath, we offer extra particulars in regards to the methodology.
Dam catchment areas
We mix GDAT with elevation knowledge to calculate the catchment space that comprises the reservoir of every dam. To hint the drainage movement route, we use GMTED 2010 with a decision of 250 meters. The elevation knowledge permits us to calculate how water would movement between every pixel based mostly on their relative upstream or downstream location, just like GOODD22. We first make use of PCRaster, an environmental modeling software program, to acquire catchment areas for every dam from elevation knowledge29. We use the native drainage route perform (ldd) to assign movement instructions to every pixel based mostly on the bottom elevation of the eight surrounding pixels. This enables us to acquire a flow-direction file from elevation knowledge. We then use the subcatchment perform, which makes use of every dam in GDAT as a place to begin and traces all doable trajectories that water may movement towards the dam. If there’s one other dam alongside the trail of the water trajectory, the perform stops, as all additional upstream factors would fall inside one other dam’s catchment space. We’re therefore in a position to seize all upstream factors that finally drain to the purpose of every dam, enabling us to acquire the catchment areas for almost all of dams in our database.
Information on movement route comprises noise from the slopes surrounding the river, as water would possibly movement in a unique route than the river itself22. The pixels the place the best movement accumulation happen (the trail tracing the underside of the riverbed) are the one places that precisely replicate the river’s movement. In consequence, dams which might be even barely misplaced relative to the principle river path may end in giant errors within the catchment space. To deal with this, we align dam places with river community knowledge from HydroRIVERS to make sure that dams fall exactly on the principle river path the place movement accumulation happens. HydroRIVERS captures all pixels with an amassed upstream catchment of greater than 10 km2, producing a linked community of strains representing amassed river flows31. We use the interpolate perform from the Shapely library in Python to snap the situation of every dam onto its nearest river community32.
The overwhelming majority of dams within the GDAT database are situated very near a HydroRIVERS river community line, with 30,471 dams (or 95.9 %) falling inside 0.5 levels of their snapped location. The remaining 1,309 dams have been snapped at a distance of greater than 0.5 levels. A few of them have been situated on islands not coated by HydroRIVERS, inflicting them to be snapped onto a continent. Therefore, dams with an interpolated distance of better than 0.5 levels could also be inaccurate. This is because of limitations within the knowledge decision, as 0.5 levels is the decision of the runoff and discharge layers used to calculate HydroSHEDS33. In these instances, we discard the snapped location and as an alternative use the unique geo-coordinates. Our remaining catchment-area calculation process, subsequently, consists of three phases: acquiring drainage route from elevation knowledge through the ldd perform from PCRaster, snapping every dam onto its respective river community through Shapely, and utilizing the drainage route and interpolated location to derive a catchment space for every dam. The output is a raster file of dam catchment areas, that are used to know how every dam contributed to modifications in surface-water protection.
Calculating floor water modifications inside dam catchment areas
We calculate the change in floor water protection inside every catchment earlier than and after dam completion. This enables us to know how the dam has inundated the encompassing land. We use knowledge from the International Floor Water (GSW) Explorer developed by the European Fee Joint Analysis Centre, which comprises floor water protection knowledge between 1984 and 201930. The dataset divides the world into pixels coded as 1 for no protection, 2 for seasonal protection, and three for annual protection. We masks all pixels which might be coded as 0 (no knowledge). Not all years comprise strong knowledge because of cloud cowl and satellite-imaging anomalies. Due to this fact, for every dam, we get hold of the median worth of all pre-dam pixels inside its catchment space and create a single snapshot of the median pre-dam water protection. We do the identical for the post-dam pixels (inclusive of the 12 months of completion), acquiring a single post-dam snapshot of the median water protection. Because of the masking of no-data pixels, some pixels could have fewer years of information. The doable median-value pixels embrace no water (1), various values of seasonal water (1.5, 2, 2.5), and everlasting water (3). As soon as we’ve the median-value snapshots of pre-dam and post-dam water protection throughout the buffer zone of every dam, we calculate the distinction within the quantity of space coated by each seasonal and everlasting pixels. On condition that GSW knowledge spans 1984 and 2019, pre- and post-dam comparisons are solely doable for dams with catchment knowledge and completion years that fall throughout the GSW time vary. This yields us 5,508 dams for which we are able to calculate water-coverage modifications.
Imputation of dam completion years
Calculating floor water modifications utilizing dams with accessible completion-year knowledge limits us to solely 5,508 dams out of greater than 30,000 in GDAT. To develop our evaluation to extra dams, we developed a machine-driven imputation process that calculates the dam’s completion 12 months based mostly on surface-water knowledge from GSW. As a result of the imputation process requires deriving completion years from surface-water modifications, we are able to solely apply it to dams with catchment knowledge however with out recorded completion years. This totals 8,357 observations. Utilizing GSW knowledge, we get hold of the annual pixel counts throughout the catchment space of every dam for every year between 1984 and 2019. The results of this step is a knowledge body of 8,357 dams spanning 36 years, with counts of no-water (1), seasonal-water (2), and permanent-water (3) pixels for every year. Years that comprise lacking knowledge are stuffed in with knowledge from the earlier 12 months or the newest 12 months with knowledge.
We use two strategies to calculate the imputed completion 12 months. The primary methodology identifies the 12 months with the biggest pixel change because the imputed completion 12 months. We conduct 4 runs for this methodology: utilizing 1 pixels solely (no water), 2 pixels solely (seasonal water), 3 pixels solely (everlasting water), and a couple of + 3 pixels (any water). For every run, we get hold of the pixel distinction between the present and prior years for every year throughout the GSW knowledge, which yields 35 observations of year-to-year variations. This yields us a complete of 4 imputation outcomes utilizing the largest-change methodology. The second methodology includes the calculation of the structural break utilizing the Wald take a look at. A structural break happens when the development of a time collection abruptly modifications34,35,36. We apply the STATA estat sbsingle take a look at, which assumes a single structural break with an unknown break date within the knowledge37,38,39,40. That is applicable for estimating the completion 12 months of a dam when the 12 months is unknown, on condition that the structural break within the water-coverage development would happen when the dam is accomplished. As with the primary methodology, we apply the structural-break take a look at with 4 runs: 1 pixel solely (no water), 2 pixels solely (seasonal water), 3 pixels solely (everlasting water), and a couple of + 3 pixels (any water). This yields 4 extra imputation outcomes for every dam. The above two strategies mixed yield a complete of eight imputation outcomes. Essentially the most frequent (mode) 12 months among the many eight imputations is the ultimate calculation for the imputed 12 months. If there are a number of modes, then the earliest mode is taken. Taking earlier years would account for the truth that it takes time for dams to refill after completion.
Use case outcomes: dam-induced world floor water modifications
Dams have considerably altered the situation and persistence of floor water world wide. Determine 8a exhibits a cumulative improve of no less than 49,715 km2 of seasonal and everlasting water in dam catchment areas between 1984 and 2018. That is probably a lower-bound estimate as a result of absence of satellite tv for pc knowledge earlier than 1984 and the dearth of completion 12 months info for all dams. The quickest dam-induced floor water growth has occurred in Asia (Fig. 8b), significantly from the fast tempo of dam building in China and India. Development is slower however nonetheless appreciable in South America and Africa, the place hydroelectric dams energy financial improvement in Brazil, Venezuela, Zambia, and D.R. Congo. North America witnessed numerous dams firstly of the pattern interval, however few new dams have been accomplished after 1984 (Figs. 2,3). Just like North America, dam building peaked in Europe across the Nineteen Sixties, and floor water protection inside dam catchment areas has remained regular since 1984.
Abstract of modifications in water protection round world dam catchments between 1984 and 2018. (a) International cumulative improve in water-covered space (seasonal + everlasting) inside dam catchments. (b) Cumulative improve by continent. (c) Cumulative improve by pixel class for pre- and post-dam intervals, displaying a rise in permanent- and seasonal-water pixels by round 50,000 km2 and a lower in no-water pixels. (d) Dam-induced floor water modifications by continent.
Evaluating the change in seasonal and everlasting water earlier than (pre-dam) and after (post-dam) dam completion (Fig. 8c,d) provides us perception into the impression of dam building on floor water dynamics. In Asia, post-dam catchment areas witnessed the biggest floor water improve by greater than 23,400 km2 (Desk 2). That is adopted by South America and Africa, with greater than 7,000 km2 of extra floor water because of dam building. We additionally observe a rise within the complete pixel depend globally and a pronounced improve in Asia because the spatial and temporal resolutions of satellite tv for pc knowledge there have improved considerably over the previous twenty years. It is a knowledge high quality limitation for quantifying the long-term floor water dynamics utilizing satellite tv for pc knowledge–a caveat documented within the GSW dataset30.
We map modifications in world dam catchments by nation to know the spatial distribution of floor water dynamics (Fig. 9). The continental summaries masks appreciable spatial heterogeneity throughout international locations. The biggest contributors to dam-induced floor water change come from creating international locations with fast hydropower set up, equivalent to China and Brazil. China alone accounts for greater than half of the entire floor water improve in Asia. In Brazil, 3,459 dams contribute to greater than 80% of the water growth in South America. Developed international locations with traditionally giant inventories of dams, equivalent to Canada and america, additionally see a major rise in floor water. The growth in floor water, nevertheless, is probably not a common characteristic of dams. In a couple of international locations, most notably Zambia, Eire, and the United Arab Emirates, floor water could have decreased barely after dams are constructed. We can not rule out the likelihood that the decline is because of idiosyncratic modifications in satellite tv for pc knowledge high quality or measurement error, as we’ve discovered the floor water loss to be lower than 5 km2 in these three international locations. Echoing prior findings30, these outcomes recommend that losses in everlasting and seasonal water from dams could also be extra concentrated than positive factors.
Floor water change in world dam catchments between 1984 and 2018 by nation. The biggest improve in floor water protection are in developed and quickly creating international locations with vital dam counts, equivalent to Brazil, Canada, China, India, america, Mexico, Turkey, and varied Southeast Asian international locations. The information captures modifications within the median floor water pixel counts between pre- and post-construction intervals inside dam catchment areas. Along with dam building, a portion of the floor water modifications in some international locations, particularly these in Asia, may very well be attributed to enhancements in satellite tv for pc knowledge high quality.
On this part, we current an outline of the information assortment course of and reveal an utility the place GDAT is used to calculate dam-induced world floor water modifications. The primary subsection, “International Dam Tracker (GDAT),” presents the GDAT database and describes the information sources. The second subsection,“GDAT Use Case,” particulars our utility of GDAT and descriptions an algorithm for calculating floor water modifications inside dam catchment areas.
International dam tracker (GDAT)
The International Dam Tracker (GDAT), a database with 35,140 dams in all continents besides Antarctica, is an authentic knowledge compilation effort24. We construct upon present world and regional dam databases, together with AQUASTAT (www.fao.org/aquastat), the International Reservoir and Dam Database (GRanD – globaldamwatch.org/grand)25, and the World Assets Institute (WRI) database (datasets.wri.org/dataset/globalpowerplantdatabase)26. For every nation, we consolidate these world and area databases with accessible knowledge from authorities businesses, lecturers, and nonprofits. Greater than 90% of dams in GDAT are geocoded (n = 31,780), and the coordinates are extensively verified utilizing Google Earth and different geospatial software program. This course of creates an in depth cross-validated database that can be utilized for evaluation at varied geographical scales.
Description of regional knowledge sources
This part supplies an outline of main regional or nationwide knowledge sources that contributed considerably to GDAT. Detailed info on the information sources for every nation is offered within the Area Highlights part. Current dam databases have restricted protection of creating and low-income international locations. As such, our knowledge assortment effort focuses on international locations in Africa, Asia, and South America, the place we collected major knowledge from administrative and different sources. Information for international locations in Europe, North America, and Oceania are primarily obtained from present world dam databases.
Dams in India have been collected from the India Water Assets Data System (WRIS), a database that’s maintained and funded by the federal government of India (indiawris.gov.in/wris). Reservoir quantity knowledge are usually not offered by the WRIS and, therefore, are usually not presently accessible in GDAT for dams in India.
Dams in China have been compiled from a number of nongovernmental sources: AQUASTAT, GRanD, Worldwide Rivers, and the Consultative Group for Worldwide Agricultural Analysis (CGIAR). Whereas the Ministry of Water Assets of the Folks’s Republic of China maintains a database on dams in China, it isn’t launched to the general public. Therefore, regardless that the GDAT database has compiled greater than 1,000 dams in China, the precise dam depend is anticipated to be a lot larger, on condition that many smaller dams could be lacking from the database.
For Continental Southeast Asia, GDAT database builds upon three essential knowledge sources: AQUASTAT, GRanD, and CGIAR Larger Mekong (wle-mekong.cgiar.org). The CGIAR database focuses on the key river basins of continental Southeast Asia (Mekong, Purple, Salween, Irrawaddy) and comprises info for greater than 800 dams that span China, Thailand, Vietnam, Laos, Cambodia, and Myanmar.
Dams within the Center East got here from AQUASTAT and GRanD, besides Turkey and Iran, the place we obtained knowledge from authorities web sites. The Turkish Common Directorate of State Hydraulic Works (www.dsi.gov.tr)publishes info for greater than 600 dams, and the Iran Water Assets Administration Firm (www.wrm.ir) publishes info for greater than 1,000. Neither database, nevertheless, is geolocated. Therefore, the geographic coordinates for these two international locations are manually verified by way of GeoNames and Google searches.
Dams in Brazil have been compiled from 4 sources: AQUASTAT, GRanD, the Brazilian Nationwide Dam Security Data System (SNISB), and Dams in Amazonia (archive.internationalrivers.org). The SNISB for 2016, the newest model accessible, contains virtually 23,000 dams. We excluded all dams and reservoirs used completely for tailings and unsafe waste storage. The SNISB contains constructions with a peak better than or equal to fifteen meters and a complete reservoir capability better than 3hm3. In consequence, the GDAT database is probably going lacking a substantial variety of small dams in Brazil that aren’t monitored by the SNISB.
Whereas a lot of the new knowledge in GDAT got here from the creating world, particularly Africa, Asia, and South America, we additionally cleaned and consolidated knowledge for Europe, Oceania, and North America from present world databases and authorities sources. Information for Oceania principally got here from AQUASTAT and GRanD, apart from Australia, which was collected from the Register of Giant Dams maintained by the Australian Nationwide Committee on Giant Dams. The GDAT database for Europe comprises info from AQUASTAT, GRanD, the WRI International Energy Plant Database, and stories for greater than 3,000 dams in Europe. For North America, knowledge have been compiled from AQUASTAT and GRanD, apart from america, which included knowledge from america Geological Survey (USGS). USGS lists greater than 8,000 dams, dikes, levees, and different water-engineering constructions. We filtered the USGS knowledge to take away non-dam constructions.
Description of dam attributes
Along with better spatial protection, the GDAT database comprises detailed attribute info for every dam. These attributes embrace primary figuring out info such because the dam title and alternate names; geographic attributes equivalent to nation, province, river basin, latitude, and longitude; dam traits equivalent to completion and building years, peak, and size; reservoir capability and space; functions; and energy-generating capability for hydroelectric dams. These attributes served as the premise for knowledge assortment, and any info present in regional and country-specific sources was matched to those attributes to be included in GDAT in a constant and standardized format.
Dams are largely concentrated alongside main river basins in Asia, Africa, Europe, and the Americas (Fig. 1 and Desk 1). Asia has the best variety of dams accomplished thus far, with 9,526 dams, or 27 % of worldwide dam building. North America and South America even have vital dam infrastructure, representing 23 % and 21 % of the worldwide dam depend, respectively (Desk 1). By way of put in capability, Asia and South America account for 50 % and 20 % of the worldwide complete put in capability, respectively, whereas Europe and North America account for 18 % and 9 %, respectively (Desk 1).
Map of dams within the International Dam Tracker database. (a) Places of all dams within the GDAT database, with every dam represented by a blue level. Giant concentrations of dams may be present in america, Brazil, India, South Africa, Europe, and East and Southeast Asia. (b) The catchment space of dams delineated utilizing drainage movement instructions.
There are notable variations within the distribution of dam completion years throughout continents. Whereas most developed international locations in North America, Europe, and Oceania have been witnessing a decline in dam building for the reason that Seventies, creating international locations in Africa, Asia, and South America have been experiencing a continued improve in dam building (Fig. 2). Most notably, the Yangtze basin in China, the Ganges-Brahmaputra basin in South Asia, and the Amazon basin in South America have many dams presently deliberate or underneath building4. A time collection of completion 12 months (Fig. 3) demonstrates a major acceleration in dam building from 1970 to 2000, adopted by a slight deceleration over the previous decade. International investments in hydropower peaked within the second half of the 20 th century in response to a rising want to diversify vitality sources and cut back dependency on fossil fuels5.
Distribution of dam completion 12 months by continent.
Cumulative distribution of dam completion 12 months by continent.
The commonest essential functions of dams in GDAT are irrigation and hydroelectricity, which characterize 25 % and 20 % of the information (Fig. 4). Extra essential functions embrace water provide, flood management, recreation, livestock, and navigation. 32 % of dams had the principle objective labeled as “different” because of restricted or lacking info within the major sources.
Distribution of essential objective of dams in GDAT – Total.
In Asia and South America, irrigation is the most typical essential objective of dams (Fig. 5). As compared, the most typical essential functions in North America, Europe, and Oceania are flood management, hydroelectricity, and water provide/storage, respectively. The variety of dams proven underneath every essential objective displays the first and secondary sources used within the knowledge assortment. For instance, Europe’s excessive share of dams used for hydroelectricity is partly as a result of excessive focus of hydroelectric energy crops within the WRI knowledge. Extra sources will must be consulted to offer a extra full illustration of the dam essential functions by area, significantly for Europe and Oceania.
Distribution of essential objective of dams in GDAT – By continent.
GDAT Use Case: Calculating Dam-Induced International Floor Water Adjustments
To reveal how GDAT would enhance our understanding of the worldwide impression of dam building, we use GDAT to calculate dam-induced world floor water modifications over the previous three a long time. Following prior work22,27, we apply an algorithm (Fig. 6a) that makes use of GDAT and varied state-of-the-art satellite tv for pc knowledge merchandise to research temporal floor water dynamics round dam catchment areas. The algorithm has 5 essential steps.
Conceptual diagram of the algorithm. We develop an algorithm to estimate the modifications in world floor water from dam building. (a) To calculate the distinction in floor water protection earlier than and after dams are constructed, we snap every dam to a river community to right for its location, calculate the drainage route, and delineate its catchment space. (b) Procedures for imputing the completion years of dams when the knowledge is lacking. This enables the evaluation of surface-water modifications to be expanded past dams with recognized completion years.
First, to make sure that dams fall exactly on a river community, we snap all 31,780 geocoded dams in GDAT onto the closest river strains utilizing a fine-scale world river community dataset at 15 arc-second decision, HydroRivers. HydroRivers supplies vectorized line community of all world rivers which have a catchment space of no less than 10 km2 or a mean river movement of 0.1 cubic meters per second, or each. It encompasses 8.5 million particular person river reaches with a mean size of 4.2 km, representing 35.9 million km of rivers globally27. This step ensures the dam coordinates are precisely aligned with the river community.
Second, we use the USGS International Multi-Decision Terrain Elevation 2010 knowledge (GMTED 2010) at 7.5 arc-second decision to generate the drainage movement route at every dam web site. GMTED 2010, developed by the USGS and the Nationwide Geospatial-Intelligence Company (NGA), is an enhanced world elevation knowledge of alternative that comes with the very best accessible knowledge sources28. The GMTED 2010 knowledge characterize a major enchancment in consistency and vertical accuracy over present world elevation knowledge, equivalent to GTOPO30.
Third, with the PCRaster library in Python, we delineate the catchment space representing the buffer zone upstream of the dam location containing its reservoir. The PCRaster code assigns a movement route worth to every pixel within the GMTED elevation knowledge, which it does by analyzing the values of the eight surrounding pixels. The code then retraces the movement of water ranging from every dam by transferring upstream till there are not any extra upstream pixels or till the algorithm hits one other dam29. The world coated by the retraced pixels is the catchment space for the dam, which is the world we use to calculate floor water modifications attributable to the dam. In complete, we get hold of 28,460 catchment areas (Fig. 1b) akin to dams in GDAT. A small subset of GDAT dams doesn’t have catchment areas because of limitations within the elevation and river community knowledge.
Calculating floor water change requires info on the dam completion 12 months. Since greater than 70% of GDAT dams have recognized completion years, we use these as coaching knowledge for an imputation process on dams with lacking info (Fig. 6b). Across the 12 months of dam completion, we anticipate to see a major change within the upstream catchment space, equivalent to in Fig. 7. Due to this fact, because the fourth step, we impute the completion 12 months for 8,342 dams by detecting development breaks within the catchment space time collection. The satellite-derived floor water protection between 1984 and 2019 comes from the International Floor Water (GSW) V1.1 dataset accessible by way of Google Earth Engine. Developed by the European Fee Joint Analysis Centre, GSW is a globally constant, validated dataset that exhibits the spatial and temporal distribution of worldwide floor water over the previous three a long time utilizing three million Landsat satellite tv for pc photos30.
Change in floor water protection for 3 exemplary dams. Dam completion 12 months in parenthesis. (a) Development for Ataturk Dam in Turkey was accomplished in 1990 and the reservoir was stuffed in 1992. A comparability of the pre-dam (left) and post-dam (proper) photos present a transparent improve in water pixels after the completion, a complete improve of 983,482 km2. The pre-dam picture displays {a partially} stuffed reservoir, demonstrating that giant dams could take an prolonged time to fill. (b) Luis Eduardo Magalhaes Dam in Brazil was accomplished in 2001. The previously free-flowing part of the river is seen within the pre-dam picture, whereas the dam flooded an space of 660,130.5 km2. (c) Mohale Dam in Lesotho, with an imputed completion 12 months of 2004, illustrates the logic behind our imputation process. The precise 12 months of building is 2002, and the dam was formally commissioned in 2003–4. After completion, the dam flooded an space of 15,884.5 km2.
Within the fifth and remaining step, we calculate world floor water modifications induced by dam building. From GSW, we get hold of pixel counts for the “no water,” “seasonal water,” and “everlasting water” classes as annual time collection. We then use the completion 12 months info to calculate the median pre- and post-dam pixel counts. We select the median because the abstract measure to reduce idiosyncratic measurement errors from climatic or technical elements which will have an effect on satellite tv for pc picture high quality. We quantify the temporal floor water dynamics for 11,710 dams accomplished between 1985 and 2019. We drop dams with out pre- or post-completion statement, i.e., these constructed earlier than 1984 or after 2019. Beneath, we offer extra particulars in regards to the methodology.
Dam catchment areas
We mix GDAT with elevation knowledge to calculate the catchment space that comprises the reservoir of every dam. To hint the drainage movement route, we use GMTED 2010 with a decision of 250 meters. The elevation knowledge permits us to calculate how water would movement between every pixel based mostly on their relative upstream or downstream location, just like GOODD22. We first make use of PCRaster, an environmental modeling software program, to acquire catchment areas for every dam from elevation knowledge29. We use the native drainage route perform (ldd) to assign movement instructions to every pixel based mostly on the bottom elevation of the eight surrounding pixels. This enables us to acquire a flow-direction file from elevation knowledge. We then use the subcatchment perform, which makes use of every dam in GDAT as a place to begin and traces all doable trajectories that water may movement towards the dam. If there’s one other dam alongside the trail of the water trajectory, the perform stops, as all additional upstream factors would fall inside one other dam’s catchment space. We’re therefore in a position to seize all upstream factors that finally drain to the purpose of every dam, enabling us to acquire the catchment areas for almost all of dams in our database.
Information on movement route comprises noise from the slopes surrounding the river, as water would possibly movement in a unique route than the river itself22. The pixels the place the best movement accumulation happen (the trail tracing the underside of the riverbed) are the one places that precisely replicate the river’s movement. In consequence, dams which might be even barely misplaced relative to the principle river path may end in giant errors within the catchment space. To deal with this, we align dam places with river community knowledge from HydroRIVERS to make sure that dams fall exactly on the principle river path the place movement accumulation happens. HydroRIVERS captures all pixels with an amassed upstream catchment of greater than 10 km2, producing a linked community of strains representing amassed river flows31. We use the interpolate perform from the Shapely library in Python to snap the situation of every dam onto its nearest river community32.
The overwhelming majority of dams within the GDAT database are situated very near a HydroRIVERS river community line, with 30,471 dams (or 95.9 %) falling inside 0.5 levels of their snapped location. The remaining 1,309 dams have been snapped at a distance of greater than 0.5 levels. A few of them have been situated on islands not coated by HydroRIVERS, inflicting them to be snapped onto a continent. Therefore, dams with an interpolated distance of better than 0.5 levels could also be inaccurate. This is because of limitations within the knowledge decision, as 0.5 levels is the decision of the runoff and discharge layers used to calculate HydroSHEDS33. In these instances, we discard the snapped location and as an alternative use the unique geo-coordinates. Our remaining catchment-area calculation process, subsequently, consists of three phases: acquiring drainage route from elevation knowledge through the ldd perform from PCRaster, snapping every dam onto its respective river community through Shapely, and utilizing the drainage route and interpolated location to derive a catchment space for every dam. The output is a raster file of dam catchment areas, that are used to know how every dam contributed to modifications in surface-water protection.
Calculating floor water modifications inside dam catchment areas
We calculate the change in floor water protection inside every catchment earlier than and after dam completion. This enables us to know how the dam has inundated the encompassing land. We use knowledge from the International Floor Water (GSW) Explorer developed by the European Fee Joint Analysis Centre, which comprises floor water protection knowledge between 1984 and 201930. The dataset divides the world into pixels coded as 1 for no protection, 2 for seasonal protection, and three for annual protection. We masks all pixels which might be coded as 0 (no knowledge). Not all years comprise strong knowledge because of cloud cowl and satellite-imaging anomalies. Due to this fact, for every dam, we get hold of the median worth of all pre-dam pixels inside its catchment space and create a single snapshot of the median pre-dam water protection. We do the identical for the post-dam pixels (inclusive of the 12 months of completion), acquiring a single post-dam snapshot of the median water protection. Because of the masking of no-data pixels, some pixels could have fewer years of information. The doable median-value pixels embrace no water (1), various values of seasonal water (1.5, 2, 2.5), and everlasting water (3). As soon as we’ve the median-value snapshots of pre-dam and post-dam water protection throughout the buffer zone of every dam, we calculate the distinction within the quantity of space coated by each seasonal and everlasting pixels. On condition that GSW knowledge spans 1984 and 2019, pre- and post-dam comparisons are solely doable for dams with catchment knowledge and completion years that fall throughout the GSW time vary. This yields us 5,508 dams for which we are able to calculate water-coverage modifications.
Imputation of dam completion years
Calculating floor water modifications utilizing dams with accessible completion-year knowledge limits us to solely 5,508 dams out of greater than 30,000 in GDAT. To develop our evaluation to extra dams, we developed a machine-driven imputation process that calculates the dam’s completion 12 months based mostly on surface-water knowledge from GSW. As a result of the imputation process requires deriving completion years from surface-water modifications, we are able to solely apply it to dams with catchment knowledge however with out recorded completion years. This totals 8,357 observations. Utilizing GSW knowledge, we get hold of the annual pixel counts throughout the catchment space of every dam for every year between 1984 and 2019. The results of this step is a knowledge body of 8,357 dams spanning 36 years, with counts of no-water (1), seasonal-water (2), and permanent-water (3) pixels for every year. Years that comprise lacking knowledge are stuffed in with knowledge from the earlier 12 months or the newest 12 months with knowledge.
We use two strategies to calculate the imputed completion 12 months. The primary methodology identifies the 12 months with the biggest pixel change because the imputed completion 12 months. We conduct 4 runs for this methodology: utilizing 1 pixels solely (no water), 2 pixels solely (seasonal water), 3 pixels solely (everlasting water), and a couple of + 3 pixels (any water). For every run, we get hold of the pixel distinction between the present and prior years for every year throughout the GSW knowledge, which yields 35 observations of year-to-year variations. This yields us a complete of 4 imputation outcomes utilizing the largest-change methodology. The second methodology includes the calculation of the structural break utilizing the Wald take a look at. A structural break happens when the development of a time collection abruptly modifications34,35,36. We apply the STATA estat sbsingle take a look at, which assumes a single structural break with an unknown break date within the knowledge37,38,39,40. That is applicable for estimating the completion 12 months of a dam when the 12 months is unknown, on condition that the structural break within the water-coverage development would happen when the dam is accomplished. As with the primary methodology, we apply the structural-break take a look at with 4 runs: 1 pixel solely (no water), 2 pixels solely (seasonal water), 3 pixels solely (everlasting water), and a couple of + 3 pixels (any water). This yields 4 extra imputation outcomes for every dam. The above two strategies mixed yield a complete of eight imputation outcomes. Essentially the most frequent (mode) 12 months among the many eight imputations is the ultimate calculation for the imputed 12 months. If there are a number of modes, then the earliest mode is taken. Taking earlier years would account for the truth that it takes time for dams to refill after completion.
Use case outcomes: dam-induced world floor water modifications
Dams have considerably altered the situation and persistence of floor water world wide. Determine 8a exhibits a cumulative improve of no less than 49,715 km2 of seasonal and everlasting water in dam catchment areas between 1984 and 2018. That is probably a lower-bound estimate as a result of absence of satellite tv for pc knowledge earlier than 1984 and the dearth of completion 12 months info for all dams. The quickest dam-induced floor water growth has occurred in Asia (Fig. 8b), significantly from the fast tempo of dam building in China and India. Development is slower however nonetheless appreciable in South America and Africa, the place hydroelectric dams energy financial improvement in Brazil, Venezuela, Zambia, and D.R. Congo. North America witnessed numerous dams firstly of the pattern interval, however few new dams have been accomplished after 1984 (Figs. 2,3). Just like North America, dam building peaked in Europe across the Nineteen Sixties, and floor water protection inside dam catchment areas has remained regular since 1984.
Abstract of modifications in water protection round world dam catchments between 1984 and 2018. (a) International cumulative improve in water-covered space (seasonal + everlasting) inside dam catchments. (b) Cumulative improve by continent. (c) Cumulative improve by pixel class for pre- and post-dam intervals, displaying a rise in permanent- and seasonal-water pixels by round 50,000 km2 and a lower in no-water pixels. (d) Dam-induced floor water modifications by continent.
Evaluating the change in seasonal and everlasting water earlier than (pre-dam) and after (post-dam) dam completion (Fig. 8c,d) provides us perception into the impression of dam building on floor water dynamics. In Asia, post-dam catchment areas witnessed the biggest floor water improve by greater than 23,400 km2 (Desk 2). That is adopted by South America and Africa, with greater than 7,000 km2 of extra floor water because of dam building. We additionally observe a rise within the complete pixel depend globally and a pronounced improve in Asia because the spatial and temporal resolutions of satellite tv for pc knowledge there have improved considerably over the previous twenty years. It is a knowledge high quality limitation for quantifying the long-term floor water dynamics utilizing satellite tv for pc knowledge–a caveat documented within the GSW dataset30.
We map modifications in world dam catchments by nation to know the spatial distribution of floor water dynamics (Fig. 9). The continental summaries masks appreciable spatial heterogeneity throughout international locations. The biggest contributors to dam-induced floor water change come from creating international locations with fast hydropower set up, equivalent to China and Brazil. China alone accounts for greater than half of the entire floor water improve in Asia. In Brazil, 3,459 dams contribute to greater than 80% of the water growth in South America. Developed international locations with traditionally giant inventories of dams, equivalent to Canada and america, additionally see a major rise in floor water. The growth in floor water, nevertheless, is probably not a common characteristic of dams. In a couple of international locations, most notably Zambia, Eire, and the United Arab Emirates, floor water could have decreased barely after dams are constructed. We can not rule out the likelihood that the decline is because of idiosyncratic modifications in satellite tv for pc knowledge high quality or measurement error, as we’ve discovered the floor water loss to be lower than 5 km2 in these three international locations. Echoing prior findings30, these outcomes recommend that losses in everlasting and seasonal water from dams could also be extra concentrated than positive factors.
Floor water change in world dam catchments between 1984 and 2018 by nation. The biggest improve in floor water protection are in developed and quickly creating international locations with vital dam counts, equivalent to Brazil, Canada, China, India, america, Mexico, Turkey, and varied Southeast Asian international locations. The information captures modifications within the median floor water pixel counts between pre- and post-construction intervals inside dam catchment areas. Along with dam building, a portion of the floor water modifications in some international locations, particularly these in Asia, may very well be attributed to enhancements in satellite tv for pc knowledge high quality.
