Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia

Hydrological Sciences Journal ISSN: 0262-6667 (Print) 2150-3435 (Online) Journal homepage: http://www.tandfonline.com/loi/thsj20 Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia RAMIRO PILLCO ZOLÁ & LARS BENGTSSON To cite this article: RAMIRO PILLCO ZOLÁ & LARS BENGTSSON (2006) Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia, Hydrological Sciences Journal, 51:1, 98-114 To link to this article: https://doi.org/10.1623/hysj.51.1.98 Published online: 19 Jan 2010. Submit your article to this journal Article views: 2388 View related articles Citing articles: 15 View citing articles Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalinformation?journalcode=thsj20

98 Hydrological Sciences Journal des Sciences Hydrologiques, 51(1) February 2006 Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia RAMIRO PILLCO ZOLÁ * & LARS BENGTSSON Department Water Resources Engineering, Lund University, Sweden lars.bengtsson@tvrl.lth.se Abstract Lake Poopó, within the large Altiplano basin of Bolivia, is connected upstream to Lake Titicaca and downstream to the salares, the big salt fields. Small changes in precipitation and river inflows strongly affect the extent of the lake surface area. For times when there are no satellite images, it is difficult to determine the extent of the lake from observations. Water balance computations were performed to create a water-level series for Lake Poopó extending back in time. The dominant water inflow to Lake Poopó is from the River Desaguadero, which constitutes the outflow of Lake Titicaca. The water-balance computations confirm the crude peasant knowledge about historical lake status. It is found that if the lake level is less than 1 m during the wet season, there is a risk that this shallow lake dries out in the dry season. Key words Altiplano; ENSO; Lake Titicaca; palaeolake; water balance Variations à long terme et extrêmes du niveau d eau du Lac Poopó peu profond, en Bolivie Résumé Le Lac Poopó, au sein du grand bassin de l Altiplano bolivien, est lié à l amont au Lac Titicaca et à l aval aux salars, les grands champs de sel. De petits changements dans les précipitations et les apports fluviaux modifient fortement l extension de la surface du lac. Pour les périodes où aucune image satellitale n est disponible, il est difficile de déterminer l extension du lac à partir d observations. Des calculs de bilan hydrologique ont permis de reconstituer des séries passées du niveau de l eau du Lac Poopó. Les entrées d eau principales dans le Lac Poopó proviennent du Rio Desaguadero, qui est ellemême l exutoire du Lac Titicaca. Les calculs du bilan hydrologique confirment les informations empiriques des paysans relatives à l histoire du lac. Il apparaît qu il y a un risque d assèchement de ce lac peu profond en saison sèche si son niveau d eau est inférieur à 1 m durant la saison humide. Mots clefs Altiplano; ENSO; Lac Titicaca; paléolac; bilan hydrologique INTRODUCTION Lake Poopó in Bolivia is a very shallow lake, covering an area up to 3000 km 2. The inflow of water to the lake is from the Desaguadero River, constituting the outflow from the upstream Lake Titicaca, from regional rivers and from the precipitation on the lake. Almost all the rain falls during December March and, consequently, the small rivers are intermittent and the flow in the Desaguadero River highly seasonal. In the dry season, the lake surface area decreases to less than half of the wet-season area. The lake has dried out on several occasions. In 1994 it dried out and remained almost dry until 1997. The biomass of the lake was destroyed. For many years, even after 1997, when there was again water in the lake, the fishermen and their families could not earn their living. Partly to assure water inflow to Lake Poopó, a regulation system was built at Lake Titicaca in 2001. However, the main reason for the regulation is to maintain the high water level of Lake Titicaca. The Authority of Lake Titicaca (ALT) says that it has the tools to protect the Altiplano ecosystem in the future (Revollo, 2001). However, to assure substantial inflow via the Desaguadero River to Lake Poopó requires that the regulation scheme is planned with a very long lead time. * Also at: Institute of Hydraulics & Hydrology, San Andres Major University, Bolivia. Open for discussion until 1 August 2006

Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia 99 The Poopó region is very poor; the poorest people are the fishermen. When the lake area is small (i.e. lake water levels are low), it is hard to reach the water and there are few or no fish in the lake. The already high concentrations of heavy metals increase. It is important to know the dynamic behaviour of the lake area in response to meteorological variations to be able to project impacts on the social conditions among the fishermen population. The objective of this paper is to show how the water level of Lake Poopó and thus the lake area varies, and how these are related to the Titicaca Lake level and to meteorological conditions in the region around Lake Poopó. Such knowledge gives a lead time when planning for dry periods, restricting upstream irrigation and controlling how upstream Titicaca water can be used. Palaeological and historical lake conditions The Altiplano of Bolivia is the largest and highest endorheic basin in the world, occupying the western part of Bolivia and the southeastern part of Peru. Today there is a water system extending from Lake Titicaca in the north through the Desaguadero River to Lake Poopó and further to the two dry salt lakes in the south, Salar de Coipasa and Salar de Uyuni. The system is called the TDPS-system (Titicaca Desaguadero Poopó Salares system cf. Fig. 1). Enormous changes in the hydrology of the Altiplano have taken place in the last 200 000 years (Fornari, 2001; Fritz et al., 2004). Several lacustrine layers have been found in sediment cores taken from the Salar de Uyuni, of which the oldest dates back 190 000 years (Lake Escara). During the Minchin phase 35 000 70 000 years ago, first described by Minchin (1882), the Altiplano was wetter than at present (Argollo & Mourguiart, 2000). The last glacial maximum (LGM, approx. 26 000 14 000 BP) was a wet period (Baker et al., 2001), which was followed by intermittent dry and wet periods. During the Tauca phase (26 000 15 000 BP) (Fritz et al., 2004), the lower part of the Altiplano constituted one big lake including the salt flats and Lake Poopó, and reaching north almost up to Lake Titicaca. In a drier climate, this palaeolake developed into three lake basins, Lake Poopó and the two salares (Argollo & Mourguiart, 2000). The beginning of the Holocene was marked by an arid phase. The water level of Lake Titicaca dropped so much that the southern Titicaca basin, Lake Huinaimarca, was separated from the main part of the lake. The mid-holocene is interpreted as climatically unstable, with humid and dry periods. From about 3900 BP, the climate has been similar to that of today, although there have been climate fluctuations. The collapse of the Tiwanaku civilization (c. AD 1100) has been attributed to an extended dry period (Binford et al., 1997). Later, Lake Titicaca levels were very high in the 16th and 17th centuries, which means that the inflow via the Desaguadero River to Lake Poopó was high and, hence, Lake Poopó overflowed and covered a larger area than today. The present mean level of Lake Poopó is at 3686 m a.m.s.l. and that of Lake Titicaca at 3806 m. According to Rouchy et al. (1996), who studied carbonate algal biotherms, the depth of Lake Minchin was 45 m and the depth of Lake Tauca 100 m. Today the maximum water depth of Lake Poopó may vary by 2 3 m from the state that the lake is dry. The lake may be considered a terminal lake. It rarely spills over. The last event of this kind occurred in 1986, and there is no evidence that it has occurred in the previous 80 years. Therefore salt accumulates in the lake. At the outlet sill level,

100 Ramiro Pillco Zolá & Lars Bengtsson Fig. 1 The TDPS (Titicaca Desaguadero Poopó Salares) system. the lake area is about 3000 km 2. The outlet, when there is any flow, is the Laka Jawira River. The lake was dry between 1994 and 1997 and, according to local people, also dry or nearly dry in the early 1940s and in the early 1970s. Lake Poopó includes two lake basins: Lake Poopó proper and Lake Uru-Uru. At high water levels the two lakes form a single water body. The Desaguadero River divides into two branches before reaching the lakes, one easterly toward Lake Uru- Uru, carrying water only when the river discharge is high, and one westerly towards the main Lake Poopó. Lake Uru-Uru is shallower than the main Lake Poopó, only about 1 m, and dries out every year. The Desaguadero River changed its course in 1985. Prior to 1985, according to Iltis (1993), the entire Desaguadero River ran through Lake Uru-Uru before reaching Lake Poopó. Iltis also reports that Lake Uru- Uru did not exist 100 year ago. There are water level observations for Lake Poopó from 1920 to the present (Marin & Quitanilla, 2002), but for low water levels those observations just give indications that the water level has been low; it is not possible to move over the soft bottom to reach open water, when the water level is very low. From the mid-1970s, satellite images are available and, hence, information about the dynamic lake area variation can be obtained. The lake seems from the records of Marin & Quitanilla (2002) to have been dry between 1939 and 1944, and nearly dry in 1970 1972. The record shows that the lake level was very high in the mid-1930s, again in 1964 1965

Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia 101 and in the mid-1980s. The lake spilled over in 1985 1986. In early 2001, the lake almost reached its sill level. The water depth is related to the aerial extent of the lake. The absolute depth scale may have changed over the years, since the difference between the lowest recorded levels in the 1940s and the high levels in the late 1980s is almost 6 m, which is much more than the maximum lake depth. Climate and geomorphology Lakes Poopó and Uru-Uru are surrounded by two mountain chains of the Andes (cf. Fig. 1). From a morphological perspective, the Poopó basin (Fig. 2) can be divided into three regions: (a) the eastern mountain region, which occupies about 40% of the total Poopó basin area, ranging in altitude between 3800 and 5000 m a.s.l.; (b) the western mountain region. occupying only a small part of the basin at altitude 3800 4800 m a.s.l.; and (c) the inter-andean flat area, constituting half of the Poopó basin at 3700 3800 m a.s.l., where Lake Poopó, the smaller Lake Uru-Uru and the lower reaches of the regional rivers are located. The area of the entire Lake Poopó basin, excluding the Desaguadero River, is about 15 000 km 2. There are 22 identified intermittent river inflows, the river basins of which are shown in Fig. 2. These regional rivers originate in the mountains surrounding the lake and dry out in the dry season each year. Since the 1950s there is a dam in the Tacagua River (Fig. 2, sub-basin no. 9). Basically all the water is used for water supply and is released to Lake Poopó only during very rainy seasons. T i t i c a c a L a k e N D e s a g u a d ero 5 11 15 10 W S E 3 R i v er S e m i a r i d zone B o u n d a r y o f c l i matic zones Raingauges L a k e s S u b - b a s i n s I n f l o w a n d o u t f low rivers 8 2 14 Uru-UruLake 7 17 12 9 Poopó Lake 4 Outle 13 16 6 Arid zone 1 30 0 3 0 60 Kilometers Fig. 2 The Poopó (Poopó and Uru-Uru lakes) basin, showing the 22 regional river basins (17 of them are numbered but the smallest ones are not), and the five meteorological stations with records since 1960. The area north of the dashed line is considered semiarid and the area south of the line arid.

102 Ramiro Pillco Zolá & Lars Bengtsson The climate of the Altiplano is classified as semiarid-cold for the northern and middle parts, while the southwestern part is arid. It is characterized by a wet season in November March and a dry season in April October (TDPS, 1993; Garreaud et al., 2003). Garreaud (2000) showed that the moisture source is from the continental lowland east of the Andes. The seasonal variability of precipitation over the Altiplano is related to changes in the upper troposphere circulation. During the Austral summer, an upper level cyclone is established to the southeast of the central Andes, due to warming. In the northern part of the Bolivian High, easterly winds prevail and allow the influx of moisture from the central continent over the plateau (Garreaud, 1999; Vuille et al., 2000). The mean annual rainfall in the Altiplano Basin varies between 800 and 200 mm in a north south direction; in the Poopó basin, according to Roche et al. (1992), it is about 350 mm. In the regional Lake Poopó basin, the annual precipitation in the northern part is 420 mm and in the southern part 270 mm (see below). As a result of the circulation transitory phenomenon, the precipitation over the Altiplano is sensitive to variations in the large-scale circulation patterns. These variations are expressed as inter-annual variability in the precipitation over the Altiplano. The effects of El Niño-Southern Oscillation (ENSO) on the atmospheric circulation are observed as variations in the amount of precipitation in the Altiplano, so that El Niño years are related to below-normal precipitation and La Niña years to the opposite (e.g. Aceituno, 1988; Lenters & Cook, 1999; Garreaud, 1999, Vuille, 1999). This signal is less clear in the northern Altiplano than in the dry southwestern part and the physical links relating the variations in large-scale circulation and precipitation over the Altiplano are not completely understood. There may be warm El Niño episodes associated with wet or normal precipitation years, as well as La Niña episodes associated with dry or normal ones. Vuille et al. (2000) and Garreaud & Aceituno (2001) explain that precipitation over the Altiplano is most closely related to winds in the high troposphere in such a way that predominant easterly wind anomalies are associated to wet years and westerly ones to dry years. The influence of the climate of the Andes on greenhouse gases has very recently been discussed by Coudrain et al. (2005) and by Carracso et al. (2005). The soil of the Altiplano is heterogeneous sedimentary infill. It has a fluvial lacustrine and alluvial origin. The soils in the mountainous areas are shallow and not well developed, while the soils along the rivers and on the hillsides have a higher level of evolution. Along the shores of Lake Poopó the soil is saline with a thin layer of salt deposited on top of the sediments. These are areas of little vegetation, occasionally covered by latifolia herbaceous vegetation type. The soil has a compact structure, which impedes infiltration (PROBONA, 1995). The vegetation consists of tropical alpine herbs with dwarf shrubs, which changes with the increased aridity towards the south of the Altiplano (Rivera et al., 1996). Rainfall and evaporation Rainfall is measured at 14 stations within and close to the Poopó basin. However, only five of them have records extending some decades back (1960 2002). These stations are evenly distributed around the lake (see Fig. 2) and, together, are representative of the precipitation regime. As is shown below, the five stations with long-term records

Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia 103 also represent well the precipitation in the Poopó basin. The majority of the data used in this study were obtained from the Servicio Nacional de Meteorología e Hidrología de Bolivia and some from Proyecto Piloto Oruro (PPO, 1996). Some gaps (2 13% of daily station data for different years) from one or two of the five stations for the 40- year period could be filled by correlating with the neighbouring two stations. Precipitation is unevenly distributed over the year (Fig. 3). The monthly values shown are computed as the mean of the five long-term stations for 40 years. As may be seen from Fig. 3, most of the precipitation falls in December March. Since the wet period extends from an old into a new calendar year, the inter-annual precipitation variation is best described by comparing 12-month precipitation for the hydrological year (1 October 30 September). The annual rainfall (a simple mean of the five stations) for hydrological years 1960 2002 is plotted in Fig. 4. The average annual rainfall for the 42-year period is 372 mm. There are four years with precipitation less than 250 mm and two with precipitation exceeding 500 mm. Observed annual precipitation over the Lake Poopó basin (Fig. 4) shows some evidence of the influence of ENSO, even though the correlation between annual precipitation and the Southern Oscillation Index (SOI) is not significant. Evidence of ENSO can be characterized by the dry El Niño years of 1965/66, 1982/83 and 1997/98, 90 80 Rainfall (mm) 70 60 50 40 30 20 10 Rainfall Standard deviation 0 J F M A M J J A S O N D Months Fig. 3 Long-term monthly average rainfall in the Poopó basin. 600 500 Precipitation (mm) 400 300 200 100 0 60-61 62-63 64-65 66-67 68-69 70-71 72-73 74-75 76-77 78-79 80-81 82-83 Years 84-85 86-87 88-89 90-91 92-93 94-95 96-97 98-99 00-01 Fig. 4 Annual rainfall the Poopó basin for hydrological years 1960 2001.

104 Ramiro Pillco Zolá & Lars Bengtsson which all had very low precipitation, less than 250 mm. Also, 1968/69 and the period 1991 1995 were El Niño years and, although the precipitation was below the normal value, the precipitation signal was less pronounced. The wet years of 1996/97 and the mid-1970s correspond to La Niña years. However, the very high precipitation of 1984/85 is not related to La Niña. Data from all 14 stations are available from 2001, when the present study was initiated, and for the period 1974 1985. The precipitation pattern was determined using these monthly precipitation data, applying kriging with an exponential variogram. There is clearly a gradient toward less precipitation in the south. There is an east west gradient only in the very south, where the precipitation is higher to the east than to the west. North of Lake Poopó including Lake Uru-Uru and quite far south along the eastern side of Lake Poopó, the annual precipitation in the period 1974 1985 was 450 500 mm. South of the lake and quite far north along the western side, the annual precipitation was less than 300 mm. The mean potential evaporation rate over the entire Altiplano is estimated at more than 1500 mm year -1 (Roche et al., 1992), and varies from 1500 to 1800 mm year -1 in a north south direction (TDPS, 1993). Pan observations are available from six stations near or within the Lake Poopó basin for the period 1990 1995. By comparing the pan observation data with the evaporation computed using the Penman equation, a pan coefficient of 0.87 was estimated. The monthly corrected pan evaporation values vary between 110 and 170 mm month -1, with an annual value close to 1700 mm. Using the precipitation information from all the meteorological stations, as well as the information about potential evaporation, it was possible to divide the regional Lake Poopó basin into a semiarid northern part, which constitutes about two thirds of the basin where the annual 40-year precipitation is about 420 mm, and an arid southern part, one third of the area with annual precipitation of 270 mm. The boundary between the two zones of the Lake Poopó basin is shown on the map of Fig. 2. This means that the mean for the entire Poopó region is very close to the annual precipitation of 372 mm, determined as the mean of the five long-term stations. The Desaguadero River Water is transported from Lake Titicaca to Lake Poopó via the 300-km long Desaguadero River. The river receives water from several tributaries on its way to Lake Poopó. The main contribution is from the Mauri River (mean flow 21 m 3 s -1 in 1977 2001) southwest of Lake Titicaca. Nevertheless, just upstream from Lake Poopó the discharge in the Desaguadero River is very dependent on the water level or depth of Lake Titicaca, as shown in Fig. 5. The average annual discharge in the downstream Desaguadero River at the inlet to the two Poopó lake basins is 66 m 3 s -1 for the 52 years, 1960 2002; the maximum annual discharge is 279 m 3 s -1 and the minimum is 16 m 3 s -1. Lake Titicaca water level was low in the 1960s and high in the 1970s and 1980s. The water level dropped several metres from the mid-1980s to the mid-1990s, but has recovered slowly in the last years. The maximum water level of Lake Titicaca in the last 40 years occurred in April 1986 when it was 3811.73 m. A low of 3807.37 m was observed in December 1996. The Titicaca water levels over time (1961 2001) are shown in Fig. 5 and can be compared with the Desaguadero River flow.

Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia 105 Q (m 3 s -1 ) 300 250 200 150 100 50 0 Desaguadero River discharge Lake Titicaca water level 312 311 310 309 308 307 306 Water level +3 500 m 1961 1966 1971 1976 1981 Years Fig. 5 (right axis) Annual mean water level of Lake Titicaca observed at Huatajata and (left axis) discharge observed at Chuquiña just upstream of lakes Poopó and Uru-Uru. 1986 1991 1996 2001 400 300 m 3 s -1 200 100 0 1977 1979 1981 1983 1985 1987 1989 1991 1993 Fig. 6 Monthly Desaguadero discharge at Chuquiña hydrological station. There is a clear seasonality in the Desaguadero River flow, which is demonstrated in Fig. 6 for the period 1977 1985. The inflow to Lake Poopó usually starts to increase in early December and is high in the first three months of the year; then it drops to low values from April onwards. The highest monthly recorded discharge (504 m 3 s -1 ) was in March 1986 and the lowest monthly flow (3 m 3 s -1 ) was recorded in October 1998. Water is withdrawn from the Desaguadero River along its course to Lake Poopó for irrigation and also to be used within the mining industry. Even downstream of the measuring station at Chuquina, water is withdrawn. From interviews and observations of lake levels, the withdrawal downstream of Chuquina was estimated at 6 10 m 3 s -1. Regional river flows Lake Poopó receives water from precipitation on the lake, from the Desaguadero River, as well as from 22 regional rivers. The biggest of these is the Marquez River,

106 Ramiro Pillco Zolá & Lars Bengtsson 12 Q observed (m 3 s -1 ) 10 8 6 4 2 Sevaruyo River Antequera River 0 J F M A M J J A S O N D J F M A M J J A S O N D Months (2001, 2002 year) Fig. 7 Observed discharge in the Sevaruyo and Antequera rivers. with a drainage basin area of 2577 km 2. In the last few years, hydrometrical stations have been established in 13 of the rivers, but there are continuous measurements in only two of them: the Sevaruyo (Fig. 2, sub-basin no. 6), which has its river mouth in the southeast, and the Antequera (Fig. 2, sub-basin no. 12), flowing into the northeastern part of Lake Poopó. The seasonal flow variations in these two rivers for the years 2001 and 2002 are shown in Fig. 7. The areas of drainage basins of the rivers Sevaruyo and Antequera are 852 km 2 and 227 km 2, respectively. Although the Sevaruyo basin is 3.5 times larger than the Antequera basin, its discharge is hardly even double that of the Antequera. The Antequera River basin is representative of the northern semiarid part of the regional Poopó basin, which constitutes about two thirds of the basin, and the Sevaruyo River basin is representative of the southern arid part. The annual rainfall in the Antequera River basin is about 430 mm, and that in the Sevaruyo River basin is about 270 mm. Observations in the other rivers have been intermittent, but reveal a similar seasonal pattern. It is therefore thought that the inflow to the lakes from these two rivers is representative for the regional river inflow to the lakes. A simple area-to-area relationship was used to estimate the regional runoff from the two representative rivers. Since measurements of regional river inflow are available only from 2001 and 2002, it is necessary to simulate the river inflow, if the total inflow of water to Lake Poopó over long times is to be estimated. The monthly runoff from the entire Poopó basin was simulated using a conceptual rainfall runoff model, SIMULA, developed at the Centro de Estudios Hidrológicos de España and transformed to Visual Basic at the Instituto de Hidráulica y Hidrología de La Paz. The model distinguishes between surface flow and groundwater. The separation between surface and groundwater is related to the water excess during an event and thus to the infiltration capacity. The surface water is assumed to reach the river directly, while the distribution of groundwater storage and runoff is related to the groundwater storage. The sub-surface water is stored in many layers and the groundwater runoff determined using a recession coefficient. The most important parameters are storage capacity of the soil, maximum infiltration in a month, and the recession coefficient. Monthly rainfall and number of rainy days in a month were used as input. Monthly potential evapotranspiration, also required as input, was taken from the pan observations previously mentioned; the same monthly values were used for all years.

Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia 107 (a) 12 10 Q (m 3 s -1 ) 8 6 4 Observed Computed 2 (b) Q (m 3 s -1 ) 0 9 8 7 6 5 4 3 2 1 0 Jan Mar May Jul Sep Nov Jan Mar May Jul Sep Nov Observed Computed Jan Mar May Jul Sep Nov Jan Mar May Jul Sep Nov Years 2001-2002 Fig. 8 Observed and computed discharge in (a) the Sevaruyo River and (b) the Antequera River. 250 200 150 100 50 0 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 Years Fig. 9 Computed monthly regional river inflow (m 3 s -1 ) to Lake Poopó. The SIMULA model was calibrated versus the runoff in the two rivers, the Sevaruyo and the Antequera. The comparison between observed and modelled monthly discharge for both the rivers is shown in Fig. 8. Since data existed only for two years, there was no possibility to validate the simulations. However, as is shown further down the regional inflows are minor compared to rain on the lake and especially compared to the Desaguadero River inflow, so it is mainly important to have a crude estimate of the river inflows. By extending the modelling period to include all months from 1960 to 2002, taking two thirds of the modelled Antequera River specific runoff and one third of the

108 Ramiro Pillco Zolá & Lars Bengtsson Sevaruyo River specific runoff and multiplying by the Lake Poopó basin area excluding the lake itself, the regional monthly river inflows to the lake were computed. Computed monthly discharges for the period 1960 to 2002 are shown in Fig. 9. The mean annual regional river inflow to Lake Poopó in the investigated period 1960 to 2002 was 8 m 3 s -1. Bathymetry Since Lake Poopó is so shallow and the shores are almost flat, a small change in the water level results in considerable change in the water surface area. Reduced lake area means that the evaporative loss from the lake is reduced, and if it rains, so is the rainfall input. The water balance of the lake is dependent on the lake area. To determine the water level from river inflows, rainfall and lake evaporation, the relationship between water depth and lake surface area must be known. The first bathymetric map was constructed already 100 years ago by Neveu- Lemaire and a second, more detailed, map much later by Boulange et al. (1978). The map of Boulange was based on three transverse sections. Boulange found the lake area to be larger for low lake levels than Neveu-Lemaire did for higher levels. This could be attributed to difficulties in measuring and the fact that the depth area relationships were based on only few depth measurements, but may also be due to changes in the bottom topography over time. Naveu-Lemaire suggested a maximum lake area of 2530 km 2 and Boulange a maximum area of 2650 km 2. Iltis (1993) claims that the lake was 3500 km 2 when it spilled over in 1985 1986. An updated bathymetry map (Fig. 10) was constructed from the old maps, satellite images extracted from the database http://www.eot.jsc.nasa.gov, measurements of lake level along two cross-sections in 2002 performed by the first author and from interviews with local people. In Lake Poopó there is an island, Panza Island, which is not visible at high water levels. The extent of the island area above water and marks on the island were used when interpreting the satellite images. The relative depth determined by comparing satellite images and land marks was converted to an absolute depth scale by setting the deepest measured point as zero. When the lake spills over, the maximum depth is 2.37 m, the surface area is 3011 km 2 and the volume is 3.415 km 3. The relationship between surface area (A, in km 2 ) and depth (h, in m) is: A + 2 = 400 + 400h 300h (1) The area in equation (1) above does not include the water storage of Lake Uru-Uru, whose surface area is 280 km 2, and the volume is 0.082 km 3, when the depth is at its maximum 0.75 m, which corresponds to the spill-over depth of Lake Poopó. Thus, as long as the water level of Lake Poopó is less than 1.62 m, Lake Uru-Uru is dry. There is a connection between Lake Uru-Uru and Lake Poopó, when the Lake Poopó surface area is more than 2000 km 2. However, when the Desaguadero discharge is high, there may be inflow to Lake Uru-Uru already before Lake Poopó has reached such a large extent. Since no better information is available for Lake Uru-Uru, a linear relationship between depth and lake area is assumed so that: AUru -Uru = 370h (2)

Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia 109 Fig. 10 Bathymetric map of Lake Poopó, 2002. The elevation curves are given as water depth (h) and as level above mean sea level. Measurements in 2000 and 2001 reveal that, at high Desaguadero River levels, about 15% of the river flow is directed towards Lake Uru-Uru. Water balance and lake level computations Knowing the input of water from rainfall (measured), the regional rivers (simulated from rainfall and potential evapotranspiration) and from the Desaguadero River (measured), and the evaporative losses from the lake (standard monthly values from corrected pan observations), and also knowing the relationship between surface area and water depth, it is possible to determine the lake level variations of Lake Poopó from simple water balance computations: dv/dt = Q Des + Q reg Q out + (P E lake ) A lake (3) or with a point description dh/dt = (Q Des + Q reg Q out )/A lake + P E lake (4) where V is lake volume, A lake is surface area, h is water depth, Q Des is monthly inflow from the Desaguadero River, Q reg inflow from the regional rivers, Q out outflow from the lake, P is monthly precipitation on the lake and E lake is lake evaporation. In a steady-state situation, when there is no outflow, the lake surface area is related to the river inflow, precipitation and evaporation as (Bengtsson & Malm, 1997): A lake = (Q Des + Q reg )/(E lake P) (5)

110 Ramiro Pillco Zolá & Lars Bengtsson For the mean flow of the Desaguadero River, 56 m 3 s -1, with the assumption of 10 m 3 s -1 withdrawal 10 m 3 s -1 regional inflow, 1700 mm annual evaporation loss and 370 mm rain on the lake, the lake surface area at steady state should be 1500 km 2, which, from the area depth equation (equation (2)), corresponds to a depth of almost 1.4 m. Of the input of water to Lake Poopó, the flow from the Desaguadero River constitutes about 65% and the contribution from the regional rivers is 10%. There is very seldom any outflow from Lake Poopó. When there is, an outflow equation relating lake level and outflow is required to determine the water level from water balance computations unless the outflow is measured. Outflow from Lake Poopó occurred in June 1986. The discharge in the Laka Jawuira River was measured in mid- June to be 60 m 3 s -1. This value was used for the outflow as long as the computations showed that the lake level was above the outlet sill level. However, when the time resolution of the computations is as coarse as a month, it can just as well be assumed that all the water that is computed to reach above the sill level spills over, since the storage between the sill level and the actual water surface is small. The result of the long-term Lake Poopó water level computations is shown in Fig. 11. The water level of Lake Titicaca is plotted in the same figure. Depths deciphered from interpreted satellite images using equation (1) are also shown. The satellite images, a total of 19, show good agreement with the water balance computed lake levels. 5.0 4.5 4.0 3.5 Computed Lake Poopo water depth Lake Titicaca level (+3807 m) Lake Poopo depth from images 3.0 2.5 2.0 1.5 1.0 0.5 0.0 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 Fig. 11 Computed Lake Poopó depth (m) and observed water level in Lake Titicaca. Relationship between Lake Poopó and Lake Titicaca levels More than 60% of the input of water to Lake Poopó comes from the Desaguadero River, whose flow is related to the water level in Lake Titicaca. Thus, there ought to be a relationship between the water levels in the two lakes. Indeed, when inspecting Fig. 11, such relationship can be seen, although the level of Lake Titicaca fluctuates more than that of Lake Poopó; also, the level of Lake Poopó did not fluctuate with the level of Lake Titicaca in 1985 to 1989, when levels of both lakes were high and Lake Poopó was near, at, or above its outlet sill level. The Lake Poopó monthly water level dependency on Lake Titicaca level is described by the relationship:

Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia 111 h = 0.13 + 0. 70 (6) P h T which explains 90% of the variance. The two depths with index P for Lake Poopó and T for Lake Titicaca refer to the bottom depth of Lake Poopó and the sill level of Lake Titicaca. The equation does not hold when there is outflow from Lake Poopó, nor when Lake Poopó is dry. The estimates by regression and by water balance are compared in Fig. 12. 3.5 3.0 2.5 2.0 1.5 Ccomputed from water balance Computed from regression Computed from water balance 1.0 0.5 0.0 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 Fig. 12 Lake Poopó water depths (m) computed from water balance and using regression with Lake Titicaca levels. ENSO-effect on Lake Poopó levels El Niño Southern Oscillation is the most important factor to explain interannual climate variations on the Altiplano. Therefore, this phenomenon is likely to be relevant for the Lake Poopó levels. The lake level record in Fig. 11 shows that the El Niño event of 1982/83 resulted in fast declining water level through 1983, and that the low levels of 1992 1996 can be attributed to El Niño events of 1991 1995. The lake was low in the El Niño years 1965/66 and 1968/69. The increase in lake levels during the first half of the 1970s coincides with a wet period over the Altiplano, which Vuille et al. (2000) and Garreaud et al. (2003) have linked to ENSO variability. The lake level grew from dry conditions following La Niña of 1996/97. As reported already in the precipitation section, not all wet or dry years and consequently not all high or low water level periods coincide with La Niña or El Niño events. Drying out and lake recovery As may be seen from Fig. 11, Lake Poopó recovers quite fast from low to high levels. In the first five months of 1984, the computed water level increases from 0.5 m to more than 2 m; the latter level is confirmed from satellite images. An even faster recovery occurred in early 2001, when, in three months, the lake recovered from being almost dry to close to spill-over level. The mean Desaguadero River discharge in these three months was 360 m 3 s -1, which is the highest three-month mean recorded; the regional river inflow was 140 m 3 s -1 during the first two months and 30 m 3 s -1 in the

112 Ramiro Pillco Zolá & Lars Bengtsson third month. The precipitation and evaporation are of minor importance, when the lake area is small, e.g. in the beginning of a wet period. In the third month of 2001 the evaporation excess was about 0.1 m. The computations showed an increase in the water level to 1.4 m, 2.1 m, and 2.3 m in the first, second, and third month of 2001, respectively. The latter value was confirmed by satellite images. Simple water balance computations (Adh/dt = Q), assuming that the evaporation from the lake is balanced by the rainfall, show that, when the total inflow is constant at 600 m 3 s -1, the lake level grows from dry conditions to spill-over in two months; when the inflow is 500 m 3 s -1 it takes two and a half months, and for 400 m 3 s -1 three months. Drying out is a much slower process. The decline cannot be faster than the lake evaporation; thus, if there is no precipitation and no inflow, the drop in lake level is 0.8 0.9 m over the half year. The lake usually retreats in April December, when rain and inflow from the regional rivers are low. The Desaguadero River inflow is about 30 m 3 s -1. The monthly lake evaporation exceeds the precipitation on the lake by 140 mm. Using these values, the lake level drop can be computed. Depending on the initial lake area (when the dry period begins), the lake level drops at a rate of 0.05 0.1 m month -1. Lowering of the lake level of about 0.6 0.7 m in the dry seasons of the 1990s was observed (cf. Fig. 11). A decline over a longer time is found 1982 1983. The lake level was computed to have dropped from 2.1 m in April 1982 to 0.6 m in late December 1983, i.e. at the rate of 0.1 m month -1 during one and a half years. The Desaguadero River flow decreased continuously over that period. The precipitation in December March is usually 200 300 mm, being mainly concentrated in January and February. In the rainy period of 1982/83, the precipitation was only about 150 mm and distributed over several months. There was very little rainfall excess to produce regional runoff. The season-to-season variations in Lake Poopó level are smaller than the lake level variations over longer periods. The lake level was rather high even in the dry seasons from the mid-1970s until 1990. In the mid-1980s, when the lake was at or near spillover for many years (Fig. 11), the Desaguadero River mean flow was 150 m 3 s -1 before withdrawal for irrigation, the regional river flow somewhat more than 10 m 3 s -1, the annual precipitation 400 mm and the annual lake evaporation 1700 mm. In the mid- 1990s, when the lake was very low, the mean Desaguadero River inflow was 20 m 3 s -1 before withdrawals, the regional river inflow less than 4 m 3 s -1 and the annual precipitation a little more than 300 mm. When inserted into the steady-state lake area relationship (equation (5)), the equilibrium lake area for the 1990s is found to be 500 km 2, which corresponds to almost zero depth. The lake dried out in dry seasons during these years. The lake was in quasi-equilibrium from 1975 to 1982. The water level fluctuated every year with amplitude a little more than 0.5 m. In these years the Desaguadero flow was about 150 m 3 s -1 in January March, and 60 m 3 s -1 for the other nine months. The annual precipitation was 330 mm. Also during the extended dry period in the mid- 1990s, discussed above, the lake was in quasi-equilibrium. The water level increased from a completely dry lake to about 0.5 m in the wet season, and the Desaguadero River did not carry more than 50 m 3 s -1 in the three wettest months of any year. The 12-month precipitation has, as seen in Fig. 4, never dropped below 200 mm. The lake evaporation is stable. The regional river flows are minor, especially in dry years. Thus, the Lake Poopó level is mostly dependent on the inflow from the

Long-term and extreme water level variations of the shallow Lake Poopó, Bolivia 113 Desaguadero River. The lake level may drop 0.1 m every month in the 8 9 month dry period of a year. If the water depth is less than a metre in the wet season, there is a risk that the lake will dry out in the following dry season. For the lake level to increase from a very low level to 1 m in three wet months, the Desaguadero River flow must be almost 100 m 3 s -1. Thus, to avoid a disastrous situation for the local people, action should be taken when the water level is also low in the wet season. Water can be released from the now regulated Lake Titicaca, but only if there has been long-term planning of the management of the entire TDPS-system. Since there is an ENSO effect on the Titicaca Lake level and on the precipitation on the Altiplano there should be possibilities for strategic planning of the water distribution through the Desaguadero River. CONCLUSIONS The water level in Lake Poopó depends on regional precipitation and evaporation, but more on the Desaguadero inflow. Since the Desaguadero is the outflow from Lake Titicaca, there is a strong relationship between the mean annual water levels in Lake Poopó and Lake Titicaca as long as Lake Poopó does not spill over and does not dry out. Since precipitation variability in the northern part of the Altiplano is ENSOrelated, there is also a relationship to Lake Poopó. Since Lake Poopó is shallow and flat, a small drop in the water level results in a marked reduction in the water surface area. Water balance computations indicate that Lake Poopó was dry between 1994 and 1997 and was very low during 1969 1973. The computations also show that the lake can recover from almost dry conditions to normal or even to spill-over depth within a year. Drying of the lake to a very small surface area takes a longer time. However, it was found that, if the water depth is less than a metre in a wet season, there is a risk that the lake will be dry in a following dry season. This information, in combination with ENSO forecasts, can help in estimating lake levels with prolonged lead time and, further, in taking action so that low Lake Poopó levels fluctuations are not disastrous for the fishermen population. Acknowledgements This study was carried out within the framework of a collaborative research project between the Department of Water Resources Engineering, Lund University, Sweden and the Hydraulic and Hydrologic Institute (IHH) of San Andrés University, La Paz, Bolivia, financially supported by SIDA (Sweden). The authors are grateful to the Servicio Nacional de Hidrometeorología de Bolivia and to Proyecto Piloto Oruro for providing important data, and to all those who were involved in collecting and measuring hydrometeorological data in 2001 2002. The comments from the anonymous reviewers contributed to a considerable improvement of the paper. Dr Cintia Uvo shared her knowledge about ENSO. REFERENCES Aceituno, P. (1988) On the functioning of the Southern Oscillation in the South American sector. Part I: surface climate. Monthly Weather Rev. 116, 505 524. Argollo, J. & Mourguiart, P. (2000) Late Quaternary climate history of the Bolivian Altiplano. Quatern. Int. 72, 37 51.

114 Ramiro Pillco Zolá & Lars Bengtsson Baker, P. A., Rigsby, C. A., Seltzer, G. O., Fritz, S., Lowenstein, T., Bacher, N. & Veliz, C. (2001) Tropical climate changes at millennial and orbital time scales on the Bolivian Altiplano. Nature 409, 698 701. Bengtsson, L. & Malm, J. (1997) Using rainfall runoff modeling to interpret lake level data. J. Paleolimnol. 18, 235 248. Binford, M. W., Kolata, A. L., Brenner, M., Janusek, J. W., Seddon, M. T., Abbott, M. & Curtis, J. H. (1997) Climate variation and the rise and fall of an Andean civilization. Quatern. Res. 47, 235 248. Boulange, B., Rodrigo, L. A. & Vargas C. (1978) Morphologie, formation et aspects sédimentologiques du lac Poopó (Bolivie). Cah. ORSTOM, Ser. Geol. X(1), 69 78. Carrasco, J. F., Casassa, G. & Quintana, J. (2005) Changes of the 0 o C isotherm and the equilibrium line altitude in central Chile during the last quarter of the 20th century. Hydrol. Sci. J. 50(6), 933 948. Coudrain, A., Francou, B. & Kundzewicz, Z. W. (2005) Glacier shrinkage in the Andes and consequences for water resources Editorial. Hydrol. Sci. J. 50(6), 925 932. Fornari, M., Risacher, F. & Féraud, G. (2001) Dating of paleolakes in the central Altiplano of Bolivia. Palaeogeogr. Palaeoclimatol. Palaeoecol. 172, 269 282. Fritz, S. C., Baker, P. A., Lowenstein, T. K., Seltzer, G. O., Rigsby, C. A., Dwyer, G. S., Tapia, P. M., Arnold, K. K., Ku, T. L. & Luo, S. D. (2004) Hydrologic variation during the last 170 000 years in the southern hemisphere tropics of South America. Quatern. Res. 61(1), 95 104. Garreaud, R. D. (1999) Multi-scale analysis of the summertime precipitation over the central Andes. Monthly Weather Rev. 127, 901 921. Garreaud, R. D. (2000) Intra-seasonal variability of moisture and rainfall over the South American Altiplano. Monthly Weather Rev. 128(9), 3337 3346. Garreaud, R. & Aceituno, P. (2001) Interannual rainfall variability over the South American Altiplano. J. Climate 14, 2779 2789. Garreaud, R., Vuille, M. & Clement, A. C. (2003) The climate of the Altiplano: observed current conditions and mechanisms of past changes. Palaeogeogr. Palaeoclimatol. Palaeoecol. 194, 5 22. Iltis, A. (1993) Recent limnological changes in a saline lake of the Bolivian Altiplano, Lake Poopó. Int. J. Salt Lake Res. 2, 17 28. Lenters, J. D. & Cook, K. J. (1999) Summertime precipitation variability over South America: role of the large-scale circulation. Monthly Weather Rev. 127(3), 409 431. Marin, R. P. & Quintanilla, J. A. (2002) Efectos ambientales sobre las pesquerias de los ecosistemas de los Lagos Poopó y Uru-Uru (Environmental effects of fisheries of Poopó and Uru-Uru lakes ecosystem). Report Univ. San Andrés, Inst. of Ecology, La Paz, Bolivia. Minchin, J. (1882) Notes on a journey through part of the Andean tableland of Bolivia. Proc. Royal Geogr. Soc. 4, 67. PPO (Proyecto Piloto Oruro) (1996) Ministerio de Desarrollo Sostenible y medio Ambiente Secretaría Nacional de Minería Swedish Geological AB. R-BO-E-9.45-9605-PPO, LP, Bolivia. PROBONA (Programa de Bosques Nativos Andinos) (1995) Formaciones vegetacionales del área andina de Bolivia. Programa de Bosques Nativos Andinos, La Paz, Bolivia. Rivera, M. O., Liberman, M., Beck, S. & Moraes, M. (1996) Vegetación de Bolivia. comunidades, territorios, indígenas y biodiversidad en Bolivia (Bolivian vegetation. Indigenous communities, territories and biodiversity). CIMAR / Universidad Autónoma Gabriel Rene Moreno, Santa Cruz, Bolivia. Revollo, M. M. (2001) Management issues in the Lake Titicaca and Lake Poopó system: Importance of developing a water budget. Lakes & Reservoirs: Res. Manage. 6, 225 229. Roche, M. A., Bourges, J., Cortez, J. & Mattos, R. (1992) Climatology and hydrology of the Lake Titicaca basin. In: Lake Titicaca. A Synthesis of Limnological Knowledge (ed. by C. Dejoux & A. Iltis), 63 88. Monographiae Biologicae 68, Kluwer, Dordrecht, The Netherlands. Rouchy, J. M., Servant, M., Fournier, M. & Causse, C. (1996) Extensive carbonate algal bioherms in upper Pleistocene saline lakes of the central Altiplano of Bolivia. Sedimentol. 43, 973 993. TDPS (1993). Climatología del Sistema de los lagos Titicaca, Desaguadero, Poopó y Salares Coipasa y Uyuni (TDPS) (Climatology of joint system of Titicaca, Desaguadero and Poopó lakes and Coipasa and Uyuni salt pans). Comisión de Comunidades Europeas-Repúblicas del Peru y Bolivia, convenios ALA/86/03 y ALA/87/23. LP, Bolivia. Vuille, M. (1999) Atmospheric circulation over the Bolivian Altiplano during dry and wet periods and extreme phases of the southern oscillation. Int. J. Climatol. 19, 1579 1600. Vuille, M., Radley, R. S. & Keimig, F. (2000) Interannual climate variability in the Central Andes and its relation to tropical Pacific and Atlantic forcing. J. Geophys. Res. 105, 12447 12460. Received 10 December 2004; accepted 21 October 2005