land-use effects on erosion, sediment yields, and reservoir ...

LAND-USE EFFECTS ON EROSION, SEDIMENT YIELDS, AND RESERVOIR SEDIMENTATION: A CASE STUDY IN THE LAGO LOÍZA BASIN, PUERTO RICO Allen C. Gellis U.S. Geological Survey 8987 Yellow Brick Road Baltimore, Maryland 21237 Rick M. T. Webb U.S. Geological Survey Box 25046, MS 413 Denver, Colorado 80225-0046 Sherwood C. McIntyre 7207 W. Cheyenne Street El Reno, Oklahoma 73036 William J. Wolfe U.S. Geological Survey 640 Grassmere Park Nashville, Tennessee 37211 Abstract: Lago Loíza impounded in 1953 to supply San Juan, Puerto Rico, with drinking water; by 1994, it had lost 47% of its capacity. To characterize sedimentation in Lago Loíza, a study combining land-use history, hillslope erosion rates, and subbasin sediment yields was conducted. Sedimentation rates during the early part of the reservoir’s operation (1953– 1963) were slightly higher than the rates during 1964–1990. In the early history of the reservoir, cropland comprised 48% of the basin and erosion rates were high. Following economic shifts during the 1960s, cropland was abandoned and replaced by forest, which increased from 7.6% in 1950 to 20.6% in 1987. These land-use changes follow a pattern similar to the northeastern United States. Population in the Lago Loíza Basin increased 77% from 1950 to 1990, and housing units increased 194%. Sheetwash erosion measured from 1991 to 1993 showed construction sites had the highest sediment concentration (61,400 ppm), followed by cropland (47,400 ppm), pasture (3510 ppm), and forest (2050 ppm). This study illustrates how a variety of tools and approaches can be used to understand the complex interaction between land use, upland erosion, fluvial sediment transport and storage, and reservoir sedimentation. [Key words: erosion, reservoir sedimentation, land use, Puerto Rico.]

INTRODUCTION Sedimentation reduces reservoir storage worldwide (Nordin, 1991; Palmeiri et al., 2001; Nagle et al., 1999). White (2001) and Mahmood (1987) estimated that the worldwide reservoir sedimentation rate is 1 percent per year. Most reservoirs are 39 Physical Geography, 2006, 27, 1, pp. 39–69. Copyright © 2006 by V. H. Winston & Son, Inc. All rights reserved.

40

GELLIS ET AL.

Fig 1. Photograph showing complete sedimentation in Comerío Reservoir in 1989. The reservoir was constructed in 1914 with a storage capacity of 6.06 × 106 m3 and by 1935, was close to 70 percent filled with sediment (Nevares and Dunlop, 1948).

designed to be usable for 50 to 100 years, before they are rendered useless by sedimentation (Hotchkiss, 1995). The effect of reservoir sedimentation has been studied in detail in Africa (Ward, 1980; Jordan, 1989; Shahin, 1993), Asia (Ongkosongo et al., 1992), Australia (Chanson, 1998), and the United Kingdom (Butcher et al., 1992). The rate of reservoir sedimentation can be accelerated by natural or human modifications to the watershed (Ongkosongo et al., 1992; Renwick, 1996). The development of effective strategies to reduce sedimentation rates requires distinguishing between background erosion rates in undisturbed settings and humanaccelerated erosion in disturbed settings (Ongkosongo et al., 1992; Palmeiri et al., 2001; Nagle et al., 1999). Reservoir sedimentation in Puerto Rico was first studied in the 1940s (Nevares and Dunlop, 1948; Fig. 1). They determined that sedimentation rates from three reservoirs in Puerto Rico—Comerío, Coamo, and Guayabal—exceeded the sedimentation rates calculated for two-thirds of the 65 reservoirs examined in the United States. The high erosion rates in the three Puerto Rican reservoir basins were attributable to geologic conditions, deforestation, and intensive cultivation of steep slopes. Lago Loíza, the reservoir analyzed in this study, was impounded in 1953 and provides metropolitan San Juan with over one-half of its water supply. The problem of sedimentation in Lago Loíza was evaluated by Collar and Guzmán-Ríos (1991) using the 1985 bathymetric survey by Quiñones et al. (1989), which indicated an exponential increase in the sedimentation rate over time such that Lago Loíza might be rendered useless by 1998. Their findings led to the development of strategies to alleviate sedimentation (U.S. Department of Agriculture, 1995). A later study by

EROSION AND SEDIMENTATION IN PUERTO RICO

41

Webb and Soler-López (1997) reconciled differences in previous survey methods and indicated there was not an exponential increase in sedimentation. Nevertheless, sedimentation in Lago Loíza and the loss of storage capacity became an important public issue during the drought of 1994, as the limited reservoir capacity led to water rationing in San Juan (Garcia, 1994; Larsen, 2000). Public concern about reservoir sedimentation led the Puerto Rican government to fund a study by the U.S. Geological Survey (USGS) to investigate land-use change in the Lago Loíza Basin and its effect on reservoir sedimentation. This article describes the results of a study conducted by the USGS on the effects of land use on upland erosion rates, watershed sediment yields, and reservoir sedimentation rates for the Lago Loíza Basin, Puerto Rico. Specific objectives of the study were to: (1) determine historical sedimentation rates in Lago Loíza, (2) reconstruct historical trends in land use in Puerto Rico and the Lago Loíza Basin, (3) determine upland erosion on hillslopes of varying land uses in the Lago Loíza Basin, and (4) determine significant variables controlling basin sediment yields and sediment concentrations. SETTING The 539-km2 Lago Loíza Basin is located in east-central Puerto Rico, 22 km south-southeast of San Juan (Fig. 2). Bedrock geology in the Lago Loíza Basin is 48% volcaniclastic rock, 33% intrusive rock (granodiorite and quartz diorite), and 19% Quaternary alluvium. The intrusive bodies weather to sandy loams and loams, whereas the volcaniclastic rocks weather primarily to clays and silty clay loams. Elevations in the Lago Loíza Basin range from 18 amsl at the outlet of Lago Loíza to 1074 amsl at El Toro peak. Mean annual rainfall varies with elevation in the basin, ranging from 1610 mm at National Weather Service (NWS) Gurabo raingage (1957–1993) to 2565 mm at the San Lorenzo NWS raingage (1931–1993; Fig. 2). The Lago Loíza Basin is drained by the Río Grande de Loíza, the largest river in Puerto Rico and the Río Gurabo (Fig. 2). Streamflow to the reservoir is measured at two USGS streamflow stations—the Río Grande de Loíza at Caguas and the Río Gurabo at Gurabo (Fig. 2). The Río Grande de Loíza at Caguas contributes an average annual streamflow volume of 195 × 106 m3 (measured from 1961 to 1993), whereas the Río Gurabo at Gurabo has an average annual streamflow volume of 119 × 106 m3, (measured from 1961 to 1993). From 1984 to 1993, the average water year suspended-sediment load transported by the Río Grande de Loíza at the Caguas streamflow station was 271,200 metric tons while 164,400 metric tons of suspended-sediment was transported by the Río Gurabo at the Gurabo streamflow station (Fig. 2). As is typical of many river systems, the majority of the suspendedsediment load at the two stations was transported in a few days; 80% of the average annual suspended-sediment load was transported in 7 to 12 days at the Río Grande de Loíza at Caguas streamflow station and 80% of the average annual suspendedsediment load was transported in 4 to 24 days at the Río Gurabo at Gurabo streamflow station. Completion of the Carraizo dam on the Río Grande de Loíza in March 1953 created Lago Loíza (Fig. 2) located 22 km upstream from the river’s outlet to the Atlantic Ocean. Lago Loíza had an initial storage capacity of 26.80 ×106 m3.

42

GELLIS ET AL.

Fig. 2. Location of sites, National Weather Service raingages, and U.S. Geological Survey streamflow and sediment stations in the Lago Loíza Basin, Puerto Rico.

METHODS Documenting erosion and sedimentation rates in the Lago Loíza Basin required the collection and analysis of historical and recent data describing reservoir sedimentation, land use, hillslope erosion, and sediment transport. Reservoir Surveys Five bathymetric surveys (1963, 1971, 1979, 1990, and 1994) were used to quantify sedimentation in Lago Loíza (Guzmán, 1963; Hunt, 1975; Iivary, 1981; Webb and Soler-López, 1997). The volume calculations from all bathymetric surveys of Lago Loíza were standardized to match the initial volume in 1953. The initial volume


43

Fig. 3. Location of sediment cores for Cesium-137 analysis, Lago Loíza, Puerto Rico.

was determined from a detailed topographic survey (scale 1:2000) completed in 1947 before the reservoir was constructed (Puerto Rico Aqueduct Service, 1947). Cesium 137 Reservoir sedimentation rates for the periods 1953–1964 and 1964–1990 were calculated based on 137Cs activity in the sediment. The isotope 137Cs is a radionuclide produced in nuclear fission reactions that began to be globally distributed in 1952 with the first atmospheric high-yield thermonuclear reaction tests (Perkins and Thomas, 1980). After deposition, 137Cs becomes adsorbed on soil particles (Davis, 1963) and is carried on soil particles through the erosion cycle, which can result in deposition in lakes as a distinguishable horizon. The first globally measurable 137Cs was detected in 1952 and produced an identifiable 137Cs horizon in undisturbed sediment deposits (Ritchie and McHenry, 1973). The greatest number of atmospheric nuclear tests occurred from 1962 through 1964, which produced a second identifiable 137Cs horizon in undisturbed sediments and is assigned a date of 1964 by Ritchie and McHenry (1973). Lago Loíza was completed in 1953; therefore, for this study, the 1964 horizon was used.

44

GELLIS ET AL.

Sediment cores were taken from four locations in Lago Loíza in December 1990 (Fig. 3). The site locations were selected on the basis of previous bathymetric surveys that identified areas of deposition. The cores were collected using a gasoline-driven capstan winch and a 6-m derrick mounted on a 4-m square wooden platform supported by floats. Sampling began with a 10-cm inside diameter (ID) steel drill casing in 1.5-m sections that was extended to the lake bottom and then into the sediment about 40 cm. The casing top was kept about 60 cm above the water surface. Sediment cores were then collected inside the steel casing in a 2-m long 7.62-cm ID polyvinylchloride (PVC) core tube. The core tube penetrated the sediment by its own weight and the weight of the drill stem pipe by slacking the cable that suspended it. After retrieving the core tube, the sediment core was pushed from the tube. Compaction of the sample was checked by measuring the core before and after it was pushed from the core tube. Sediment cores were collected in sections measuring up to 1.9 m. The majority of core sections were divided into 15-cm increments. After each core section was collected, the drill casing was lowered the distance sampled plus about 40 cm. The laboratory procedure for analysis of 137Cs were as follows. Sediment samples were first dried at 105oC for 24 hours. The samples were then crushed with a mortar and pestle until they passed through a 6-mm mesh screen. The sieved samples were placed in 1-L Marinelli beakers and weighed. Samples were then analyzed for 137Cs using a multichannel analyzer with a lithium drifted germanium detector. 137Cs activity in each sample was counted for 29,000 seconds. Land-Use History and Population Changes Land-use history for the entire island of Puerto Rico was reconstructed for the period 1828–1992 (U.S. War Department, 1900; Murphy, 1916; U.S. Department of Commerce, 1938, 1943, 1952, 1961, 1972, 1980, 1984, 1989; Wadsworth and Birdsey, 1985; Birdsey and Weaver, 1987). Changes in land use for the Lago Loíza Basin were reconstructed using land-use maps created for the following years: 1950 (Brockman, 1952), 1977 (Puerto Rico Department of Natural and Environmental Resources, or PRDNER, 1977), and 1987 (Gellis et al., 1999). Land use was classified for the three time periods into the following six general categories: (1) pasture, (2) forest, (3) cropland, (4) rural, (5) urban, and (6) disturbed land. Rural land use is low-density, residential development along roads away from major urban centers and may include small mixed-crop plots. Disturbed land is bare ground or land cleared of vegetation for construction, recreation, and cropland. Small portions of some basins (Río Cagüitas near Aguas Buenas, 2.6 km2; Río Cagüitas at Villa Blanca at Caguas, 1.2 km2; Río Grande de Loíza at Caguas, 0.2 km2; and Río Turabo above Borinquen, 2.5 km2) were not covered in the aerial photographs, and therefore land use in 1987 was normalized against the basin area covered by the aerial photographs rather than the entire basin area. Land in farms for municipalities within the Lago Loíza Basin for the period 1935– 1992 were compiled from U.S. Department of Commerce (1938, 1943, 1952, 1961, 1972, 1980, 1984, 1989, 1994) and classified as cropland or pasture. Population and housing unit data from 1950–1990 for municipalities within the Lago Loíza


45

Fig. 4. Location of sediment traps operating in the Lago Loíza Basin, Puerto Rico (February 7, 1991– December 12, 1992) listing land-cover type, soil type, vegetation, and permeability. Soils and permeability data were obtained from Boccheciamp (1978). Vegetation, or absence of, was described while collecting sediment data at the traps.

Basin were compiled from U.S. Department of Commerce census of population and census of housing reports covering the period 1950 to 1990 (U.S. Department of Commerce, 1953a, 1953b, 1963a, 1963b, 1973, 1982a, 1982b, 1992).

46

GELLIS ET AL.

Sheetwash Erosion Nine hillslope sites representing principal land-use categories in the Lago Loíza Basin (forest, pasture, cropland, and construction) were selected for measurement of hillslope runoff and sediment or sheetwash erosion (Fig. 4). The period of data collection extended from February 7, 1991, through December 12, 1992. Each site was instrumented with three to five sediment traps labeled A–E. Urban and rural land were not selected for this analysis. Hillslope runoff and sediment were collected in a sediment trap or trough in unbounded plots (Gerlach, 1967). The sediment trap was a collection trough 52 cm wide and 8.5 cm deep. To prevent raindrops from entering the trap, a lid made of sheet metal was fitted with a hinge to the back of the trap. Two 1.27-cm diameter holes were drilled into the side of the trap, one at the bottom and one in the center, and fitted with a 1.9-cm hose and connected to a 18.9-L collection bucket. The center hole was designed to operate if the bottom hole became clogged with organic debris. Traps were visited at selected time intervals. If water was present in the bucket, the bucket was weighed in the field with a hand scale, water was poured out, and sediment from the collection bucket and trough was transferred to another bucket and brought to the laboratory. At selected times, a sample of the poured water was taken back to the laboratory to determine the amount of sediment in the poured water. The water-sediment mixture was dried at 1050C for 24 hours and then weighed. Selected samples were ashed at 5500C for 1 hour to determine the percent organic matter. Contributing areas of the sediment traps were surveyed using a total station (Gellis et al., 1999). In his study, Larsen (1997) estimated the contributing area of sediment traps with a metric tape and noted that the drainage area estimated in this way may be off by as much as a factor of two. Bounding the contributing area to the Gerlach Troughs with an impervious material may be a more satisfactory method (Loughran, 1989; Gellis et al., 2001). Erosion rates for each trap were determined as the volume-weighted sediment concentration and sediment yield. The volume-weighted sediment concentration for each trap, reported in ppm, was calculated by summing the sediment mass (g) retained from each sampled rainfall event and dividing by the sum of runoff (sediment and water) from each rainfall event. All of the traps were not operating concurrently during the same period. The volume-weighted sediment concentration normalizes the data for differences in the collection period by dividing sediment by runoff. Sediment yield for each trap was obtained by summing the sediment mass (g) retained from each sampled rainfall event, and dividing by drainage area (m2), and by the number of days the trap was in operation (g/m2/day). Subbasin Sediment Yields Twelve sediment stations whose contributing areas drain 87% of the Lago Loíza Basin, were selected to determine factors significant in suspended-sediment transport (Fig. 2). Discharge and suspended-sediment transport were examined in these 12 stations for water years 1991 through 1993 (October 1, 1990, to September 30, 1993;


47

Fig. 2). At each station, streamflow and sediment data collection and records computation followed standard USGS guidelines (Carter and Davidian, 1968; Guy, 1969; Porterfield, 1972; Kennedy, 1984; Edwards and Glysson, 1988). Each streamflow station was instrumented with an automatic-pump sampler capable of collecting 24 separate, sequential water-sediment samples. Sampling was initiated when the river stage exceeded a set threshold as detected by a data-collection platform. Missing discharge and sediment data were estimated at some stations. At station 50055300 (Río Bairoa at Bairoa) for the period October 1 through November 18, 1990, daily suspended-sediment loads, in metric tons, were estimated using regression analysis between mean daily discharge and daily suspended-sediment load developed for the period November 19, 1990, through September 30, 1992. Regression analysis for this station produced an R2 value of .78. At station 50055225 (Río Cagüitas at Villa Blanca), streamflow data for the period October 1 through November 27, 1990, were estimated using regression analysis of mean daily streamflow against an upstream station, 50055100 (Río Cagüitas near Aguas Buenas). Regression analysis for this station produced an R2 value of .87. The suspendedsediment load developed for this station, for the period October 1 to November 27, 1990, was estimated using a regression of mean daily discharge versus suspendedsediment load for the period November 28, 1990, to September 30, 1991. Regression analysis for this station produced an R2 value of .73. Because discharge is used in the calculation of suspended-sediment load, there is a spurious correlation between discharge and suspended-sediment load. The high R2 values reported for stations 50055390 and 50055225 are likely owed to this spurious correlation. Statistical Analysis of Subbasin Sediment Yields and Concentrations A statistical analysis was used to determine significant variables controlling sediment yields and sediment concentrations for the 12 sediment stations. Independent variables used in the statistical analysis were basin slope, land use, and soils for the 12 streamflow stations in the Lago Loíza Basin. Best subset regression was the statistical test used to determine significant variables in multiple linear regression by systematically searching through the different combinations of the independent variables and selecting the variables that best predict the dependent variable (Kuo et al., 1992). Predictions for this analysis were considered significant if a P value was less than .05. Dependent variables were sediment yield and the dischargeweighted sediment concentration. Sediment yield in the Lago Loíza Basin was quantified by dividing the average annual suspended-sediment load from 1991 through 1993 by drainage area. The discharge-weighted sediment concentration (mg/L) is calculated as the total suspended-sediment load divided by total streamflow for each year, and averaged over the study period. Guy (1964) analyzed suspended-sediment transport during storm events in seven streams in the Atlantic coast area of the United States and determined that the mean concentration of sediment (discharge-weighted sediment concentration) for a storm event was a better dependent variable than sediment load. Independent variables in the statistical analysis were basin characteristics of slope, soils, and land use. Mean slopes for each subbasin were calculated using a

48

GELLIS ET AL.

Table 1. Results of the Bathymetric Surveys of Lago Loiza, Puerto Rico (Webb and Soler-López, 1997)a Capacity in millions of m Storage loss (%)

3

1953b

1963

1971

1979

1990

1994

26.80

23.41

20.00

16.38

15.20

14.19

0.00

12.65

25.37

38.88

43.28

47.05

a

Capacity in million m3, is calculated for a pool elevation of 41.14 amsl. b Original capacity.

Table 2. Sedimentation Rates for Lago Loiza Based on (A) 137Cs, Average Annual Percentage of Cross-Sectional Area Filled and (B) Bathymetric Surveys, Annual Storage Loss, in Percent Core site (Fig. 3)

1953–1964

1964–1990

(A) 137Cs, average annual percent of cross-sectional area filled 1

0.94

1.21

2

1.39

1.02

3

1.35

1.14

Average

1.32

1.12

(B) Bathymetric surveys, average annual storage loss, in percent 1.23

1.08

geographical information system (GIS). Relief ratios were also used as a measure of basin slope and were calculated for each subbasin. A relief ratio is the elevation difference along a line drawn along the principal drainage course from the mouth of the basin to the basin divide, divided by the length of this line (Strahler, 1957). Sediment yield increases with increased relief ratio (Hadley and Schumm, 1961). Soil textures in each subbasin were taken from the U.S. Department of Agriculture Soil Survey report (Boccheciamp, 1978). Soil texture, defined as the percentage of sand, silt, and clay was used as a dependent variable and was based on the following grain size ranges: sand, 2 to 0.05 mm; silt, 0.05 to 0.002 mm; clay,