Earth Surface Processes and Landforms Earth Surf. Process. Landforms 26, 63-76 (2001) EFFECTS OF HISTORICAL LAND USE ON SEDIMENT YIELD FROM A LACUSTRINE WATERSHED IN CENTRAL CHILE MARCO CISTERNAS*, ALBERTO ARANEDA, PATRICIA MARTÍNEZ AND SERGIO PEREZ Centro EULA-Chile, Universidad de Conception, Chile Received 28 September 1999; Revised 30 May 2000; Accepted 24 July 2000 ABSTRACT Sediment yield in the San Pedro Lake watershed, inferred from sedimentation in the lake, can be related to land use changes shown on aerial photographs taken during the period 1943-1994. In this watershed, which covers 4-5 km2 of mountainous terrain in San Pedro County, central Chile, the area of native forest species decreased from 70 per cent in 1943 to 13 per cent in 1994. During this same period, the area of pine plantations increased from 4 to 46 per cent. To study effects of these changes, we took a core from the centre of the lake and estimated sedimentation rates by 2 °Pb dating, which we checked with Cs and pine pollen. The results showthat sedimentation rate ranged from 5 mg cm a in the late 1800s to 60 mg cm a-1 in the late 1960s. These rates, together with assumptions about the production and delivery of the sediment, give corresponding figures for sediment yields with maximum values close to 1 t ha-1 a-1. Sediment yield between 1955 and 1994 closely tracks the total land use change that can be detected, irrespective of land use type, on sets of aerial photographs taken four to 18 years apart. However, this measure of land use change, while convenient and successful as a predictor of historical erosion, may be unreliable because it probably excludes many changes that occurred in long intervals between successive photographs. Copyright © 2001 John Wiley & Sons, Ltd. KEY WORDS: lake sediments; sediment dating; sediment yield; land use; GIS; Chile INTRODUCTION In the history of environmental disturbance by people, the greatest impact has been made over the last two centuries. The technical and industrial revolution has significantly disturbed the planet, not only its ecosystems but also its soils, streams and lakes. Nevertheless, most environmental studies are concerned with ongoing damage. This approach has a limited perspective since it provides only one snapshot of a system that has a time scale of hundreds of years of human disturbance and thousands of years of natural post-glacial change. Over the last decade, physical geographers and environmental researches have valued the role of lake sediments as 'historical recorders' of environmental change (Dearing, 1991; Smol, 1992; Boer, 1994; Dixit et al, 1995; Millspaugh and Whitlock, 1995; Auer et al, 1996; Spliethoff and Hemond, 1996; Van der Post et al, 1997). Slowly deposited lacustrine sediments can record both natural, baseline conditions and subsequent human disturbance. In studies that use the sediments of a lake to gauge human disturbance in the surrounding watershed, the 210Pb isotope is commonly employed for estimating the sedimentation rate in the lake, which in turn provides an estimate of the sediment yield from the watershed (Battarbee et al, 1985; Dearing, 1991; Appleby and Oldfield, 1992; Boer, 1994; Van der Post et al., 1997). In this paper we quantify the sediment yield in a small disturbed lacustrine watershed by dating lake-bottom deposits with 210Pb. We check the dates with 137Cs derived from atomic-bomb testing that peaked and ended in 1963, and with pollen from pine trees introduced in the 1880s. We find that changes in sediment yield correlate with a simple but imperfect measure of land use change detected on air photos from the past 50 years. * Correspondence to: M. Cisternas, Centro EULA-Chile, Universidad de Concepción, Casilla 160-C, Concepción, Chile Contract/grant sponsor: FONDECYT; contract/grant number: 1980529 Contract/grant sponsor: DIUC; contract/grant number: 96.310.021-1.1 Copyright © 2001 John Wiley & Sons, Ltd. 64 M. CISTERNAS ET AL. STUDY AREA Chica de San Pedro Lake (36°51'S, 73 °05'W) is located in San Pedro County, central Chile, beside the Biobio floodplain close to the Pacific Ocean (Figure 1). The lake is bordered by mountains composed of metamorphic rocks except to the northwest, where the lake is impounded by fluvial sediments of basaltic composition (Aguirre et al, 1972). The drainage basin of the lake covers 4-5 km2. The lake has a surface area of 0-87km2, a maximum depth of 18 m and an average depth of 10-3 m (Parra et al, 1976; Parra 1989). The floor of Chica de San Pedro Lake has steep slopes on its west, south and east sides (Figure 1). The northern side is less abrupt. The central part of the lake floor is nearly flat. These bathymetric characteristics reflect the geomorphological evolution of the lake: soon after the last glaciation, a ravine was blocked by sediment of the Biobio River, and the resulting lake has been largely filled, in Holocene time, with sediments derived from Chica de San Pedro watershed (Cisternas et al, 1997). Many historical human activities have affected the lake and its watershed: clear-cutting of the native forest, raising of wheat, introduction of exotic trees, and urbanization (Cisternas et al, 1999). In the past 50 years, some of this land use has been influenced by economic policies of the Chilean government. Replacement of native forest by pine plantations, which began in the late 1800s, accelerated in the 1930s under a national economic policy to develop the forestry in central Chile (Morales, 1989; Unda and Ravera, 1994). Such plantations, many of which replaced native forests soon after the native trees were logged, were established without special preparation of the soils or slopes (Morales, 1989). The time of these changes includes the time recorded by our core (see results below and Figure 2) and the period of record for precipitation in nearby San Pedro County (Figure 3A). In the past 50 years, the annual precipitation was in the ranged of 1-2 m a-1 and averaged 1-55 m a-1. The annual precipitation fluctuated throughout this time without overall trends longer than a few years. Figure 1. Location of study area, bathymetric map and coring site Copyright © 2001 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 26, 63-76 (2001) LAND USE AND SEDIMENT YIELD IN CHILE 65 METHODS Sampling and laboratory analysis A core sample 50 cm long and 6 cm in diameter was obtained in 1996 by scuba diving from the central and deepest part of the Chica de San Pedro Lake (Figure 1). First, the core was X-rayed to check for biological or physical disturbance, a lack of which simplifies the isotope and pollen analysis. Then, the core was extruded vertically and cut into 1 cm slices. Each such slice was analysed, by standard methods, for dry density (drying at 105 °C) and organic carbon (loss on ignition at 550 °C). The 210Pb (half-life: 22-3 years) activity of each slice was calculated from the measured activity of its first daughter isotope, 210Po (half-life: 1384 days), by means of alpha spectrometry (Hasanen, 1977). Supported Pb (the amount in equilibrium with radium in the sediment matrix) was estimated from the Po activity of the three lowest slices; these slices showed nearly identical activities. To check the 210Pb dates, the samples were also analysed for 137Cs. Atmospheric testing of nuclear weapons in the 1950s and early 1960s produced this isotope, which has a half-life of 32 years. It reached its maximum atmospheric activity in 1963, just before signing of the test-ban treaty (Wan et al, 1987; Robbins et al, 1990). To measure 137Cs that became incorporated in the samples, we used a gamma-ray spectrometer with a solid state Ge(Li) detector (Pifiones and Tomicic, 1995). Pine pollen was analysed to provide a further check on the 210Pb chronology. Chile lacked pine trees until Pinus radiata D. Don was introduced there near the close of the 19th century. In San Pedro County, a coal mining company began widespread planting of P. radiata in 1885 (Aztorquiza, 1929; Contesse, 1987; Millán and Carrasco, 1993; Donoso and Lara, 1996). Later, in the 1930s, the plantations were spread further, throughout south-central Chile, as a policy of the national government (Contesse, 1987). Because this history makes pine pollen at Chica de San Pedro Lake no older than the 1880s, the lake's deposits of about that age should contain the deepest P. radiata pollen. Accordingly, we looked for pine pollen in 1 g of dry sediment from each slice of the core. To separate the pollen from the sediment matrix we used standard palynological techniques (Faegri and Iversen, 1975; Dupré, 1992), and to measure concentrations of the pollen, in grains per gram, we used the methods of Anderson (1974) and Kempt et al. (1974). Calculation of sedimentation rates and sediment yield Ages and sedimentation rates were determined from 210Pb activity by means of the CRS (Constant Rate of Supply) model. In this model, the supply of 210Pb from the atmosphere does not vary with time (Goldberg, 1963; Appleby and Oldfield, 1978, 1992). To estimate sediment yield from the watershed, we considered a method proposed by Dearing (Dearing et al, 1987; Dearing, 1991). In its simplest form, this method is based on three assumptions that are not necessarily valid for Chica de San Pedro Lake: (1) the sedimentation rate does not vary with location in the lake; (2) all the sediment supplied by the watershed is deposited in the lake, no sediment is stored in transit; (3) all the sediment deposited in the lake, organic as well as inorganic, comes from in the watershed, none of the sediment is produced within the lake itself. We applied this method in two ways: with and without a correction intended to obviate assumption (3). To make the correction, we subtracted the organic fraction measured by loss on ignition, to obtain the yield of inorganic sediment only. We also calculated a total sediment yield that includes organic sediment, whatever its source. Measurement of land use changes San Pedro watershed land use patterns were determined from aerial photographs taken in 1943,1955,1961, 1978,1981 and 1994. To outline the watershed on the photographs, we used topographic maps with a scale of 1:10000 and a contour interval of 10 m that we adjusted to the photograph scales. Then, using the six sets of aerial photographs, we mapped six categories of land use: native forest, bushes, deforested areas, exotic forest (pine plantation), urban-residential and grassland. Copyright © 2001 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 26, 63-76 (2001) 66 M. CISTERNAS ET AL. Radiograph Unsupported !,Tb Activity ™Pb Age ™Pb Dates (Bqkg"1) (years before 1996) (years AD) Pinus radiata pollen (grains 104g"1) 40 80 120 1996- 1993- 1989 1984 1978 1972 1968 1961 1951 1948- 1942- 1937- 1930- 1921 1912 1883 Not computed < Sediment Accumulation Rate (mg cm2yf1) 0 20 40 60 ca. 1885 - > Not detected _L_ JL___I Organic Matter (%) 13 14 15 16 J___I___|___I___I___l No data No data Figure D 2. Radiograph, unsupported 210Pb activity, chronological model, 137Cs activity, Pinus radiata concentration, sediment accumulation rate and organic matter plotted against depth in the Chica de San Pedro Lake core Using ARC/INFO GIS software, we measured areas of these land uses, adjusted to take account of land slope. To calculate slope we used the mapped topography to construct a digital elevation model (DEM), from which a digital terrain model (DTM) was derived. The DTM allowed us to increase previously calculated land use areas in proportion to the slope. These methods yielded percentages, for each of the six sets of aerial photographs, of area covered by each land use type. Changes in land use were detected by comparing consecutive photographs. To summarize changes we used ARC/INFO to compute a quantity we call the total detectable change. This computed value, expressed as a percentage of the watershed area, denotes the total land use change without regard for the kind of change (such as a change from native forest to exotic forest). We also computed the rate of total detectable change by dividing the total detectable change by the amount of time between successive photographs. These total changes are merely 'detected' because each was computed by comparing photographs that show conditions at the beginning and end of a period of time. Such photographs do not necessarily show all the land use changes that took place in this measurement period. As a result, we unavoidably failed to detect many of the land use changes that have occurred in the watershed of Chica de San Pedro Lake between 1943 and 1994. RESULTS Radiography Radiography shows the core suitable for measurement of sedimentation rates. As shown in Figure 2A, the core lacks obvious signs of burrowing, or other disruption, that could have redistributed Pb, Cs or pollen after their deposition. Instead, the entire core contains horizontal strata a few millimetres to about 2 cm thick. 210 Pb isotope flux The total residual unsupported Pb load in the core is 007569 Bqcm~ . This load implies a Pb flux rate of 23-57 Bqm below. ' a 1; this flux rate is lower than normally found in the northern hemisphere, as discussed , 210 Unsupported Pb activity and chronological model 210 210T In a simple case, where both Pb fallout and sedimentation rate are constant through time, the Pb Copyright © 2001 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 26, 63-76 (2001) LAND USE AND SEDIMENT YIELD IN CHILE 67 activity in a core declines smoothly with depth from a maximum at the top to a constant value - the supported 210Pb activity - in equilibrium with radium in the sediments (Goldberg, 1963; Appleby and Oldfield, 1978). Under these simple conditions, the downward decrease in unsupported 210Pb activity is controlled solely by the age of the deposit, which controls progressive decay of the 210Pb that had accumulated as fallout. In our core, unsupported 210Pb activity shows an overall decrease downward (Figure 2B). Interpreted with the CRS model, in which 210Pb fallout does not change with time (see Methods), this trend implies that the top 15 cm of the core spans a little more than a century (Figures 2C and D). The earliest date inferred from this model, for deposits 15 cm below the surface, is 1883. For the top of the core, we fit the model to a date of 1996-the year the core was taken-and we compute intermediate dates at 1 cm intervals. This series of dates is mostly consistent with 137Cs and pollen data in Figure 2E and F, as discussed below. The unsupported 210Pb profile contains several fluctuations that are superimposed on the overall downward decrease in activity (Figure 2B). If 210Pb fallout is assumed constant, these fluctuations imply relative changes of sedimentation rate (Figure 2G). The relative minima in 210Pb activity imply relative maxima in sedimentation rate because, if constant 210Pb fallout produces a constant 210Pb flux to the lake bottom, rapid sedimentation dilutes the 210Pb activity (Battarbee et al, 1985; Dearing 1991; Olavi et al, 1990; Alvisi and Frignani, 1996). The highest peak is defined by unsupported activity that is highest not at the top of the core but at a depth of several centimetres. Additional peaks are present lower in the sequence. 137 Cs and pollen tests of the 210Pb chronology The profile of 137Cs activity is dominated by a broad peak that agrees exactly with dates inferred from 210Pb (Figure 2E). The likely date of this peak is 1963, when 137Cs reached its maximum in the atmosphere. The peak corresponds to the time between 1961 and 1968 in the 210Pb chronology. The activity declines above 6 cm depth in the core, probably because almost no 137Cs has been added to the atmosphere since signing of the test-ban treaty in 1963. 1 ^^7 01 C\ 1 ^^7 The lower part of the Cs profile, however, conflicts with the Pb chronology. Cs remains detectable to depths corresponding to 210Pb dates in the 1930s. These 210Pb dates may be about 20 years too old, for atmospheric testing of nuclear weapons produced little 137Cs before the early 1950s. Alternatively, 137Cs migrated several centimetres downward by diffusion after deposition. Diffusion may have also spread the peak that we assign to 1963, but it probably did not move that maximum upward or downward. O 1 C\ 1 ^^7 O 1 C\ Despite discordance between Pb dates and the lowest part of the Cs profile, earlier dates in the Pb chronology are consistent with the lowest appearance of P. radiata pollen (Figure 2F). This lowest pollen, at a depth of 15-16 cm, may have come from the trees introduced in 1885. If so, the pollen data support the 210Pb date of 1883 that we computed for a depth of 15 cm. Sedimentation rates According with the CRS model, net sediment accumulation has varied by an order of magnitude, from 5 mg cm~2 a-1 in 1883 to 60 mg cm~2 a-1 in 1968 (Figure 2G). Changes in sedimentation rates show three pulses of sedimentation in the 20th century. The first began late in the 19th century and reached its maximum in the late 1940s (58 mg cm~2 a-1). Another pulse began early in the 1950s (24 mg cm~2 a-1) and reached its maximum in the late 1960s (60 mg cm~2 a-1). The most recent event started around 1978 (26 mg cm~2 a-1) and continued into the 1990s, when the maximum was 52 mg cm~2 a-1. The deposition of 15 cm of sediment between the computed date of 1883 and the core-top date of 1996 gives a mean sedimentation rate of 1-33 mm a-1. Because the measured mass of sediment is equivalent to 3-35 g for a column 15 cm tall and 1 cm2 in horizontal cross-section, the corresponding mass accumulation rate averages close to 30 mg cm~2 a-1. Organic content All these figures refer to the sum of inorganic and organic sediment. By measuring loss on ignition, we Copyright © 2001 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 26, 63-76 (2001) 68 M. CISTERNAS ET AL. found that organic content varies little with depth - from 13 to 16 per cent of dry weight in the top 15 cm of the core (Figure 2H). The minimum organic content of 13 per cent corresponds to a model age of 1968. We converted organic content to organic sedimentation rate by applying the obtained age-sedimentation model. The results show that the organic sedimentation rate was nearly constant for the model years 1942-1996 (Figure 3B, diagonal hatching). Some of the organic matter probably came from the watershed, while other organic material was probably produced within the lake, but the proportions that can be assigned to these sources are unknown. For these reason, we emphasize the inorganic component of the sediment yield from the watershed. Sediment yield We computed watershed sediment yield as S*AL/AW, where S is the sedimentation rate at the core site, AL the area of the entire lake bed, and^4w the watershed area corrected for land slope. This simplified approach, from Dearing et al. (1987) and Dearing (1991), depends on three assumptions stated in the Methods section. We made the computations for the top 10 cm of the core, which according to our 210Pb dating corresponds to the time between 1942 and 1996. This period contains the years of all the aerial photographs with which we measured changes in land use. Sediment accumulation rate and sediment yield, both plotted on a linear time scale, retain the same three pulses recognized in the depth plot (Figure 2G) but show that they were brief (Figure 3B). The first two peaks, around 1950 and 1970, lasted several years; the third peak, in the 1990s, about seven years. All three approached 1 t ha-1 a-1. Values below the mean yield (0-5 t ha-1 a-1) prevailed for longer periods of time. Organic matter, whatever its origin, has little effect on these patterns. Because the accumulation rate of organic matter varies little with time, organic matter merely increases the total sediment yield by a value that is relatively small (01-0-2 t ha-1 a-1) and nearly constant (Figure 3B). Consequently, the potential production of organic matter within the lake creates little uncertainty about sediment yield from the watershed. Land use changes The watershed of Chica de San Pedro Lake is dominated by a ravine that drains directly into the lake (Figure 3C). The steepest and highest areas are located in the southern headwaters. Because of topographic relief, the watershed area calculated with the DTM (4-8 km2) is 7 per cent greater than the planar area. Most of the native forest in the watershed has disappeared in the past 50 years (Table I). In 1943, the date of our earliest air photos, native forest covered 70 per cent of the area, mainly in the south-central part of the watershed. This land use shrank to 13 per cent by 1994. Concurrently, pine plantations expanded from 4 per cent in 1943 to 46 per cent in 1994. Because these plantations have progressively replaced native forest, the watershed was never completely deforested in the past 50 years. Land uses other than native and exotic forest show various tendencies. Between 1943 and 1994, heterogeneous bushes ranged from a minimum of 19 per cent in 1978 to a maximum of 30 per cent in 1955. Deforested areas, though lacking in 1943 and 1955, covered 16 per cent in 1961, 14 per cent in 1978, 22 per cent in 1981, and just 8 per cent in 1994. The area of grassland was 4-5 per cent in all surveyed years through 1978; thereafter, grassland covered just 1 per cent in 1981 and was absent altogether in 1994. The urban and residential land use began in 1961 (1 per cent) and reached a maximum of 6 per cent in 1981. Land use change, mapped to include all the measured kinds of land use change (the total detectable change defined in the Methods section), migrated progressively southward in the steepest part of the watershed between 1943 and 1994 (Figure 3C and D). Most of this migration was due to the spread of pine plantations, that is, to the expansion of exotic forest described above (Table I). The area of total detected change, when summed for the entire watershed and expressed as a percentage of the area of the watershed, ranges from 30 per cent for the comparison between 1943 and 1955 air photos, to 64 per cent for the comparison between 1981 and 1994 (arrows and dots in Figure 3E). The total detectable Copyright © 2001 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 26, 63-76 (2001) LAND USE AND SEDIMENT YIELD IN CHILE 69 Q —. Q. >. S? ■£ Q. S 60 0 >- C D) 1 £ OT 40 0 i^-Organic I Inorganic ^^^ « « a 1.0 m ^^ >> i r- ,:» 0.5 E 5 ■n (D C/J 0.0 1940 1950 1960 1970 1980 1990 2000 B 400 mi 1940 1950 1960 1970 1980 1990 2000 Years AD F Figure 3. Precipitation record, sedimentation rate, sediment yield, DEM, total detectable change and total detectable change rate plotted against time Copyright © 2001 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 26, 63-76 (2001) 70 M. CISTERNAS ET AL. Table I. San Pedro watershed land-use percentages determined from aerial photographs taken in 1943,1955,1961,1978, 1981 and 1994 Land use 1943 1955 1961 1978 1981 1994 Native forest 70-1 52-8 33-9 34-4 28-8 12-6 Bushes 21-6 30-2 24-8 18-7 28-5 28-5 Deforested areas 0-0 0-0 15-7 14-1 21-7 8-1 Exotic forest 4-4 12-3 19-8 22-6 13-6 45-7 Urban-residential 0-0 0-0 1-3 4-9 6-1 5-1 Grassland 3-9 4-7 4-4 5-3 1-2 0-0 change fluctuated between 30 and 40 per cent for the intervening comparisons (1955 and 1961, 1961 and 1978, 1978 and 1981). As noted above (see Methods), these percentages probably exclude many land use changes that took place during the years between photographs. Such undetected changes probably explain the inverse relation between the rate of total detectable change (bars in Figure 3E) and the length of time between successive air photos. The lowest rate (2-5 per cent a-1) was obtained from air photos taken 17 years apart, while the highest rate (13 per cent a-1) comes from air photos just three years apart. Relationships between sediment yield and land use We looked for relationships between sediment yield inferred from the core and three kinds of land use measurements obtained from air photos (Figure 4). Only the last one, total detected change, shows a consistent relationship. The estimated sediment yield appears unrelated to the first and most basic of our measures of land use - the percentage of the watershed covered by each of six types of land use (Table I). As shown in Figure 4A-F, we made a linear correlation analysis between sediment yield (from the sum of organic and inorganic rates; Figure 3B) and areal extent of different land uses (from the air photos taken in 1943, 1955, 1961, 1978, 1981 and 1994). Sediment yield may be related to land use in just two cases: it tends to increase with area of exotic forest (Figure 4D; r2 = 0-44) and to decrease with area of grassland (Figure 4F; r2 = 0-36). But the small sample size (five snapshots) makes even these trends statistically insignificant. Estimated sediment yield has also no obvious correlation with the rate of total detected change - a rate normalized to the duration of the period between air photos (Figure 4G). The main cause of this non-correlation may be failure to detect land use changes that took place early in long periods between photographs. Such failure may explain why the average rate of total detected change varies inversely with the length of the period between photographs (as shown by the inverse relationship between the width and height of grey bars in Figure 3E). For example, average rates of detected change are lowest for the long intervals 1943-1955 and 1961-1978 (Figure 3E). The average rate is also low for the long interval 1981-1994, even though that time coincides with economic growth in Chile and immediately follows the period (1978-1981) having the highest average rate of detected change. The estimated sediment yield apparently tracks total detected change in land use, but only if that change is expressed without regard for the length of time during which the change occurred. This somewhat surprising relationship is shown in Figure 3F, which compares sediment yield (sum of organic and inorganic rates selected from Figure 3B, replotted as grey bars) with the total detected change (from Figure 3E; arrows and dots). For graphic simplicity, the sediment yield data selected for Figure 3F represent only those four 1 cm core depth intervals that, according to our 210Pb chronology, include the air photo years 1955, 1961, 1978, 1981 and 1994 (Figure 2D). The resulting comparison shows that sediment yield increases with the total detected change. This relationship gives an apparently high correlation (R2 = 0-95), but the sample size is Copyright © 2001 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 26, 63-76 (2001) LAND USE AND SEDIMENT YIELD IN CHILE 71 100 50 0 .c CO ní to -#—< o CO 0 (U CO ■o c CC o 100 50 10 0 15 ■S-5 £