with te equal to 5 to 10 years potentially provide a good record of interannual to decadal fluctuations (their respective HF band). They should also exhibit negligible lag with respect to secular (LF) climatic variations. Large, closed lakes with te greater than 50 years can provide good coverage of decade-to-century (HF) variations. Whenever a response in the HF band is considered, however, allowance needs to be made for a phase shift of -p/2 with respect to the climate signal.
Time series of lake levels have traditionally been compiled from observational, historical, or paleolimnological (geological) data (Street-Perrott and Harrison, 1985). Few lakes, even in developed countries, possess instrumental (gauge-board) records extending back before A.D. 1875. In some cases, however, historical observations of the emergence or drowning of specific landmarks permit lake-level time series to be extended for several centuries further back, at least in outline (de Terra and Hutchinson, 1934; Street-Perrott and Harrison, 1985; O'Hara, 1993).
In data-poor regions, or for the period before the start of reliable historical observations, high-resolution paleolimnological studies provide qualitative and, in some cases, quantitative information about past variations in lake depth and area, based on direct dating of former shorelines, or analyses of stratigraphical, geochemical, and paleoecological data from sediment cores and surface exposures (Street-Perrott and Harrison, 1985). The time resolution of these studies, however, is dependent on the dating framework. The most precise data should in principle be obtainable from lakes with finely laminated mud, such as Lake Turkana, Kenya (Halfman and Johnson, 1988); however, these are generally rare in the tropics.
Recent studies at the Mullard Space Science Laboratory, University College London, have established the feasibility of measuring variations in lake level by satellite radar altimetry (Mason et al., 1984, 1990) and variations in lake area with imaging instruments such as the NOAA Advanced Very High Resolution Radiometer (AVHRR) or the Earth Resources Satellite-1 (ERS-1) Along-Track Scanning Radiometer (Harris and Mason, 1989; Mason et al., 1990; Harris et al., 1992). The overall accuracy is expected to be around ±0.25 m for water level and 1 percent for surface area (Mason et al., 1990). Figure 2 shows the variations in area of the closed Lake Abiyata, Ethiopia (7°N) (te ca. 4 to 9 years) between 1985 and 1991, derived from AVHRR images (Harris et al., 1992).
Lake Titicaca (16°S) is a large, open lake, 8,100 km2 in area, situated at an altitude of 3,812 m above sea level in the Altiplano of Peru and Bolivia. The equilibrium response time te of the lake is probably less than 5 years. It is fed mainly by tropical summer precipitation. Wet (dry) summers in the altiplano are characterized by a poleward (equatorward) displacement and weakening (strengthening) of the mid-latitude upper westerlies over South America (Kessler, 1974). High water levels were observed in the 1920s, 1930s, 1950s, 1960s, late 1970s, and late 1980s, with low levels in the 1910s and 1940s (very marked) and around 1970 (Figure 3) (Kessler, 1974; Künzel and Kessler, 1986; Martin et al., 1993).
The time series of the yearly rise (maximum minus previous minimum level) of Lake Titicaca between 1915 and 1981 was analyzed by Künzel and Kessler (1986) using maximum-entropy spectral analysis. They identified sig-