THE LIMNOLOGY OF
A Large, High A1 titude Tropical Lake
Peter J. Richerson
Division of hv.1:ronmentaL Studies ard Institute of EcoZogy Universi* of CaLi,com?ia, Wis
Carl Widmer
ELbert Cove12 College Uniuer~ityof the PackjYc Stock ton, CaZ-ifimCc
Timothy Ki ttel
Dfviston of ,%~irortwntatStudies and Ee"coZogy Graduate Group University of California, Davis
Institute of Ecology Pub? ication #14, June, 1979 - - esta sincera - Dedicamos monograffa a1 pueblo del Altiplano, con una gratitud profunda y par su ayuda y pacimcia para con nosotros, y a nuestros colegas del Instituto del Mar del * Perb y de la Unjversidad Nacional Tgcnica del A1 tiplano, sin cuya colaboraci6n continua no
hubiera sido posible este trabajo.
Estamos concientes de que estas personas viven, estudian y enfrentan su realidad para
transformarla y enr-iquecerla acorde con el proceso peruano nacional actual. Querems ofre- cer esta monograffa corn nuestra propia contribuci611, aun limitada, de participacidn en
ese proceso.
The authors are grateful for the financial support provided by the National Geographic Society, the Faresta Institute for Ocean and Mountain Studies, the University of Cal ifornia
(through a Rockefel 1er Foundation institutional development grant), the Jastro-Shiel ds
Scholarship Comi ttee and the University sf the RacifSc. Widmer was supported by an Organ-
ization of American States fellowship and Richerson was partly supported by National Sci-
ence EMS during this study. -\ Foundation grants GA-34099 and 75-14273 We are a1 so indebted - to the Instituto del Mar del PerG, the Universidad Nacional Tecnica del A1 tiplano, the Ser- - vf cio Nacional de Meteorelogfa e Midrol ogfa, and to individual s including Victoria Val cgr-
cel , Roger Srni th, Fl orentino Tito, Gerald Fisher , John Me1 ack , dswal do Zea , Linda Thorpe , Tim Long, Mayne Wurtsbaugh, Kenso Kawahira, Leigh Speichinger, Fathers Patrick, Eugene and
John, and Sister Margaret. Antonio Landa C., in charge of Peruvian studies on the lake,
was particular1y he1 pful in many ways. His field personnel , Alejandre Ardiles and Edgar
Farfan, provided important logistical servSces, and Sr. Farfan has done us several inoor-
tant favors since our leaving the lake. Charles R. Goldman gave us valuable advice, criti-
cism, and the equipment to count 14-C. Eleodoro Aquire was very helpful in sending us
weather data. Vern Scott and the Independent Study Committee were instrumental in allowing
Ki ttel to spend an extended period in Perd. Leonard Myrup and Tom Powel 1 gave us valuable
advfce regarding the heat budget, and we thank them and Dan Godden for the unptrblished Lake
Tahoe heat budget figure. Matilde Lopez and A7 Mahood helped very substantially with algal
species identifications. Ben Orlove furnished unpublished calculations of fish harvest.
Uy-lhanh Ly drafted rn a n y of the figures and Dolores Durnont, Wanda Greene, Lyn Schonewise,
and Cliff Shockney typed the drafts. The editorial efforts of Geoffrey Handesforde-Smith
are grateful 1y acknowledged.
i i TARLE OF CONTENTS
Resumen - Abstract
Introduction
General Descri ption of take Ti ticaca
Physical and Chemical Limnology
Physical Measurements
Cherni cal Measurements
Riological Limnoloqy
Phytopl ank ton Populations
loop1ankton Ropul a tions
Fish Fauna
Primary Production
Production of Hiqher Trophic Levels
Concluding Discussion
Literature Cited RESUMEN
2 El lago Titicaca es un gran lago de altura(8.100 Km ; 281 m prof. maxi 3.803 msnm) en el Altiplano andino peruano-boliviano, ubicado a 16' 1at.S. Una serie de 21 mediciones de variables lirnnol6gicas bdsicas fueron hechas en la parte honda cerca del pueblo de
Capachica, Perb, en 1973. Pas estaciones fueron muestreadas a intervalos de aproximadamente dos semanas desde enero a diciembre. Las mediciones incluyeron perfiles de temperatura, profundidad-Secchi, penetracidn de lur, pardmetros quTmicos especfficos y producci6n primaria. Mues tras de fitoplancton y zooplancton fueron col ectadas y enurneradas. Los datos de 3973 junto a la informaci6n recopilada por otros investrigadores fue usada para comparar la 1 irnnologia del lago Titicaca con lagos temperados y tropicales de las tierras bajas. El lago Titicaca es semejante a 10s lagos bajos tropicales en casi todos 10s aspectos de su 1 imnologfa a pesar de su temperatura fria. Sus perfiles termales muestran un epi 1irnnion ampl io, con relativamente pequezas diferencias entre temperaturas epilimnPticas (rnax. 16%) e hipolimngticas (11.l0~jy un perTodo de remocibn mds o menos cornpleto en la estacion seca, Julio a Septiembre. El balance cal6rico es aproximadamente equilibrado durante todas las estaciones, principalmenre por la evaporaciBn, la que se a proxima a Zm a%- ' . Consecuentemente, la estratificacidn es d6bil pero persistente y la rnagnitud de la rnezcla en profundidad variaria substancialrnente de afio en azo. El hipolirnnion no se mezcl6 por debajo de los 150m en 1973. El lago es bastante transparente parque la biomasa de fitoplancton circula en un epi limnion ampl io y la bjomasa de crus tzceos predadores es pemanentemente grande; 1as profundidades-Secchi fluct'uan entre 4.5 a 10.5rn.
El lago es abundante en 561 idas di sue1 tos (780 rng I-') debido a que su perdida de agua es principalmente por evaporacihn. Bajo lar condiciones cl irniticas actuales, las concentraciones de s61 idos di sue1 tos esth amen tando. Los cambios cl im5ticos probable- men te afectarian tales concentraciones substanci almen te. Las pobl ac iones de f itoplancton incluyen prineipalmente algas uerdes y azules-verdes aunque la diatamea Stephanodiscus astraea es la doininante durante la estaci6n seca. La biomasa pramedio er 3.0 cj~mn*~.La diversidad (HI ) es a1 rededor de 3 bits- indiv. -' , pero baja durante 10s perhdos de "florecim-iento" (bloom). Porque el 1ago esta' regul armente mezclado bajo de la profundidad de compensaci6n, hay poca diferenciacih vertical de las poblaciones. Algunas diferencias son aparentes entre la flcra de Is73 y la que ha sido descrita par otros investigadores.
Tasas de sucesi6n estacional son bajas. La fauna zooplanctUnica estd dominada por el cop6podo calanoide Boeckella titicacae. La biomasa prornedio del sooplancton es alrededor de 0.89~~-rn-~.Se produc~nvariaciones estacionales en la biornasa y en la proporci6n de
10s distintos estados de 10s ciclos de vida (life history staqes), per0 la sucesibn estacional de las poblacinnes esti virtvalmente ausente. La fauna de peces es pobre en taxa wperiores y estg dominada por un cardurnen end6mico del Cyprinodon genera
Orestias. Par lo menos T9 especies de este g4nero estan presentes. La riqueza aparente de ~specieses baja para un lago tropical de su tamaio.
El ?ago Titicaca cs roderadamente eutr6fica son un proniedio diario de productividad -2 -1 primaria neta de 1.45 gC-rn .dia . La varfaci6n de la produccidn primaria estacionaT -2 . -1 f7uctGa entre aproximadamente 0.75-2.75 gCmm *dla , y es dificil de explicar. La
produccihn neta ests correlacionada con la produccibn por unidad de biomasa a1 nive7 de
saturaci6n de luz, tal que estz implicado el estado de nutrientes. La producci6n de 10s niveTes tr6f icos superiores puede ser gruesarnente estirnada .
El rendimiento ictiol6gico rniximo sostenible puede ser del orden de la$ 50.000
toneladas rn6tricas por ano.
Los contrastes estructurales y funcionales entre lagas temperados y trapicales
estan relacionados can 10s diferentes patranes de variacien ambiental. La estacionali- dad esti suprimida en 10s lagos tropicales, por consiguiente la produccidn anual y la
biomasa no son detectables. A una escala de tiernpo mayor, sin mbarqo, 10s lagos tropi- cales son prabablemente tanto o m5s variables que 10s ternperados. La estratfficaci6n termal d6bi 1 pero persi stente es la que parece causar las diferencias de a50 en a60 en
las condiciones quirnicas de 10s nutrientes. Las altas tasas de evaporaci6n hacen el agua y el balance de s6l idas disuel tos sensi bles a 1as fluctuacjones del cl ima. So-
larnente la fauna ictiol6gica de los lagos tropicales responde a estos moldes con un
increment0 en la diversidad y entre Tos peces como Pas del lago Titicaca las excepciones ABSTRACT
Lake Titicaca is a large (8100 h2,281 rn max. depth), high altitude (3803 m asl.) lake in the bltiplano of the Peruvian and Bolivian Andes at 16's latitude. A series of
21 measurements of basic IimnologScal variables were made in deep water near the village of Capachica, PerG, in 1973. The stations were sampled at approximately hi-weekly in- terval s from January to December. Measurements included temperature prof: 1es, hcchi depth, light penetration, selected chemical parameters, and primary production. Phyto- plankton and zaoplankton samples were collected and enumerated. The data from 1973, together with infarrrtati on col Tected by other investigators, is used to compare the
Iirnnology of Lake Titicaca to temperate and lowland tropical lakes.
Lake Titicaca resembles lowland tropical lakes in almost all aspects of its limnology, despite cool temperatures. Its thermal profiles show a thick epilimnion, re'latively little difference between epilimnetic (rnax. 16*~]and hypolimnetic (~1.1~~)temperatures and a period of more or less complete overturn in the dry season, July to September. The heat budget is near1y balanced during a1 3 seasons, chiefly by evaporation which approaches
2 rneyr-l. Consequently, stratification is weak but persistent and the extent of deep nixing probably varies substantially from year to year. The hypolimnion did not mix be1 ow T 50 m in 7 973. The lake is qui te transparent because the phytoplankton biomass circulates in a thick epilimnion and crustacean grazer biomass is perennially high;
Secchi depths range from 4.5 to 10.5 rn. The lake is high in dissolved s~lids(780 mg -1-I ) because its loss of water is mainly by evaporation. Under present climatic conditions, concentrations of dissolved solids are increasing. Changing climate probably affects such concentrations substantially.
Phytoplanktan populations include mainly greens and blue-greens , a1 though the diatom -2 Stephanodiscus astraea is a dry-season dominant. Mean biomass is 3.0 gC-rn . Diversity -1 (H') is about 3 bitsaindiv. but falls during bloom periods. Because the lake is usually mixed to below the compensation depth, there is little vertical differentiation of populations. Some differences are apparent between the flora of 1973 and that de- scribed by other investigators. Rates of seasonal succession are low. The zooplankton fauna is dominated by the cal anoid cepepod Boeckel 1a titi cacae. Zooplankton bjornass averages about 0.89 qc*rnm2. Seasonal variation in biomass and in the ratio of life history staues occurs, but seasonal successfon of populations is virtually abs~nt. The fish fauna is poor in hiqher taxa and 7s dominated by an endemic flock of the Cyprinodont senus Orestias. At least 19 species of this genus are present. The apparent species richness is low far a tropical lake of its size.
lake Titicaca is moderately eutrophic with a mean daily net prirary production of -2 1 .a5 g C rn day -I. The seasonal variation of primary production ranges from about 0.75 to 2.75 g C . nm2 . day-', and the pattern is difficult to explain. Net production is
correlated with production per unit biomass at light saturation sa nutrient status is im- plicated. Production of higher trephic levels can be roughly estimated. Maximum surtain-
abTe fish yields may be on the order of 50 thousand metric tons per year.
The structura7 and functional contrasts between temperate and tropical lakes are r~latedto different patterns of environmental variation. Seasonality is suppressed in tropical lakes, so annual production and biomass variations are muted. On lanqer time scales, however, tropical lakes are probably as variable or more variable than temperate ones. Weak'but persistent stratification is likely to cause substantial year to year differences in physical and nutrient chemical conditions. High evaporation rates make water and dissolved solids budgets sensitive to fluctuations in climate. Only the fish fauna of tropical 1akes respond to these patterns with increased diversity, though examples
of low fish diversity, as in Titicaca, are comn. INTRODUCTION
Until recently, l imnology has been principally a science of temperate lakes, and work in tropical environments has lagged behind temperate ones in both extent and sophisti- cation. It is important to rectify this lack of informatlion because of the intrinsic jn- terest of tropical lakes and because the different conditions of such habitats may provide valuable cornparati ve data regardi ng broader ecological questions. This general 1i mnologi- cal description of Lake Titicaca was undertaken to characterize a system which is unusual not only because of its tropical latitude but also because of its altitude and size. Lake
Titicaca is subject to a unique combination of slight seasonaljty, low temperature, and great depth and area. The paper is presented in the spirit of comparative natural history
Its aim is to elucidate the similarities and differences behrreen Lake Titicaca and other temperate and tropical lakes in an effort to shed light on the general problems of lacus- trine ecosystems.
Large freshwater tropical 1akes of lower elevations tend to be characterized by several properties which distinguish them from temperate lakes of simjlar area and depth. In stra- tified tropical lakes, epilirnnia tend to be thicker and transparency higher for a given trophic state than in temperate lakes (Talling, 1969). Because seasonality is low, hypolim- netic and epilimnetic temperatures differ little and arc usually above 20°C. Primry pro- duction is usually high and relatively aseasonal (Brylinsky & Mann, 1973). The phytoplank- ton assemblages of tropical lakes are roughly comparable to temperate ones in instantaneous diversity and few forms are restricted to tropical environments. Seasonal succession of the planktonic flora is more or less suppressed by the muted or absent climatic variability.
The zooplankton and possibly other invertebrates also tend to have only normal or lower instantaneous diversity ((Green, 1972a; Hubendick, 1962) in exception to the usual tatftu- dinal gradients of increased diversity in the tropics. Seasonal changes are substantially reduced, giving planktonic faunas a mnotonous character. Fish faunas are frequently characterized by impressive endemic species flocks and high total dli vers ity (Lowe-McConnell ,
1975). Many components of the biota are too poorly known to permit generalization.
Research on tropical lakes has accumulated in well defined phases in all parts of the world. Scattered and fragmentary reports by traveling naturalists began to appear by the late 1800%. Even to the present day, most information from tropical lakes is of this fom, a1 though the variety and sophistication of measurements has increased. For Lake 5 Titicaca, this phase is represented by the work of Agassis (1876), Agassit and Gaman (2876), Neveu-Lemai re (1906), Schindl er (1955), and Qerkosch and LBffler (1961 ) . The second phase general Iy consjsts of short but intensfre expeditionary investigations which concentrate on surveys of physical condition, morphometry, and taxonomic coll ections. Juday's (1915) and
Deevey'r (19571 work on Central American lakes, Ruttner's (1931, 1952) observations in fndonesi a dlrrinq the Geman Sunda expedition, Cunnington ' s (1920) expedition to the East
African Lakes in 1904-7905, and the Percy Sladen Trust Expedition to take Titicaca in 1937 are examoles of this phase. Under the leadership of H. Cary Gilson (1939, 1960, 1g55), the
Percy Sladen Expedition spent six months, from April to September, at the lake. Their observations and collections resulted in twenty reports of which most are taxonomic studies.
The third phase is the collection of a complete series of modern limnological data for an annual cycle or more. Information of this type is available from a few tropical lakes including bakes Victoria (Talling, 1966) and George (Greenwood and Lund, 1973) in East
Africa and Lake Lanao (Frey, 1969; Lewis, 1973a, 1973) in the Philippines. The East African work, especially at Lake George, is notable for the varjety of measurements obtained as part of an International Biological Programme production study. The work on Lake George perhaps represents a fourth phase, since it included experimental investigations as well as descrip- tive 1 imnalogy. Other reasonably extensive studies of tropical lakes include those on East
African saline lakes (Melack and Kilham, 1974; La Barbera and Kilham, 1974; Hecky and Kilharn,
1!373), Lake Kariba (Balon and Coche, ?974), Lake bti tlan (llorris and Sumrfeld, 1969;
Dorris, 1972; Weiss, 1971), Lake Izabal (binson and Nordlie, 2975), and Lake Yolta (Biswas,
1969; Viner, 1970). A rather substantial literature of less complete studies and of more specjatized investioations is now available on a wide variety of tropical fresh waters, but only for African lakes is this information surmnarized in a convenient form (Beadle, 1974).
This paper reports data on the physical and biological properties of Lake Titicaca collected from January through December, 1973. Durina the year. a series of 21 excursions inta the main basin of Lake Titicaca were made at ap~roximatelytwo-week interva7s to measure primary production and chemical conditions and to sample the plankton. Most meas- ure~entswere made in water of 100 meters depth at a station (I in Figure 1 ) 250 meters to the east of Isa:ata IsTand, near the village of Capachica. Department of Puno, Per6.
Because the basin drops auickly to considerable depths at this location, essentially open fake conditions could be sampled with a small befit a short distance from shore. This location was also chosen becausp it is in the same area used by the Percy Sladen Trust
Expedition as a base for its investiqations (Gilson, 1939). Samples for dissolved oxygen 6 and measurements of water temperature and light absorption were usually taken at a deep- - water station (O in Figure 1) about 7 kilometers east of the Isa6ata station in water of
175 to 200 meters depth. The primary production measurements were made at the Isazata
station because of the danger of remaining too long at the deeper station in a small boat.
A three-day cruise of the north end of the lake was made to estimate horizontal hetero-
geneity. No attempt was made to sample the shallow water regions such as Puno Bay, which contains extensive Sci rpus marshes, or Lago Pequeiio (Hui namarca) . These rather isolated basins are somewhat different ecosystems from the deep, steep-sided Lago Grande.
I.', I.',
L--P -.. -. - -"/.. PERU !; BOLIVIA $<-
- I
Figure 1. Map of Lake Titicaca. Sampling stations used in this study are marked by capital letters. Routine work was done at stations I and 0.
7 GENERAL DESCRIPTION OF LAKE TITICACA
Lake Titicaca is located near the narthern end of a large, hish altitude internal drainage basin, the Altiplano, which includes parts of PerG, Bolivia, Chile, and Argentina.
The midpoint of the lake is about 15~45'Slatitude and 69" 25'W longitude, The lake itself is shared between PerG and Bolivia. The lake surface is about 3803 m as1 and has an 2 area of 8100 km . Gf? son (1939, 7964) prepared a morphometrie map of the lake based on soundings from the Percy Sladen Trust Expedition and those of Agassiz (1876) and Neveu- 2 Lemaire (1906). The lake is composed of a large, deep main basin, Lago Grande (6315 km ), 2 2 and two smaller, shallower basins, Lago Pequeiio (1260 kin ) and Puno Bay (525 km ) (Figure
1). The mean depth and maximum depth of the 1 ake are 107 rn and 281 rn, respectively. The total volume of the lake is thus 866 krn" The ehhorel ine length of the lake is 1141 km, including islands. Shoreline development is 4.2. The hypsographic curve of the lake's basin is given in Figure 2. A11 length, area and volume measurements were made with a digitizing table from Newell's (1949) map whose soundings are based on Gilson (1939).
Gilson (1939, 1964) reported somewhat different areas and mean depths. Our area calcula- tion compares very closely to Monheim (1956). The area of the catchment basin is about 8 times the total surface area of the lake.
The climate of the Lake Titicaca region is cool and semi-arid, The weather pattern near the lake at Puno, PerG, for 1973 is presented in Table 1 . Mean monthly temperatures varied from 5.7' C in July to 10.7'~ in January and November; monthly rainfa17 varied from none in June to 238 m in January. Total precipitation for the year was 797 m. The rainy season (months with mre than 100 m of rainfall) usually begins in December and lasts until March but is quite variable. The Cake fiticaca region has a distinct seasonality of temperature. However, the typical daily ranqe in temperature greatly exceeds the variation in ~nthlymeans. Mean daily ranges of temperature varied from 8.8'~ in January to 14.9'~ -1 - in June in 1973. Average wind speeds ranged from 0.E9 rnesec-I in March to 1.3 mosec ~n
August, 1973. Monheim (1956) reported that winds at Lake Titicaca seldom surpass Beaufort
3-4. However, the authors have frequently observed strong afternoon and evening winds up to 10-14 mmsec-' (Reaufort 6-7). Figure 3 summarizes, in the form of a cl imograph, mean monthly weather data from Puno, for the years 1964-1 973. Additional weather data are sumarired by Kessler and Monheim (196R).
The dominant native t~rrestrialvegetation in the vicinity of the lake is locally known AREA ~rn~ 1000 3000 5000 7000 9000 I I I I 20 - 40 - Lago Grande 60 - 80 - I00 -
Figure 2. Hypsographic curve for Lake Titicaca. as puna. This vegetation Cs dominated by 1arge bunch-grasses , especially Stipa, Ee_s_txa, and Catamagrostis species. tower elevations around the lake are extensively cultivated.
Potatoes, oca, quiiiua, barley, and broad beans are the principal crops. Sheep, cattle, llamas and alpacas are grazed on fa1 f ow fields and at higher elevations. The rural pepu- lation is dense in the regions adjacent to the lake. Extensive use is made of lake re- sources including native and introduced fish and totora (Scirpus totora) , which is used for construction and fodder. See hizarraga --et a?. (1965) and Bertholet (1969) for descrip- tions of the local environment, economy and social system and Laba (1977) for a history of the devel opmen t of 1a ke resources.
The geological formation of Fakes related to the modern system could have begun as early as the Miocene (Moon, 1939). Part of an old peneplain was raised and delimited by overthrusts to form an intermontane basin, the A1 tiplano, isolated from ocean drainage. In the latest Pliocene-early Pleistocene, Lake Ballivisn formed as a massive lake that included both the present day Lake Titicaca and Lake Poop6 basins (Newell, 1949). Based on remnant wave cut terraces, Newel l concludes that, during the Pleistocene, Lake Ball ividn reached a level of more than 100 rn above the present level of Lake Titicaca and remained for a time at the 85 rn level. The drop of the lake to form Lake Titicaca near its present level occurred rapidly, as no intermediate terraces are discernabl e. Pl eistocene or Recent re-
- newed block faulting deepened the 1ake floor to its present depth. See James (1971) for a
modern interpretation of Central Andean geology. The isolation of the a1 tiplano lakes from other drainage systems since the PI iocene explains the substantial endemism of the biota
(Brooks, 1950).
The level of Lake Titicaca fluctuates substantially and its water budget has been the
subject of several studies (Monheirn, 1956 ; Kessler and Monheim, t 968; Kessler, 3970, 1974 ;
Cehak and Kessler, 1976). A reliable daily record of the fluctuations at Puno since 1912 is available (Monheim, 1956; Del Castillo, 1977). Annual fluctuations of 0.5 - 1.0 meter
are superimpred on longer term fluctuations which include a record maximum in 1933 of
about 1.2 m above the arbitrary zero mark an the Puno gauge and a minimum in 1943 of -3.7 rn.
In recent decades the record inclvdes levels near the historic maximum in 1963 and a -1 -7 m
minimum in 1970. The average level in 1973 stood at about -0.7 rn; approximately the
historic average. The variance spectrum of the record shows no well-defined cycles other
than the annual one (Cehak and Kessler, 1976). Kessler (1974) attributes changes in JeveT
to variation in rainfall rather than evaporation, since the annual average evaporation,
calculated from water budgets (1957-1 961 ) is rather constant, compared to rainfall
variation and changes in lake level. During this same five-year perfod evaporation account-
ed for 985: of water losses and river outflow via the Rio Desaguadero to Lake Poop6 only 2Y7.
Incoming water was estimated to be 58% from direct precipitation and 425 from stream inflow
(Kessl er and Monheim, 1968).
Given this regime of high evaporation, it is not surprising that the lake is high in
dissolved sol ids (780 mg/l , Gil son, 1964). Gi1 son (1964) gave a water residence time for
Lake Titicaca of approximately 280 yr., but a calculation from the water budget data of
Kessler and Moqheim (1968) and our estimate of volume gives a water residence time of 70 yr.
and a conservative non-volatile constituent residence time af 3,440 yr. Newel 7 (1949) gives
evidence that the lake has been stable near the present level for some time. The separa-
tion of Lago Umayo and Laguna Arapa from Lake Titicaca probably occurred with an & rn drop
in the lake level in relatively recent times. Newel1 states that this shrinking of the
lake was probably caused by the cutting of the basin rim at Desaguadero and observes that
the outlet is not currently being degraded. Monheim (1956) also concluded that the level
of the lake is presently stabilized by the flow regirne of Rio Desaguadem. PHYSICAL AND CHEMICAL LIMNOLOGY
Physical Measurements
Methods
Temperature measurements were made with a thermistor (February-May , September-December) and a water bottle thermometer (January, June-August ) both cal ibrated against a laboratory standard thermometer. Temperature profiles were adjusted to a 11.1 'C hypo1 imnetic minimum temperature to remove small inconsSstencies in the measurements. Heat content was calcula- ted from these temperature data and from the hypsographic curve as the heat required to be last to cool the lake to O'C.
A secchi disk and a homemade photometer, with a sensitivity peak in the green, were used to make rough measurements of light penetration. Light extinction coefficients were then determined from the slope of light penetratjon vs. depth plotted on semi-log paper
(Hutchinson, 1957).
12. IY, 16, 12. IY. 16. 12. I'I. 16. 32. 14. 16. 12. 19. 16. 12. 19. 16.
Figure 4. Temperature profiles from Lake Titi caca , 1473. Thermal Regime
Lake Titicaca is a warm monomictic lake according to Hutchinson's (1057) system of thermal types. During the summer and fall, the lake was stratified with a 40 m thick epilimnion in March deepening to 70 m in June (Fiqure 4). Durinq mid-winter (July 301, the lake became nearly isothermal at 11.1 to 11.2~~.Weak shallow epilirnnia were often present in the least stratified period and on no measurement date was the lake completely isothermal. These data and the chemistry proffles discussed below indicate tbat Lake
Titicaca mixed to at least 100 rn by July 30 and that thorough deep rnixtnw did not occur in
7973. Stronger stratification was re-established in September and continued into the following summer.
During stratification, there was only a small difference between epilimnetic
(12.0 - 15.7'~) and hypalirnnetic (11.1'~) temperatures. At times, multiple-stepped thermoclines occurred in the epilimnion (February 26, March 11, and November 30). This was soon followed by atelomixis (March 27 and December 14), the mixinq of thermally-divergent layers in th~epilimnion without erosion of the main therrnecl ine (L~wisl973a).
Heat Budget
The heat budqet of lakes is im~ortantfor two reasons: one, the pattern of heat qain and loss controls stratificatfon which in turn sets a physical framework for chemical and biological processes in the lake, and two, the heat budget fs intimately related to evaporation rates which affect the geochemistry of the lake and the climate of the lake shore area.
The heat content of Lake Titicaca declined as the lake became isothermal and later increased during restratification (Table 2). Approximately 19,300 cal .crnm2 of heat was
Tost from the February maximum to the July minimum. Because of decreased seasonality from temoerature to equatorial latitudes, Lake Titicaca's total annual heat budget is lower than that of larse temperate lakes such as Baikal , 65,500, Michisan, 52,40D, and Tahee,
34,800 cal.crnw2 (Hutchinson, 1957) but is high compared to that of equatorial Lake Victoria
(0' 05' S), 9.000 - 11,000 cal-cm-' (Tall ing. 1966). The Titicaca budqet resembl er more that of tropical Lake AtitIan (Guatemala, 14' 40' N), 22.1 10 cal .ern-? (Hutchinson, 1957).
Using data from the Capachica weather station, 30 km northeast of Puno. an ~stirnat~ Date Heat content Secchi depth Light extinction-l m coefficient, rn
8 Feb. 168,700 4.5 - 24 Feb. 169,100 4.75 0.17 11 Mar. 165,400 4.5 0.13 27 Mar. 164,800 5.8 0.13 11 Apr. 165,300 6.5 0.10 2 May 165,300 7.4 0.07 18 May - 7.0 0.07 2 June - 8.25 0.07 22 June 7 55,300 8.5 0.07 15 July 151,900 8.75 0.06 30 July 149,800 10.0 0.07 15 bug. 150,300 10.5 0.08 4 Sept. 150,200 10.5 0.05 28 Sept. 1 56,500 8.75 0.05 14 Oct. 153,600 9.25 0.05 30 Oct. 157,100 9.25 0.04 16 Nov. 156,500 7.5 0.04 30 Nev. 1 58,600 6.2 0.06 14 Dec. 162,900 6.0 0.06 28 Dec. 156,900 6.0 0.07 Table 2. Heat content and light penetration parameters in Lake Titlicaca, 1973. was made sf the various components of the heat budget. The principal methods used are summarized in Neumann and Pierson (1966). The basic equation for the heat budget of a water body is
S=R(I-a) - 1 - H- LE+h t where 5 is storage rate, Rt is total incident solar radiation, a is surface albedo, I is net long-wave radiation, H is sensible heat transfer, SE is evaporative heat Toss, and h is a residual term. This equation neglects advective effects due to inflow, outflow, and rainfall. Monthly averages of 5. were calculated from the heat content data by assumina that storage changed at a constant daily rate between measurements of heat content. The average monthly rates were taken to be the number of days at each rate in that month times that rate, summed, and then divided by the number of days in the month. Na correction was made for the first week in February ngr the last three days of December. Rt was calculated by rnultiplyinq the theoretical value of solar radjation from List (1951) times the Y of theoretical insolation r~portedfrom the Puno weather station. The surface albedo was as-
- S R+(3-a) I& -I+ - LE -H A B E E pan Wa
Feb -14 442 651 -788 -267 -58 +6 -22 4.6 5.3 3.0 Mar -103 440 616 -795 -271 -67 -26 -25 4.5 4.2 2.8 Apr +I2 506 617 -783 -269 -63 4-4 .23 4.6 4.4 2.9 May -187 531 548 -376 -300 -69 -1 15 -23 5.1 4.2 3.0 3un -186 478 519 -759 -258 -66 -100 -25 4.4 4.5 2.6 Jul -138 501 513 -747 -372 -79 -14 .25 5.3 4.1 3.4 +12 530 549 -754 -333 -69 +89 -21 5.7 5.4 3.7 Sep +24 511 589 -757 -308 -67 +56 .21 5.2 4.9 3.7 Oct +I97 621 615 -774 -335 -52 +I22 .16 5.7 6.2 3.5 NOV +51 559 620 -777 -369 -55 +73 .15 6.3 6.5 3.8 Dec -96 486 617 -776 -372 -65 +I4 -17 6.3 5.7 4.0
Table 3a. Heat Budget for Lake Titicaca, 1973. Negative values represent loss of heat from the lake. Energy fluxes are given in calmcrn-Z . day-'. B is the Bowen's ratio, and E is the calculated estimate of daily evaporation. Monthly mean wind speed, %. (rn sec-' ), and pan evaporation, E pan, (rrm day*' ), data from the Capachica station are also given.
1. F' -Rout -AV = Residual +Rin -E
+I016 +724 -1900 -31 -375 y -566
2. P -AV = E +in +I016 +724 -31 -375 = 1334
Table 3b. Water Rudget for Lake Titicaca, 1973. Values are in m-year-'. 1: Evaporation term from heat budget. 2: Evaporation estimated as water budget residual ,
Sensible heat transfer is small and a loss throughout the year, with little or no fluctuation. This results from the lake being consistently warmer than the mean air temp- erature for all months. Evaporative lass is high and relatively constant during the year. The pattern of LE only loosely follows Rn (r = .60, p - .05, n = 21), and is not signifi- cantly related to the storage term Cr = .36, p > .05, n = 11). Because of this pattern and the high magnitude of LE, the storage flux results mainly from the small imbalances of LE and Rn.
The storage term is correlated to the sum of source and sink terms, Rn - LE - H (r =
-84, p c .01, n = 11). During some months the residual is high relative to other budget terns. This may be in part the result of some af the simplifying assumptions. Advective heat flux from rivers, advection from precipitation, and the change in heat content per unit surface area caused by the seasonal fluctuation of lake volume are probably each re- sponsi bie for no more than i 5 cal day-' . The residual term is correlated with the storage term (r = .91, p < .01, n = ll), which indicates that, while the pattern of the fluctuation of S is explained by the source and sink terms, the magnitude of the fluctua- tion is not accurately estimated. The residual is probably largely the result of having to use western shore-station weather data rather than accurate average overwater data to estimate the terms. Temperature, hurnidity,annd windspeed are influenced by topography and the land surface's heat budget. Errors or biases may also arise fram applying Jacob's
(1951) evaporation equation to a high elevation lake and from the fact that relatively few temperature profiles were used to estimate S. Residuals in heat budgets can usually be made small only by using longterm averages of parameter estimates (T. Powell , personal com- munication).
Water Budget
A water budget for Lake Titicaca was estimated for the year 1973. The budget is given as a balance of inputs, ouptuts, and change in lake level :
P + R. E - Rout bV = 0 in - - where P is precipitation on the lake surface, Rin is river inflow, E is evaporation, Rout is river outflan, and AY is the net change in lake level. Ground water input and output were assumed to be negligible.
Precipitation was estimated from the average of annual rainfall at three near-shore stations, Capachica, Puno, and JuF i. Kessler and Monheim's (1968) mean estimate of river input and output for the years 1957-1961 was used. Since Kessler and Monheim's extensive summary of rainfall data clearly indicated that the west shore of the lake is drier than the east, their chart was used to adjust the data from our three stations upward to reflect precipitation onto the lake surface as a whole. Likewise, since 1973 had slightly higher rainfall than Kessler and Monheim's average of 5 years, river inflow was adjusted propor- tionately from their mean. The lake level at Puno was measured to have increased by 14 3/4 inches (375 mn) during 1973 (Oswal do Zea R. , personal comuni ca tion). Annual evaporation was calculated from the heat budget values. The monthly evaporation for January was esti- mated assuming the same daily rate in February. In addition, evaporation was estimated fram the residual of the other terms. The water budget is sumnarized in Table 3b.
Evaporation estimated as the budget residual, 1334 m, is much lower than that calculated from the heat budget but only a 1i ttle lower than Monheim and Kessler's water budget estimates. The water budget relies heavily on an adequate estimation of river in- 18 flow and outflow data which are not available to us for 1973, Kessler and Monheim's anal- ysis of the water budget for the period 1957-1961 indicates that variation in annual out- flow via Rio Desaguadero, Rout' was small : ? 11 m. Kessler (1970) also showed a consi- derable discrepancy between 5 year average estimates of evaporation based on a heat budget method (1714 m) and the water budget method (1480 m), a1 though the difference is less than ours.
Light Penetration
Secchi disk depths and light extinction coefficients are shown in Table 2. Secchi depth was 4.5 - 4.75 m in February. As the rainy season ended, clarity increared and the Secchi depth gradually went to a maximum of 10.5 rn in August. With the onset of light spring rains in September, followed by an increase of phytoplankton biomass, Secchj depth decreased to 6 m by December. Because of the sensitivity of the photometer in the green, extinction measurements have only relative value. Extinction ctpeff icients followed a pattern influenced by rainfall during the first half of the year and by phytoplankton biomass during the latter half =
Regression Analysis
Pairwise simple correlations were calculated between most physical measurements ob-
tained in 1973. Two physical parameters, the extinction coefficient and the depth of the
epilimnion, were investigated using multiple regression to determine their statistical re-
lationships to other physical and biological conditions. Details an the method of comput-
ing phytoplankton biomass is given on page 30. The results of the simple correlations
among physical factors are shown in Table 4, (see also Table 17).
The equations for the light extinction coefficient and the depth of mixing are highly significant. Rainfall and phytoplankton biomass account for 75.5% of the variance of light extinction. The two independent variables are not significantly correlated (r = 0.31, p < -05, n = 21). The equation is interpretable: light extfnction increases with turbidity caused by sediment load (brought in by runoff) and by algal biomass. The correlation is somewhat stronger with rainfall than with biomass, which is rather surprising. Apparently, the small fluctuations in biomass combined with high turbidity of inflows during the rainy season cause the higher dependence on rainfall compared to biomass. The authors on one occasion in May observed a sediment plume from the Rio Aamis reaching several km into the lake. The depth of mixing can be predicted empirically by air temperature and wind speed with 76.1% of the variance explained. Air temperature and wind speed are essentially inde- pendent (r = 0.06, p x- .05, n = 21). Pow air temperatures may cool the lake surface, causing convective mixing, and strong winds across the lake create water currents that may cause vertical turbulence. The regression equation gives air temperature as being the pri- mary factor although turbulent mixing is generally considered to be mare important than t hemal convection.
Discussion
Most tropical lakes of intermediate depth appear to have a thermal pattern simi lay to
Lake Titicaca, more or less substantial ty mixing once annually (Talling, 1969). Lake Titi- caca is perhaps a marginal member of the oligomictic class of lakes defined by Hutchinson and toffler (19561, but it is probably best categorized as primarily rnonomictjc with incorn- plete circulation in some years. Apparently no really good example has yet been recorded of oligomixis uncomplicated by density stratification of the deeper waters by electrolites as in the cases of Tanganyi ka and Malawi or by smaf 1 size and great protection as in Bun- yon$ (Talling, 1969; Baxter et al., 1965; Eccles, 1974).
The heat budget of Lake Titicaca is in striking contrast to temperate lakes. Godden
(1976) has sunmarired the information for several such lakes, all of which show very large storage terms on the same order as a,., and smaller and seasonally fluctuating LE. Godden's budget for Lake Tahoe, which is more or less typical of temperate monomicitic lakes, is shown in Figure 6 for comparison with Lake Titicaca. H and LE are both much less variable in Lake Titicaca than in temperate lakes. In Titicaca, sensible heat always represents a loss term, while in temperate lakes it is a source of heat in the summer. Rn and LE are both large and nearly balance each other in every month, hence the storage rate term and total storage remain small. Other tropical lakes wi7l probably be found to exhibit a simi- lar pattern, although it will be surprising if lowland lakes do not usually gain sensible heat. High Rn and LE throughout the year, perhaps combined with generally low mean wind speeds, are probably responsible for the stepped thermoclines during the sumner months and for the persistence of weak stratification during the winter.
Because of the large seasonal changes in R and air temperature and the high specific n heat of water , temperate 1akes 1ag considerably behind the seasonal weather. Tropical lakes, with much less seasonal environments, are always much closer to thermal equil ibrium with existing conditions. In addition, the relatively small total heat storage in tropical 20 I AUG- MOV- FEE- MAY- AUG- UCT JAN APR JULY QCT
Figure 6. Summary of the analytical heat budget for Lake Tahoe. Seasonal averages based on 38 months of data. Courtesy of D.R. Godden (1976), L.O. Myrup, and T.M. Powell. lakes results in weak stratification. Consequently, relatively small changes in the large and nearly balanced values of LE and Rn may easily disrupt or noticeably strengthen strati- fication in a short period of time. The depth of mixing during the winter approach to iso- themy may then be very dependent on the particular weather conditions in a given year.
A1 though tropical lakes are much less seasonal than temperate ones, the pattern of strati- fication may be highly variable from year to year if weather patterns vary much between years. Hence, in some respects, tropical 1akes may be physically less stable systems than temperate ones.
The general characteristics of Lake Titicaca's water budget are clear from Kessler and
Monheim's (1968) data. The great bulk of water (approximately 98Z) is lost by evaporation rather than by outflow via the Rio Desaguadero. Nevertheless, the substantial disparity in 21 both our and Kessl el-' s (1970) calculations between evaporation estimated by water budget and heat budget methods begs further attention.
Chemical Measurements
Methods
Routine chemical measurements were performed on samples taken at Station I including oxygen, a1 ka1 inity, pH, silicate, phosphate, nitrate, calcium, magnesium, chloride, and sulfate. These measurements were made with a Hach DR-EL, whose methods are adopted from those recornended in Standard Methods of Water and Waste Water Analysis (A.P.H.A., 1971)
(Table 5). The approximate accuracy of the methods are indicated in the table. Accuracy was general ly acceptable for a1 1 measurements except ni tra te and phosphate whose concentra - tions were normally too low for reliable results with the Hach kit. One analysis of cal- cium, magnesium, copper, and zinc was performed using atomic absorption techniques by G.
Smith at the University aF California, Davis, on a sample of water concentrated by evapo- ration.
Results
The epilimnion pH was around 8.6 for most of the year and decreased to 8.5 at the time of isothermy. Likewise, the pH of the deeper water was slightly under 8.4 except during the isothermal period when it rose to 8.5. Most of the inorganic carbon available - to photosynthesis is present in the form of bicarbonate, MC03 . Total alkaTinity remained very constant at 120 mg*lml CaC03 throughout the year and did not vary with depth.
The silica data are shown in Figure 7. From January to the end of May, silica concen- trations in the epilimnion were observed to ranqe from 0.49to1.18 rng-l-'. Shortly there- after, concentrations fell to low levels (0.06-0.18 rnq.1-'1. Thrmghout this time (January-
July 15), hypolimnetic silica at 100 meters depth ranged from 1.82- 2.60mg*l-l. As the lake became isothermal at the end of July, the epilirnnetic silica concentration rose to between
0.25 and 0.46 rng.lm', and the concentration in the hypolimnion fell to 0.34 mg-l-' at 100 meters. After this period, epilimnetic silica cancentrations again became lower before
starting a slow build-up. Silica concentrations at 100 m and deeper followed a similar
pattern but increased to considerably higher values during this period. At the end of the year. dissolved silica had risen to 3.7 mg-lwl, itr highest value, at 150 meters depth.
Other reported values in Lake Titicaca are in the 0.5-1.0 rng-l-' range (LGffler, 1960;
Rohrhirsch -et --37.- 1969). The concentrations of sil ica in the epil imnion reported here could be Jimiting to diatom growth during certain seasons of the year (Hughes and Lund, 22 Ext~nction Secchl Daily Hours Air Relative Wind Mixed Coefficient Depth Insolation of Sun Tmp. Humidity Speed Rainfall Si02 Depth
Extinction Caef.
Secchi Depth
Daily Insolation
Air Temp.
Rel. Humidity
Uind Speed
Rainfall
Si02
Mixed Depth
Tabl e 4. Pai rwi se pmduct moment correlation coefficients between selected physical and chemical measurements. One asterisk indicates a coefficient signifi- cantly different from 0 at the .05 level, and two asterisks indicate the -01 1eve1 .
Total Alkalinity was masured titrimtrlcally dth standard sulfuric acid uslng Bromcresol Gwen-Methyl Red indicator. t2 1 Phenolphthalein alkallnlty was absent. The results were expressed in mg CaCO 3 per litpr. rng-l- . Chlorjde was determ$ned by titrattng wjth standard mercuric nitrate to the dl phenylcarbatone endpolnt . Results were ex- pressed as mg chlarlde ton per liter. $5 rng.1-'.
Calcium expressed as mg CaC03 per I1ter was detemined by titratlon at pH 12-13 wl th a standard soTutjon of the sodlum salt of ethylenediamine tetra-acetic acid (EOTA) to the Cal Ver I1 Calcium Indicator endpoJnt (red to blue). ~5 mg-I-'. AagnesIum expressed as mg CaC03 per liter was determined by the difference between calcium hardness and total hardness. The latter was ma~uredby titration at pH 10.1 with standard EOTA (Sbdtum salt) to the Calmaglte endpoint (red to 1 blue). Calmagite and Cal Ver I1 Calcium Indicator are structurally related azo dyes. !I0 mg-1- . Nitrate was determined by a modificatton of the cadmium reductton mth& In whEch the nltrate fomd stoichimtncally is measured by diazotization of sulfanilic acid and coupling with gentisic acid to give a dye uhich is wasuwd colorimtrically. Results Here expressed ar mg nltrogen-nitrate ion per liter. 7.03 mg-1". Otssol ved Oxy~was determined by a modification of the Winkler nethod in which MTI'~, converted to manganese hydroxide, reacts w~thO2 to fonn manganese dioxide. This oxIdIzes iodlde to elemental iodine in acid solution. The iodine is tftrated with a standard reducing solution of phenylarslne oxide (P4O). The results are glven In terns of nq O2 per liter. t.2mq-I-'. @ was determined calorimetrically using a narrow range indicator. Khynml Blue. :.05 pH unit. Ortha-Phosphate was masured colori~tricallyby the amunt of heteropoly blue fomd after treatment with mlybdate and ascorbic acid. Results were given In rng phosphate per 11ter. * .03 m.1-l, Sillca was determined c~lorimetricallyby the helertlpoly blue method whlch entails eliminating mlybdophosphoric acid with oxalic acid prIor to reductton with 1-amino-2-naphthol-a-sulfonic acfd solution. Stllca was reported as mg 510 per liter. 1.05 rng.l-ll. 2 Sulfate was measured turbidlmetrlcally accordfnq to the amount of barium sulfate precjpitated From the water under spec{- 1 fied conditions and reported as rg sulfate per liter. !10 mq.1- .
Tabl e 5. Methods used for chemical measurements and their approximate accuracy.
1962). Diatoms are not a very important component of the plankton except during the autumn and winter months when the dominant phytopl ankter is Stephanodi scus astraea. This species has a definite seasonal succession pattern (Figure 9) which probably influences the season- a1 changes in silicate concentration. Talling (1966) observed that dlatom populations in-
creased markedly in Lake Victoria during times of isothermal mixing throughout the water
column, and that the increase in diatoms depleted the surface concentrations of silica. A
regression analysis of variations in silicate concentration showed a hjghly significant re-
lationshjp to rainfall and a significant relationship ta mixing depth (fable 4). The most
important source of silScate during 1973 appears to have been runoff during the rainy sea-
son, while the recycling during mixing was of less significance. Diatom uptake of the si-
1 icate renewed by winter overturn along with the near absence of diatoms from the plankton
during the summer rainy season may be partly responsible for this pattern. Most of the phosphate determinations were near the 1imi t of sensitivity of the method.
However, re1ativel y high val ues were obse~vedin deeper water. averaging 35 pg-l-' PO4-'?.
The data in Table 6 suggest a pattern similar to that of silica, with a short-lived build-
- - pg~03-~1 -' mg S~O~-I-' Date Shallow Deep Shallow Deep Shall ow Deep
26 Jan 8 Feb 24 Feb 11 Mar 27 Mar 11 Apr 3 May 18 May 2 Jun 22 Jun 15 Jul 30 Jul 15 Aug 4 Sep 28 Sep 1.4 Oct 30 Oct 15 Nov 30 Nov 14 Dec 28 Dec
Table 6. Average shallow (0-30m) and deep water (>30m) concentrations of phosphate, nitrate, and silica in Lake Titicaca, 1973.
24 up in concentration at the tine of mixing. The values obtained are consistent with those
of Rohrhirsch --et a?. (1969) in Lake Titicaca who recorded values from 7.8 to 62 p9=1-' in
surface water.
Concentrations of nitrate measured were also too Pow to be very reliable given the
limited method. Rohrhirsch --et al. were not able to detect nitrate in Titicaca. If only
the averages of very deep measurements (100 m or deeper) are used for the estimate, the
U:P ratio in Lake Titicaca Ss low, approximately 4:1 by weight or a 10:1 atomic ratio.
NO3-N averages 170 ug-1-' at these depths and POq-P averages 38 ~~-1-l.
Average values, based on at least one epil irnnetic and one hypolimnetic sample on 21
measurement dates, for the concentrations of chloride, sul fate, calcium, and magnesi urn
ions, were 260, 282, 66, and 34 rng.1-' , respective1 y. Gilson (1964) reported 250 eg.l-l
of chloride, 246 rng-l-' of sulfate, 65 mg-l-' of calcium, and 35 mpl-' of magnesium.
Analysis for calcium and magnesium by atomic absorption on the sample of water concentrated
by evaporation contained 64.0 mg-lw' of calcium and 36 rng*l-' of magnesium ion. The zinc
concentration was 28 vg-l-', and the copper concentration was at least 2.5 ug.l"'-
Table 7a shows an apparent increase in the concentrations of some major ions between
1937 and 1973. The various past measurements are difficult to compare critically because of differences in methodoTogy, locations from which samples were taken, and number of sam-
ples obtained. The values reported by Gilson (1964) are the mast representative of the main basin, but are based on a single mixed sample of water from various depths except for
C1- for which he reports 20 determinations. Our values represent averages of approximately
70 determinations for each ion but are 1 imited by the accuracy and precision of the Hach
kit. Since the lake's loss of water is presently largely by evaporation, its chemical com-
position may be increasing under present climatic conditions and the apparent increase,
particularly in the conservative anions 1 so4=) could be real .
Using the water budget given by Kessler and Monheim (1968) and the chemistry data for
Lago Pequefio and the Rio Rarnis and Rio Huancan4, given in very sketchy terms by Gilson
(1964), an estimate of the salt balance is possible able 7b). The years 1957-1961 used
by Kessler and Monheim in the estimation of the hydrological budget were during a period when the lake stood near its historic mean level and over which the net change in lake level was a modest 0.42 rn. The agreement rri th the observed increase is surprisingly close for C1-, 1903 %lfr08 1 fr37 1954 9 973 Reveu-hemaire Posnansky Gi1 son hlfffler Authors (1906) (1911) 11964) (1 960)
Table 7a. Major ion concentrations recorded since 1906 in Lake Titicaca. Neveu- ternaire and Posnansky mixed samples, including water from Laqo Pequeh. Loffler's sample was from Puno Bay. Gilson's and the authors' samples were from Lago Grande. All units rng.l-'. Gilsan's data here are far one sample date except for C1- which is a mean of 4 dates (5 depths each date), Standard errors are indicated for anions where multiple determinations are available.
I. tlmconcentration in 1937 was 247 rng.l-' in the main basin and 274 rng*l-l in Lago Pequefio (Gi 1 son, 1964) . 2. Rio Desaguadero flow (Kessler and Monheim, 1968) = 2.5 x lo8 m3 - y-'. 10 ~1-output via Rio Desaguadera = (274 g*m-3)(2.5 x lo8 rn3 y-')= 6.9 x 10 -1 9'YV - 3. CI- input via Cncoming rivers: 3 Rio Ramis flow = 82 rn3*sec-I, Rio Huancane flow = 16 rn -set-' (Kessler & Monheirn, 1968) -1 Rio Ramis [C?-] = 45 mg-I-', Rio Huancane [Cl-] = 161 nq.1 (Gilron, 1964) Estimated mean in~utconcentration:
8 3 -1 Total river inflow (Kessler 8 Monheim, 1968) = 53.9 x 10 m *yr Ct- riverine input: 8 3 (64 9.m3)(53.9 x 10 m Vyr-') = 34.5 x 10" g-yi' - 1 4. Net annual increase in ~1-= Input - Output = 27.6 x 10~'gmyr . 3 Lake volume = 866 krn Annual increase in the concentration of C1-:
-I -1 5. Expected increase in [~l-'11937 - 1973 = (0.32 rng-1 -yr )(36 yrs) = 11.5 rng.l-' Apparent observed increase in [~l-I] 1937 - 1973 = 260 ms*l =' - 2 67 mg.l-' = 13 mq-I-'.
Table 7b. Salt balance calculations for chloride ion.
26 given the crudity of the various data and the short span of time since 1937 for observing - changes. No inflow SO4- data are available, so no check is possible using the second ion.
The general conclusion that the dissolved solids concentration of Lake Titicaca is far out
of equilibrium with present climatic conditions ir fairly strong however, since the water
budget and ion concentrations are unlikely to be in error by so large an amount as com-
pletely negate the fivefold difference between inflow and outflow mass of cI-. The lake's
~1-budget will apparently balance under the present rainfall and geochemical regime when the Lago PequeRo concentration rises to about 1400 mg-1-' . The approach to equil ibriun is
fairly rapid on a geelogical time scale. After 500 years the concentration would rise to
422 mg.1-I.
DissaIved oxygen curves (Figure 83 present a rather complex picture. At 140 meters
and belaw, oxygen concentrations (2.4-4.8 mg*l-' ) are about half that on the surface. Mix-
ing of oxygen down to 100 meters had taken place by the middle of August. At the usual
mid-day temperatures of the surface waters, around 13. ~OC,saturation with oxygen occurs
at about 6.6 rngml-', so that the somewhat higher valuer often observed represent supersat-
uration. Since sampling was done at about mid-day, it js reasonable to suppose that this
supersaturation is a result of photosynthesis. At the end of November, a notable decrease
in oxygen concentration had occurred at all depths, The lowest concentration was 2.4
rng.l-' at 150 meters. This event coincides with a time of increasing phytoplankton biomass
and production following a Spring minimum {Figure 16). Much detritus was observed on mi-
croscope slides prepared from material collected during this season, and it is probable
that the low dissolved oxygen reflects decomposition of organic matter.
The oxygen deficit in the hypolimnion can be roughly related to primary production.
In June, just before overturn, the hypo1 imnetic O2 concentration was about 2.5 mg-1-I -3 (2.5 g-m ) less than the @pilimnetic concentration. If the respiration coefficient de-
rived from metabolizing epilimnetic organic matter is 1.2, about 3.0 grams of oxygen will
be required to mineralize a gram of carbon. Since the hypolimnion will average about 100 -2 -2 rn thick, the total oxygen deficit is about 250 gO -m or 83 gC*m oxidized. This amount 2 of carbon represents about 57 days of average I4t primary production (see below). A1 -
though this simple calculation neglects a number of considerations including the actual
oxygen concentration at great depth, sedimentary storage of carbongand littoral and alloch-
thonous inputs, it strongly suggests that only abut 205 of net annual production is min-
eral ized below the themcline. ;nr+-+ I.?.~.:I :T..-.:.7,3 o ! r.?.3.o.1.?, , I r%rl.rr~z~. I \ R :. :I I :'! ZO.l%,??:' 1?.~,73" 1 \ r Ti! , , 1 \ :stm 8 > i t I \\
Figure 7. Silica profiles from Lake Titicaca, 1973.
r.I ';s;~-:!EI: nvV::frl R'. 0: 'I
Figure 8. Dissolved oxygen profiles from Lake Titicaca, 1973.
28 Qiscussi on
The chemistry data raise a number of interesting points. One is that Lake titicaca, in comon with many other tropical lakes, has a low N:P ratio. Tall ing (1 966) remarks on the same condition in Lake Victoria, and Lewis (1974) reports an extremely low ratio of 0.2 by weight in the epilirnnion of Lake Llano. Of the well studied tropical lakes, only Lake
George, in whish N fixation is important (Horne and Viner, 19711, has a nearly normal ratio of N: P (8:1, Dunn eta1, , 1969). Low N:P ratios may prove to be general features of tropi- cal lakes.
A second interesting feature of the chemical regime is the failure of complete mixing of the lake in 1973, confirming the indications based on temperature profiles. Low concen- trations of oxygen persisted below 100 rn in the profiles from July 15 and September 28
(Figure 8). On August 15 and September 28, the 100 rn measurements approach saturation, but the deficit at 170 rn was still about 2 rng.l-' on September 28 and increased thereafter. In
1974, by contrast, Wayne Wurtsbaugh (personal comunication) observed saturated values at
150 m during isothermy.
Finally, the hjgh evaporation and small outflow make the lake" dissolved solid con- centration sensitive to cl imatic fluctuations. The climatic fluctuations of the Pleirto- cene particularly may have caused several-fold changes in salt concentration, even without major lake-level changes. Such long-term instabilities in chemical segjme are well known for African 1akes (Livingstone, 1975). BIOLOGICAL LIMNOLOGY
Phytoplanktan Populations
Methods
Phytoplankton were collected from the same Van Dorn bottle samples as the production and chemistry samples. Usual 1y nine samples were obtained from the euphotic zone and samples were collected from depths up to 100 m on some occasions. 125 ml samples were pre- served with Lugol's fixative and fifty ml were filtered onta 0.45 p millipore filters which were mounted on slides using the method of Dozier and Richerson (1975). Identifications were made using settled material and a Wild inverted microscope to supplement the slide mounts. The slides were optically good, but there was some cell distortion and the pigments were severely bleached. Unfortunately, no Iiving material could be exarni ned.
Organisms on the slides were enumerated by counting from 50 to 100 randomized fields at 1200 x using phase contrast, representing a volume of about 0.05 - 0.7 ml. Except in very sparse samples from below the euphotic zone, a minimum of 250-300 cells were counted.
Particularly when the large colonies of Anabaena sphaerica were abundant, well over 1000 cells were frequently enumerated. Biovolume and bioarea of the phytoplankton populations were determined by approximating each species by one or a few simple geametrical forms. For each species, a sample of cell sizes was assembled by measuring a number of cells (up to about 100 for the dominant organisms) from different experiments, and computing an average volume. Care was taken to ensure that any size differences between dates did not unduly affect the average, but no major changes in the cell size of species from time to time was observed.
The carbon content of the biomass was estimated by three methods. Multiplication of biovolume by 0.1 gives the conventional linear approximation of carbon content. In addi- tion, the regression equation of Mull in, Sloan and Eppley (1966) was used to estimate car- bon content for individual species. the equation used is a fractional power function,
LoglO C = 0.76 LoglO V - 0.29 where C is biomass per cell in picograms carbon and V is cell vatume in cubic micrometers.
The third estimate of biomass was made using the relationship reported by Mullin, Sloan and
Eppley (1966) between cell surface area and carbon content: 0.18 times cell surface area in 2 - v gives cell carbon in picograms. A1 though Mu1 1in, Sloan and Eppl ey's regression equations are based on marine diatoms,
+ the power function equation is most Ii kely a better estimate of carbon biomass than the linear biovolum based equation for the Titicaca species as well. The power function
equation takes account of the fact that smaller cells have smaller or absent vacuoles and
hence a greater proportion of carbon biomass per unit volume. The averall average biomass
estimates compare fairly well, but there are systematic differences. The area based esti-
mate is about 10% higher than that based on the volume power function equation. The linear
volume estimate is much lower than the other two in the sumr when cells are small, but
equals or exceeds them in the winter when large cells predominate. The estimate used for
further calculations and in graphs is that derived from the Mullin, Sloan and Eppley volume
equation unless otherwise noted.
Community patterns in the phytoplankton were examined by several methods. The diver-
sity of the epil imnetic assemblage (down to 30 m) was calculated for each measurement date.
The index employed is the information theoretic statistic, calculated using the unbiased estimator (Pie1ou , 1969), as we1 1 as the conventional Shannon-Weaver fomul a. The differ-
ences between these estimators is quite small for these data. Diversity spectra, an index
of the spatial heterogeneity of the assernbl age (Margalef , 1969), were computed over depth
by unaveraged and averaged methods. The unaveraged method begins with a given sample (the
surface), for which the ordinary Shannon-Weaver index is calculated. Then the data of
succeeding depths are added to the sample and the index recalculated. The average plots
first the average index of all single samples, then of all adjacent pairs, then triplets,
until the whole set is combined to calculate the same final value as the unaveragd
method. If much between-habitat differences in assemblages exists, initial diversity will
be low, but will increase rapidly as samples are added. Conversely, relatively flat
diversity spectra indicate less between-habi tat differences.
Two different succession indfces were computed to quantify seasonal changes in com-
munity composition. One index, first used by Armstrong (1969), is based on the changing
contribution of species to the Shannon-Weaver index diversity, and can be interpreted as a
fractional rate of change of information:
where Sab is the daily rate of movement of the cornunity through diversity space, (b-a) is the time interval in days,and fia and fIb are the fractional contributions of species i to ,- diversity on the days of succeeding measurements such that
-- log f. = X- x- H - where xi is the numbers or biomass of species i at date a or b, X - is the total numbers or - biomass on the corresponding date, and H- Ss the information theoreti c measure of diversity en that date.
The second index was developed by Jassby and Goldman (1974a) and is the rate of change of biomass compos iti on :
where
and bi - is the biomass of the i th species at a particular time, and a-b is the time inter- val between two separate measurements. S . is analogous to the Amtrong index, but does J not weight a species contribution to the change jn community composition by a diversity
function.
Results
The comner and more easily identified phytoplankton species occurring in Lake Titi-
caca during 1973 are listed in Table 8. Also listed jn the table are some comparissns with
earlier reports of a7gal occurrences in Lake Titicaca based on the diatom determinations of
Frenguell i (1939), the net plankton col 1ections of the Percy Sladen Trust Expedition
(Tutin, 1940), and Thornasson's (1956) list from Puno Bay. The patterns of biomass change
of dominant and comnon organisms are also shown in Figure 9.
The species composition in 1973 was substantially different from 1937 as judged from
the l argely minter-time col 1 ections of the Percy 51 aden Trust Expedition. For example, the
overwhelrni ng dominant of 1937, Botryococcus Brauni i, was probably absent in 1973, and
Stephanodiscus astraea reported as rare from the 1937 material, was the dominant during the
winter of 1973. U6no (1967) also reported that .R.- Braunii dominated the phytoplankton - . Species Abundance 1973 Other Occurrences rCI Cyanophyta 1. Lyngbya vacuol ifera Skuja Dominant September to None reported from Titi caca . November. (Tutin, 1940: L n b a aestuari i (Mert*i~mann in Lake Poop6.) 2. Anabaena sphaerica var. Co-dominant January - None reported. attenuata Bharadwaja February. Dominant November - December. 3. Anabaena sp. Uncommon. Thomasson (1956) : Ana baena sp. in Puno Bay. 4. Nodul aria Harveyana (Thw. ) Comon. January - Tutin (7940): fairly Thuret March, November - frequent. December. 5. Gloeothece incerta Skuja Perennial. Common, None reported. but not abundant. 6. G1oeocapsa punctata Naegel i Comon, June - August None reported.
1. Elakatothrix cf. viridis Corrrmon, never abundant. None reported. (Snow) Printz 2. Ulothrix subtil issima Perennial , dominant or Tutin (1940): comnn. Rabenhorst co-domi nan t . 3. Schroederia setigera Common, never abundant, None reported. (Schroed.) Lemmermann 4. Pediastrum du lex var. Rare. Thomasson (1956): in Puno cl athraturn &am) Bay.
5. Pedi astrum cf. Kawra is kyi Rare. Tutin (1940): present. Schmi dl e 6. Pediastrum Boryanum (Turpin) Rare. Tutin (1940) : uncorrrmon . Meneghini Thomasson (1 956) : in Puno Bay. 7. Coet astrum cf. mi croporurn Rare. Tutin (1940) : uncommon. Naegel i Thomassan (1 956) : in Puno Bay. 8. Oocystis Borgei Snow Perenni 91 , common to Tutin (1940) : Oocystis co-dominant. gigas Archer var. Bor ei Lemmermann , (synonym-f . Thomasson (1956) : in Puno Bay. 9. Ankis trodesmus fa1catus Perennial, common, Tutin (1940) and Thornasson var. acicul ari s (A.n) never abundant . (1956): in Puno Bay. G.S. West 10. Closteriopsis longissima Perennial, uncommon. Tutin (1940): var. troplca W. and G.S. West Anki s trodesmus Ion issima (Lemm.1 WiIle, & 11. Selenastrum minutum Comon, but not None reported. (Naeg. ) Col 1 ins abundant, June - Sept 12. Mou eotia cf. viridis Cornon to subdominant Jutin (1940) : Mougeotia sp. hittrock except January; (Sterile material only) October - December. Chloraphyta (continued) 13, Closterium sp. Perenni a1 , moderately Tutin (1940): Closterium abundant. acerosum (Shrank) Ehrenb, from wetted mud. Thomasson (1956): 3 species in Puno Bay.
14. Staurastrum gracile Ral fs Perennial, cormnon. Tutin (1940) : Staurastrum-. . paradoxurn Meyer, synonm QY similar tax. ra~ile sccordi to Skui?o- rnasson 18956) : taurartrum sp. in Puno Bay. 15. Eotr ococcus Brauni i Kitsing Very rare or absent Tutin (1940): heavily ddoubtful col any observed) dominant, June and July. Thomasson (1956): lin Puno Bay.
Chrysophy ta 1. tri bonema ambi guum Skuja Uncommon, August to Tutin (1 9401 : Tribonema sp. December. from wetted mud, Capachica stream.
Baci 1lariophyta 1. Cyclotella Whiniana Common, May - June Frenguelli (1939). Tho- Kijtzing masson (1956): in Puno Bay, also Cyclotella sp. 2. Cyclotella stelligera Corrmon, May - June Frenguelli (1939). C1. & Grunow 3. Stephanodiscus astraea Cumon to dominant, Frenguelli (1939). Tutin var. minutula (m March - September. (7940): rare plankter. Grunow 4. Fragilaria sp. Rare. Frenguef 1 i (1939) : several species from Puno Bay. 5. Achnanthes lanceolata var, Rare. None reported. dubia Grunow 6. Cacconei s 1acentul a var. Rare. Frenguel 1i (1939). mx&JmF--. - 7. Navicula radiosa Kiltzing Rare. Frenguel 1i (1939): in Puno Bay and Lago PequeEo. 8. Amphiprora-- alata Kijtzing Raw. None reported. 9. Epithemia argus Kutzing Rare. Frenguelli (1939): in Puno Bay. Ryrrophyta 1. Wnodinium sp. Perennial, common. None reported.
2. Peridiniurn spp. Raw -e. Tutin (1940): fairly (at least one larse and common deep plankter. Thomasson (1956) : Peri - --dinium ~iliei and m- --dinium sp. in Puno Bay.
Table 8. Planktonic algal species identified from 1973 Lake Titicaca samples and thei r re1 ative numerical abundance. Approximately 200 ramp1 es were examined. Previous reports of occurrence in the Titicaca plankton and nearby habitats are also given. Anoboena sphoerica
120 Ulolhrix subtilissima EO O 80 60 - g? 40- 2" 20- Y+-. m 7 40- Lyngbya vocuolifera 20 L
60b Maugeatio cf. viridis 1 d
- Stephanodiscus as traea - -
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
- Closterium sp - $ - - 2E 5- - - -A-- * - 20 15- Nodularra Morveyono -- - Staurestrum grocl le 7
- 5 - - A=1 15 Floeocapso punctato - 10 - - 5 - - 1 1 1 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Mev Dec
Figure 9. Patterns of mean epilimnetic (top 30 m) biomass fluctuation of the more important species of planktonic algae in Lake Titicaca, 1973. Vertical bars are +I S. E. based on enumeration error and vertical heterogeneity. Note scale differences between the two sets of species. during April, 1961. Some important species are cornon to the two years, however, includinq
lllothrix subtil issima, Oocystis Borgei , Closteriopsis lon~issima,Mougeotia sp., and Staurastrum graci 1e.
The pattern of diversity of the epil imnetic cornunity during 1973 is shown in Figure
10. Divers; ty is low during the periods of dominance of Anabaena sphaerica and Ulothrix
subtilissirna in the late spring and early sumer period. Values are rather constant during
the fa1 1 and winter, though there is a noticeable decline during the period of deep mixing.
The annual pattern is most pronounced for cell numbers because of the strong effect of the
numerous small A. sphaerica cells during the summer.
oL 1 m a I r m r r r r 1 1 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure TO. Variation of epf 1 imnetic phytoplankton diversity. 1973.
Few signj ficant correlations were found between diversity stat isti cs and other vari - ables. Some significant relationships were found with weather variables, reflecting the surrmer low and fall -winter maximum of diverzity. For example, diversity based on numbers has a correlation with air temperature of r = -.66 (p < -01, n = 21). Correlations with factors of presurnabl y more direct causal significance 1 ike nutrient concentration. mixing depth and primary production statistics are generally weak. Table 9 shows some 05 these correlations. The expected negative relationship to production statistics and biomass is
36 present but is significant only for three of the four comparisons with Mullin, Sloan and
Eppley biomass estimates. The major variation in diversity was caused by the midsumer dominance of the flora by Anabaena sphaerica and Ulethrix subtil issirna-, but the causes of this dominance pattern appear to be obscure. The most that can be said is that there is a weak but generally detectable tendency for high diversity to be associated with more 07 iga-
tmphic periods. This is the expected re1 ationshi p between diversity and a1 igotrophy
according to Margalef (1968).
ZP ZP n n Hours of Dajly Secchi Hl xed (day -'I (hr. 'I) En B10volm Eloarea Btmass PI5 P/B/R, C Sun Insnlat~on Depth Ralnfall SfO, Depth
Table 9. Product-mment correlations beweween the Shannon-Weaver divers1 ty Indlces based on eel 1 rider and bimss and selected bloloqical and phys~calvariables. One asterlsk ~ndicatera sign1 ficant dif ferencr fmm 0 dt the .05 level, and wo asterirks Indicat~the -01 level. Symbols are d~srrlbed in th* text.
The diversity spectra are fairly flat on almost all dates (Figure 11). This confirms
the idea gained from inspection of the ran biomass data that relatively little vertical
structure is present in the phytoplankton assemblage (Table 17). Individual species never
show marked differences with depth in the euphotic zone.
The succession rate patterns in Lake Titicaca are shown in Figure 12. The substitu-
tion of numbers for biomass causes some a1 teration of patterns and the difference between
the two indices is substantial, although all curves have some features in comon. Table 10
shows the correlation between indices. In the early part of the year, patterns differ sub-
stantially between the Amstrong and Jassby-Goldman indices, while later in the year both
indices show a peak in August, a minimum in September, and a second peak in October or
November, The August peak is associated with the drop in production at the beginning of
August and the change in dominant populations from Stephanodt scus astraea and Mougeoti a- cf.
viridis to Lyngbya- vacuoll ifera. The second peak is associated with the decl fne of -L. vacuolifera and the bloom of Anabaena sphaerica. The generally high succession rates early
in the year, evident in a17 analyses except the Jassby-Goldman biomass index, are associated
with blooms and declines of Ulothrix swbtilissima and Anabaena sphaerica. The low or de-
cl injnq rate through the fall to mid-winter period is associated with smother and more
modest changes in the population of most species. On the whole, the Jassby-Gwldman index
is more variable, while the log-functional nature of Amstrong's index produces a lower- variance pattern .
- - I I-,4----- i't. -1.: . . 7- .1- 'I . - -- *-*.-- -+.. a-, _...-- -
Fiqure 11 . Diversity spectra, showing how diversi ty changes as sample size is increased by adding successive depths beginning at the surface (unaveraged method). Jassby-Go1 dman Index Armstrong Index
Biomass Number 'Biomass Number
Jassby-Go1 dman Index
Biomass 1 .OO
Number .41 1.00
Arms t rong Endex
Biomass .35 -38 Number .22 .34
Table 10. Product moment correlation coefficients of the different succession indices Double asterisk indicates significant difference from zero at the .O1 level of significance. The critical value for the -05 level is .44.
=' JAN FED MPR &PA MAY JUN JUL RUG SEP 3CT NW DEC JAN FER MbR APR MAY JUN Jut AUG SEP OCT W3.I Mt
Figure 12. Succession rate per day of the take Titicaca phytoplankton assemblage, 1973. Upper graphs in each case are Jassby-Goldman index and the lower are the Armstrong index. Discussion
The phytoplankton association of Lake Titicaca is not strikingly different from other tropical lakes, or even from temperate lakes far that matter. The species which make up the assemblage are widespread outside Lake Titicaca and neither this lake nor other tropical lakes, except perhaps Lake Tanganyika (Cunnington, '1920), appear to be characterized by a highly specialized tropical flora. Nor is the instantaneous diversity of the phytoplankton assemblage outside the range comnly encountered in temperate floras.
Some features of the Ti ticaca flora do reflect its tropical environment. The rela- tively deep epilimnion of the lake (almost always deeper than the compensation depth) pre- vents systematic vertical differentiation of the plankton habitat. In m5t temperate lakes as transparent as Lake Titicaca, some production would occur below the top of the thermo- cline, permitting a relatively stable vertical mu1 tip1ication of habitats. The complex, sometimes atelomictic, structure of the Titicaca epilimnion is not sufficiently stable to produce marked population differentiation within the euphotic zone, a1 though it might help diversify the assemblage by the non-equi 1i brium rnechanl'sms proposed by Hutchinson (I961 ) and Richersan , Amtstrong and Go1 dman (1970).
The sharp differences between the flora observed by the Percy Sl aden Trust Expedition
(Tutin, 1940) and our study in 1973 could have resulted from either a long-term secular trend in lake conditions or from the year-to-year variation of conditions resulting from weak stratification. As discussed earlier, the latter hypothesis is quite plausible for
tropical lakes in general, a1 though more extensive data from tropical lakes than is
presently available will be required to provide a definitive answer. Lake levels were very
similar in 1937 and 1973, although 1937 was in the middle of a 10 year falling and 1973 at the beginning of a 6 year rising trend.
The seasonal changes in the phytoplankton assemblage are marked and, by some indices,
the rate of change is somewhat variable seasonally. Nevertheless, insofar as quantitative comparisons with other 1 akes are possible using previous calculations of Wi11 iams and
Goldrnan(1975), Titicaca's succession rate is generally low and without seasonal differences
compared to typical temperate and arctic lakes. However, it appears to have a higher and
mere seasonal ly variable rate of succession than Lake Victoria, and more closely resembles
Lake Tahoe, another deep mnomictic lake. Williams and Goldman's (797'5) estimate of Average Re1 . % Hum. Average Theo- retical 07.- 13.7 Wind Precipi- Air Temp .(*C) 19 hrs. Speed Insola- tation Evaporation -Month -Max. Lfl. 9; (m/sec) tion (mm) Piche (m) Tank (mn) Jan. 4 0 Feb. 4 9 Mar. 5 2 dpr. 6 8 May 8 7 June 79 July 81 bug - 7 7 Sept. 6 5 Oct. 7 1 Nov . 6 1 Dec. -5 2 MEAN: 65 TOTAL:
Tab1 e 7. Weather summary for Puno, per;, 1973.
Lake Titicaca: Puno, Per6 Altitude 3825 m Mean air temperalure 8.5-C Mean onnuol precipita!ian 630 mm Meon annual evoporot ion 1655 mm (PI, 1995 rnm (TI
i?
Evaporation (PI P".\ ,.f=h..n,." x.\
Figure 3. Climograph for Puno. The data plotted represents the monthly averages for the ten-year period 1964-1 973 (1965-1 973 for evaporation data), Evaporation given for Tank method (T) and Piche evaporometer (P). All weather data courtesy of Servi cio Naci onal de Meteorologfa e Hidralogfa , Puno.