Figure 1. -- Shaded Relief Image of the
Navarro Watershed (39K)
The city of Cartago is located in the watershed and is part of the Gran Area Metropolitana (GAM). The GAM is the primary population center of the country, and increasing population growth is an important issue (Monzon, 1993). For example, the effects of urban fringe growth on water resources, is a concern of the Programa Nacional de Desarrollo Urbano Sostenible (PRODUS).
The 279 km2 Navarro watershed is one of two located at the headwaters of the Reventaz¢n River basin. The watershed exhibits diverse topography, and particularly dramatic relief (Figure 1). The elevation ranges from 3300m near the summit of the Irazu volcano to 1029m at the La Troya streamflow gauge.
Figure 1. -- Shaded Relief Image of the
Navarro Watershed (39K)
Management of surface and subsurface water in the Navarro watershed is important for hydroelectric power production and water supply maintenance. Simulating the hydrologic response of the watershed using spatially distributed characteristics could provide a valuable tool for hydrologic research and watershed management. This paper will describe the physical characterization of the Navarro watershed using geographic information technologies to enhance distributed hydrologic modeling efforts. In particular, the processing of digital elevation data will be described.
The hydrologic model used to simulate runoff in the Navarro
watershed was the Precipitation Runoff Modeling System (PRMS). The
PRMS is a modular modeling system designed to evaluate impacts of
climate and land use on surface runoff, sediment yields, and
general basin hydrology (Leavesley et al., 1983). Parameterization
of PRMS is undertaken based on hydrologic response units (HRU).
HRUs are units of a watershed partitioned on the basis of
characteristics such as vegetation type, precipitation
distribution, slope, aspect, and soil type. The PRMS was operated
within the Modular Modeling System (MMS). The MMS is an integrated
computer software system which provides an operational framework
for development of algorithms and their application towards
modeling physical processes (Leavesley et al., 1996).
Physical
Characterization
The determining factors which influenced the selection of data for this study were requirements to complete the study and availability. In comparison to other Central American countries the amount and quality of environmental data available in Costa Rica is relatively high. This may be due to a national emphasis on the environment, education, and health concerns, rather than military development. However, initial assessment of data resources revealed significant gaps in some areas (eg., soils data) and discrepancies in scales between others.
The availability of data for this watershed may be seen as deficient compared to that needed, for example, for an experimental watershed in which it is desired to apply a fully distributed hydrologic model. However, this state of data availability lends itself to testing a distributed hydrologic modeling approach utilizing HRUs. In the HRU delineation process thematic data can be aggregated, and homogenous units are created from which more generalized parameter values are extracted.
Three types of data were utilized to develop GIS layers for the delineation of HRUs and the parameterization of the PRMS: elevation, digital satellite, and precipitation.
The primary elevation data used was the U.S. Defense Mapping Agency's (DMA) Digital Elevation Terrain Data (DTED). Other elevation data considered bore constraints which were difficult to overcome. For example, topographic maps at 1:50,000 scale which covered the watershed were obtained, but digitizing the entire watershed was judged prohibitive for this study. In addition, complete coverage of the watershed by same scale stereo aerial photographs was not available for construction of a digital elevation model.
DTED data are distributed in 1 degree by 1 degree or smaller cells, with a 16-bit range of elevation values. Two cells of DTED Level 1 data were utilized in this study. The boundaries of these cells extended from 9o N to 10o N and from -83o W to -85o W. According to the DMA, the data for the cells were digitized or scanned from cartographic sources at a scale of 1:250,000, in the mid-to late 1970's.
The DTED data were provided in a three arc/second format. The resolution of this data between 9o N and 10o N latitude is 92.161m2. Elevation is represented by a regular grid of post points spaced at 100m and interpolated with an inverse distance weighted routine using 8 neighbors. The absolute horizontal and vertical accuracy of the data is approximately 200m.
A projection file was created in ARC/INFO to enable rectification of the original points from a geographic grid to a Lambert Conformal Conic projection. This projection was chosen because at low latitudes its conformality property possesses true shape of small areas, and also the ancillary maps available for the study area were represented in this projection. To enable integration with additional data sets the DTED file was resampled to 90m using a bilinear interpolation routine.
A multiple step process was enacted in ARC/INFO GRID to define the drainage pattern and boundary of the watershed. Essentially four steps were carried out to delineate the drainage network using the elevation data: removing sinks in the DEM, assigning flow direction per cell, assigning flow accumulation values per cell, and determining the threshold flow accumulation value that best represented the drainage pattern (Jenson and Dominique, 1988; ESRI, 1992).
In order to delineate the boundaries of the watershed the drainage pattern was first displayed and the location of the streamflow gauge estimated. The streamflow gauge location on the derived drainage pattern was accurate in an east-west direction but approximately 550m south of the actual location, according to a 1:50,000 scale topographic map. The topography in the area where the La Troya streamflow gauge is located opens to relatively level terrain, which may have contributed to the difficulty in defining the location of the gauge. The area draining to the streamflow gauge was identified using the Watershed function in GRID and the flow direction map. The resulting map defined the boundary of the watershed and provided a mask to use with additional GIS data layers.
Additional analysis steps were carried out to derive physical characteristics of the watershed from the elevation data, including area, elevation parameters, perimeter of the watershed, and drainage density. These parameters were compared with parameters produced in previous studies of the watershed.
In a technical report generated by Sol¡s et al. (1991) and a thesis by Baltodano and Hidalgo (1992), similar physical characteristics were extracted from 1:50,000 scale topographic maps for the Navarro watershed. In an earlier technical report, Elizondo (1979) extracted watershed characteristics from the 1:200,000 scale topographic map, San Jos‚ CR2-CM-5, published by the Instituto Geographico Nacional (IGN). The report utilized the Puente Negro streamflow gauge which, due to flood damage, was later replaced downstream by the La Troya gauge.
In this study the area of the watershed was calculated using the GRID function Zonalarea. The perimeter was calculated using Zonalperimeter. The elevation parameters, such as maximum, minimum and mean elevation values, were extracted in ERDAS IMAGINE 8.1. These parameters as well as drainage density were compared to estimates from Solis et al. (1991), Elizondo (1979), and estimates from the primary utility company in the country, the Instituto Costaricense de Electricidad (ICE) (Table 1).
| Parameters | Colby | Solis et al. | Elizondo | ICE | ICE |
|---|---|---|---|---|---|
| Station | L. Troya | P. Negro | P. Negro | L. Troya | P. Negro |
| Area (sq. km) | 279 | 282 | 273 | 274 | 273 |
| Elevation Max. (m) | 3427 | 3200 | 3300 | ||
| Elevation Min. (m) | 1057 | 1020 | 1029 | 1049 | |
| Elevation Mean (m) | 1691 | 1725 | 1620 | ||
| Perimeter (km) | 109 | 87 | 78 | ||
| Drainage Density | .66 | .88 | .46 |
Slope values for the watershed were computed in ERDAS IMAGINE 8.1 and the GIS IDRISI. The following percentage distributions were calculated for the watershed:
An advantage of using DTED data in this study was the ease with which parameters such as physical characteristics were derived. Use of previously derived digital elevation data provided flexibility and analysis capabilities in a timely fashion. The tradeoff was the lack of quality control in data development.
The drainage pattern derived from the DTED data, provided a generally representative depiction of the watershed. However, errors were encountered. Accuracy was assessed by comparisons to drainage patterns from a Landsat Thematic Mapper (TM) image, 1:10,000 scale land use maps, 1:50,000 scale topographic maps, and aerial photographs. An overlay of the drainage pattern on a false color composite TM image (Figure 2) reveals non-systematic errors in the correspondence of the location of the drainage pattern, especially along the slopes of the Irazu volcano in the northern section of the watershed. In other areas the drainage patterns matched.
Figure 2. -- Landsat TM Image, Bands
4,3,2, with Drainage Pattern Overlay (52K)
These errors are due in part to the quality of the DTED data and level of sophistication of present GIS software tools for drainage pattern extraction. The poor representation of the rivers in the northern section of the watershed raises the question of whether the cartographic source for the elevation values was produced before the 1963-1965 eruption of the Iraz£ volcano.
The most recent eruption of the Iraz£ volcano began in March of 1963 and continued to spew large volumes of lithic ash through March of 1965. Accumulation of ash on the slopes of the volcano altered the hydrologic regime of the rivers. A hard impervious crust formed on the mantle of ash which reduced infiltration, increased runoff, and resulted in increased slope erosion, frequent flash floods, and deadly debris flows. Emergency measures were undertaken such as channel improvements, the construction of levees to protect Cartago, and watershed rehabilitation by terracing, drainage diversion, contour trenching, and artificial revegetation Waldron (1967).
The area, mean elevation and drainage density values derived for the Navarro watershed fell between the values produced from other sources (Table 1). The maximum and minimum elevation and perimeter values calculated from the DTED data were somewhat higher than that produced from the 1:50,000 and 1:200,000 scale topographic maps. The absolute accuracy of the DTED data may have affected the differences in elevation values, and the perimeter differences could be due to grid cells representing the watershed outline, rather than a vector line. Overall, the quality of the elevation data and the derived drainage pattern, though not optimal, supported its application in a distributed hydrologic modeling exercise utilizing HRUs.
A digital Landsat TM satellite image from 1986 was classified to provide land cover characterization. Variable illumination angles and reflection geometry due to different slope and aspect orientations limit the effectiveness of Landsat TM classification efforts in mountainous terrain. A method for reducing this "topographic effect", which was developed in a temperate region (Smith et al., 1980; Colby, 1991; Hodgson and Shelley, 1994) was tested in the neo-tropical environment of central Costa Rica.
Aerial photographs and land use maps provided reference information for classification of the TM image. Black and white aerial photographs for the north-central section of the watershed at a scale of 1:20,000 and for the southern section of the watershed at a scale of 1:60,000 were obtained from the IGN. Also, 1:10,000 scale land use maps were obtained for the central part of the watershed.
Improved classification accuracy was obtained using the topographic normalization routines (Colby, 1995). Elevation data quality was believed to have reduced topographic normalization effectiveness and perhaps classification accuracy.
The Navarro watershed is located between diverse precipitation regimes. One of the driest areas in the country extends from Cartago through the Valle Central (Janzen, 1983). The highest rainfall rates in the Upper Reventaz¢n River basin, 7500mm a year, are found approximately 25km southeast of Cartago (Jansson and Rodriguez, 1992). The mean annual precipitation recorded at stations used in this study varied from 1329mm at the centrally located Comandancia station, to 3218mm at the Belen station located in the southeast.
To develop an image of the spatial distribution of precipitation across the watershed, a number of trend surface analyses were undertaken using the GIS IDRISI. Second and third order trend surfaces were derived using mean annual rainfall figures calculated from complete years of data for the 20 year period between 1967 through 1986 (IMN, 1988). Trend surfaces were created using from 7 to 22 stations located in and near the watershed.
Generally, third order surfaces had the highest goodness of fit
(R2) values, but did not seem to represent the spatial distribution
of precipitation in the southwest section of the watershed
accurately due to a scarcity of precipitation stations in the area.
The final image chosen was a second order trend surface created
using fourteen stations. The R2 value for this surface was 91.35
percent.
Applications
Once digital spatial characteristics of the watershed had been generated HRUs were delineated using thematic GIS data layers of watershed sub-basins, a distance buffer from the stream, precipitation distribution, and land cover categories. The sub- basins were delineated using the same techniques as described above to delineate the watershed, however, research and management criteria were also taken into consideration. A distance buffer from the stream was created based roughly on the variable source area concept (Troendle, 1985). The precipitation distribution layer consisted of a three category aggregation of the 2nd order trend surface described above. The land cover layer included the following categories: bare areas, grass, shrubs, trees, and impervious areas. HRUs were delineated using the four thematic layers and an improved GIS-based methodology (Colby, 1995).
Following delineation of the HRUs hydrologic simulation of the watershed was undertaken for a time period during the 1987-1988 Costa Rican water year. Parameters for PRMS were extracted using the HRU boundaries, the HRU thematic layers and additional data layers such as elevation, slope and aspect. The accuracy of the simulations were determined to be sufficient to proceed with a scaling analysis using multiple resolutions of land cover data.
In the scaling analysis a series of land cover patterns
aggregated at 90m2 intervals, from 90m2 to 1260m2, provided areal
dimensions to parameterize PRMS. The fractal dimension D values of
the land cover patterns were also calculated at each resolution.
Simulation of hydrologic runoff was undertaken at each resolution.
A strong correspondence was found between the range of resolutions
at which accurate hydrologic simulations were achieved and the
range of self-similarity of the land cover patterns, as measured by
their fractal dimension (Colby, 1995).
Conclusion
The data available for this study was less than optimal. An
obvious gap was the lack of available soils data at a useful
resolution. Quality of the elevation data effected the derivation
of several watershed characteristics and topographic normalization
efforts. The status of data availability for this study, however,
may be representative of many watersheds in which watershed
modeling and management are desired. For example, in many
developing tropical countries topographic maps at a scale larger
than 1:200,000 or 1:250,000 may not be available. The choice of
this watershed provided the opportunity to evaluate the
effectiveness of utilizing a distributed hydrologic modeling
approach based on HRUs. The capability to aggregate thematic data
and form hydrologic units provided flexibility to work with less
than optimal data. The HRU modeling approach delivered effective
watershed characterization and hydrologic simulation accuracy which
enabled the desired research to be carried out. Utilization of
geographic information systems provided essential capabilities for
improving the processing, management, and analysis of available
data.
Acknowledgements
The author would like to acknowledge the Organization of
American States, the U.S. Geological Survey, the U.S. Defense
Mapping Agency, and numerous organizations and individuals in Costa
Rica for their support in this research effort.
References
Baltodano P., J.A., and H. Hidalgo L., 1992.Definicion de Niveles de Inundacion en el Rio Reventado. In: IV Congreso Nacional de Recursos Hidraulicos. San Jose Costa Rica.
Colby, J.D., 1991. Topographic Normalization in Rugged Terrain. Photogrammetric Engineering and Remote Sensing 57(5):531-537.
Colby, J.D., 1995. Resolution, Fractal Characterization and the Simulated Hydrologic Response of a Costa Rican Watershed. Ph.D. Dissertation, Geography, University of Colorado.
Cortes G., V.M., and G. Oconitrillo C., 1987. Erosion de Suelos Horticulas en el Area de Cot y Tierra Blanca de Cartago. Tesis de Grado para Optar al Grado de Licenciado en Geografia. Departmento de Geografia, Universidad de Costa Rica.
Elizondo M., J.A., 1979. Estudio Hidrogeologico Preliminar de la Cuenca del Rio Navarro Provincia de Cartago. Servicio Nacional de Aguas Subterraneas Riego y Avenamiento, Departmento de Ingenieria e Hidrologia, San Jos‚, Costa Rica.
ESRI, 1992. Cell-Based Modeling with GRID 6.1: Supplement-Hydrologic and Distance Modeling Tools. Environmental Systems Research Institute, Inc., Redlands, CA.
Hodgson, M.E., and Shelley, B.M., 1994. Removing the Topographic Effect in Remotely Sensed Imagery. ERDAS Monitor 6(1)4-6.
Instituto Meteorologico Nacional, 1988. Catastro de las Series de Precipitaciones medidas en Costa Rica. San Jos‚, Costa Rica.
Jansson, M.B., and A. Rodriguez (Editors), 1992. Sedimentological Studies in the Cachi Reservoir, Costa Rica. Department of Physical Geography, Uppsala University, UNGI Rapport nr 81, 217 pp.
Janzen., D.H., 1983. Costa Rican Natural History. Chicago, The University of Chicago Press.
Jenson, S., and J. Dominique, 1988. Extracting Topographic Structure from Digital Data for Geographic Information System Analysis. Photogrammetric Engineering and Remote Sensing 54(11):1593-1600.
Leavesley, G.H., Lichty, R.W., Troutman, B.M., and L.G. Saindon, 1983. Precipitation-runoff modeling system -- User's manual: U.S. Geological Survey. Water Resources Investigations Report 83-4238.
Leavesley, G.H., Restrepo, P.J., Stannard, L.G., Frankoski, L.A., and A.M. Sautins, 1996. MMS: A Modeling Framework for Multidisciplinary Research and Operational Applications. In: GIS and Environmental Modeling: Progress and Research Issues, M.F. Goodchild, L.T. Steyart, and B.O. Parks (Editors). GIS World Books, pp.155-158.
Monzon P., J.P., 1993. Patrones de Crecimiento del Gran Area Metropolitano. Tesis de Grado para Optar al Grado del Licenciado en Ingeneria Civil, Departmento de Ingeneria Civil, Universidad de Costa Rica.
Quesada M., C.A., 1979. Effect of Reservoir Sedimentation and Streamflow Modification on Firm Power Generation. Ph.D Dissertation, Department of Civil Engineering, Colorado State University.
Smith, J.A., T.L. Lin, and K.J. Ranson, 1980. The Lambertian Assumption and Landsat Data. Photogrammetric Engineering and Remote Sensing 46(9):1183-1189.
Solis B., H., Murillo M., W., and R. Oreamuno V., 1991. Estudio Hidrologico e Hidraulico para el Control de Inundaciones en la Cuenca del Rio Purires: Valle de Guarco. Servicio Nacional de Aguas Subterraneas Riego y Avenamiento Direcci¢n de Ingeniera, Centro Agronomico Tropical de Investigacion y Ensenanza: Proyecto Renarm-Cuencas, Costa Rica.
Troendle, C.A., 1985. Variable Source Area Models. In: Hydrological Forecasting, M. Anderson, and T. Burt, (Editors). Wiley, New York. pp. 347-403.
Waldron, H.H., 1967. Debri Flow and Erosion Control Problems Caused by the Ash Eruptions of Iraz£ Volcano, Costa Rica. Geological Survey Bulletin 1241-1. United States Government Printing Office, Washington D.C.
home page