Using GIS to select drainage basins for sampling: An example from a cosmogenic 10 Be study of erosion rates within the Susquehanna River Basin Joanna M. Reuter 1 Paul R. Bierman Milan J. Pavich 1 Abstract We used GIS analysis to effectively select a series of sub-basins within the 71,250 km 2 Susquehanna River Basin for fluvial sediment sampling. Measuring cosmogenic 10 Be from these samples allowed us to develop a better understanding of erosion rates and patterns within the Susquehanna Basin and to investigate relationships between erosion rates and basin-scale characteristics. We began the analysis by obtaining regional datasets of topography, bedrock geology, glacial extent, precipitation, land cover, and physiographic province, all of which are freely available. After delineating boundaries of thousands of sub- basins at a range of scales, we summarized the landscape characteristics within the basin boundaries. We plotted the summarized characteristics to gain insight regarding the types and quantities of basins available for sampling. Guided by what we learned, we decided to base the sampling strategy on the following factors: basin scale (3-10 km 2 ), position south of the glacial margin, physiographic province, lithology, and mean basin slope. Finally, we used a series of queries to display basins with the desired characteristics. Because the pool of remaining candidate basins was small, we manually selected basins at this point, examining the digital topographic maps to screen for excessive disturbance (such as strip mines) and difficulty of access. Our results demonstrate the effectiveness of using GIS to aid in developing a sampling strategy. Using the 10 Be data generated from 60 sampled basins, we identified positive correlations between slope and inferred erosion rates, though the erosion rates for the sampled lithologies (sandstone, shale, and schist) are indistinguishable from each other after accounting for slope. Rates extrapolated from the small basins, based on the identified relationships, are close to rates measured for larger basins. This systematic approach to basin selection can be applied to any research that requires sampling of a small subset of basins from a large group of possibilities. University of Vermont, Burlington, VT } U.S. Geological Survey, Reston, VA ) Delineation of thousands of sub-basins We delineated basins using the following steps: Filled sinks in digital elevation data. Computed a flow direction grid (FlowDirection). Computed a flow accumulation grid (FlowAccumulation). Iteratively performed the following steps: Queried flow accumulation grid for basin size range of interest (maximum size < 2x minimum size, to avoid problems with nested basins and basin fragments; quarter log increments accomplish this). Assigned a unique ID to each point (StreamLink). Determined drainage basin boundaries (Watershed). Converted basins to a shapefile. Examples of basins delineated using these steps for three different size ranges: 1) Acquisition and preparation of digital spatial data We obtained the following digital data sources and projected to UTM, NAD 83: Elevation: National Elevation Dataset (NED) Slope derived from NED Precipitation: PRISM Land use: National Land Cover Data (NLCD) Geology: Digital bedrock geology of Pennsylvania Glacial margin: Physiography Digital Raster Graphics (DRGs) ftp:// 3) Summary & analysis of sub-basin characteristics Basins essentially serve as cookie cutters to summarize the data values within the basin boundaries. 5.6 – 10 km 2 56 – 100 km – 1,000 km 2 Summarizing continuous value data Examples: slope, elevation, precipitation In ArcView 3.x: Analysis > Summarize Zones... In ArcGIS: Spatial Analysis > Zonal Statistics... Output for each drainage basin polygon: pixel count, area, minimum, maximum, range, mean, standard deviation, and sum Summarizing discrete value data Examples: lithology, land use In ArcView 3.x: Analysis > Tabulate Areas... In ArcGIS: not available in menus of ArcGIS 8.x Output for each drainage basin polygon: the area of the basin mapped according to each unique value of the grid (such as forested, urban, agricultural); can be converted to % of basin (Left) Each point represents a non- glaciated sub-basin of the Susquehanna River Basin. The basins included in this plot are > 1 km 2 in area and are mapped as a single lithology. (Right) Each point represents a sub- basin of the Susquehanna River Basin. As basin area increases, the range of mean basin slope decreases. The vertical lines are artifacts of the basin size ranges that were specified when delineating basins. Examples of plots of summarized data that provided useful insight for the development of a basin selection strategy:

5) Basin selection 6) Sampled the GIS-selected basins The outlines of the basins that we sampled are shown in the context of the sampling strategy, rather than in their geographic context: 7) 10 Be results demonstrate effectiveness of GIS-based approach to basin selection We used in-situ produced 10 Be measured from fluvial sediment to infer basin-scale erosion rates on a year time scale (Brown et al., 1995; Bierman and Steig, 1996; Granger et al., 1996). Samples come from two groups of basins: GIS-selected basins: small basins selected with a GIS-based sampling strategy, grouped by physiography, lithology, and slope USGS basins: samples from USGS stream gaging stations, selected because they have sediment yield data (15 to 8,700 km 2 in area) Acknowledgments We thank E. Butler, J. Larsen, R. Finkel, and M. McGee for assistance with sample collection and processing. Research was funded by the USGS and NSF EAR and EAR J. Reuter was supported by an NSF Graduate Fellowship. References cited Bierman, P.R., and Steig, E., 1996, Estimating rates of denudation and sediment transport using cosmogenic isotope abundances in sediment: Earth Surface Processes and Landforms, v. 21, p Brown, E.T., Stallard, R.F., Larsen, M.C., Raisbeck, G.M., and Yiou, F., 1995, Denudation rates determined from the accumulation of in situ-produced 10 Be in the Luquillo Experimental Forest, Puerto Rico: Earth and Planetary Science Letters, v. 129, p Granger, D.E., Kirchner, J.W., and Finkel, R., 1996, Spatially averaged long-term erosion rates measured from in situ-produced cosmogenic nuclides in alluvial sediments: Journal of Geology, v. 104, p Matmon, A., Bierman, P.R., Larsen, J., Southworth, S., Pavich, M., Finkel, R., and Caffee, M., 2003, Erosion of an ancient mountain range, the Great Smoky Mountains, North Carolina and Tennessee: American Journal of Science, v. 303, p PhysiographyLithology Slope RangeIDDescriptionName Easting 17NAD27 Northing 17NAD27quad Area km2 Slope mean % forest% ag 1 Appalachian Plateaussandstone state forest land, short walk from long dirt road trib to Little Birch Island Run Pottersdale Appalachian Plateaussandstone 4855 some new trails/dirt roads in vicinity (does that imply logging??)Moccasin Run Keating Appalachian Plateaussandstone road (state 120) parallels stream, access from roadBig Run Rathbun Appalachian Plateaussandstone 2793 state game land?, possible 2.5 km hike in from south (easy terrain) or 3km along stream from northPebble Run The Knobs Appalachian Plateaussandstone state forest land, short walk from long dirt road; sample with trib to Little Birch Island Run Little Birch Island Run Pottersdale ) Development of basin selection strategy Guided by what we learned from the GIS analysis, we developed a basin selection strategy in order to assess relationships between drainage basin characteristics and erosion rates inferred from 10 Be in fluvial sediment. This table summarizes several important factors we considered: The basin selection strategy is also based on physiographic province, because these provinces subdivide the region by topographic and geologic characteristics. We selected only non-glaciated basins, because glaciation violates assumptions for inferring erosion rates from 10 Be. For each group of data, we wanted enough basins to have statistically meaningful results, so we limited the number of physiographic and lithologic combinations. We queried the delineated basins for desired characteristics. For example, the map (left) shows the results of a query based on these characteristics: Lithology = “sandstone” Area > 3 km 2 Area < 10 km 2 Barren < 3% (proxy for strip mines and other disturbance) Dams in basin = false The queries yielded a manageable number of basins for manual screening with DRGs (digital topographic maps). Access (public or private land) was a major concern. We prepared a list of basins to visit, allowing for some attrition (due to access, for example). Here is an excerpt from the list: EXPLANATION Appalachian Plateaus GIS-selected basins Piedmont GIS-selected basins Valley and Ridge GIS-selected basins USGS basins (Right) 10 Be-inferred erosion rates for the GIS-selected basins in the Susquehanna River Basin range from 4 to 54 m/My. This is a notably wider range of rates than for the Great Smoky Mountains of the southern Appalachians, where the sampling strategy was developed for other reasons, without utilizing GIS to seek out diverse basin characteristics. (Right) As basin area increases, the range of 10 Be erosion rates decreases. The greatest range of erosion rates are observed among the GIS-selected basins (16 ± 10 m/My, mean and standard deviation). The USGS basins, at larger basin scales, yield a narrower range of erosion rates (14 ± 4 m/My). (Below) Relationships between slope and 10 Be-inferred erosion rate for the GIS-selected basins in each physiographic province and (bottom right) for the USGS basins collectively. Correlations exist among the groups of basins which show the widest range in mean basin slope. After accounting for slope, no discernible relationship exists between lithology and erosion rate in the Valley and Ridge. Basin selection goals and 10 Be considerations Insight from GIS Slope Goal was to select basins with a range of slopes, from gentle to steep, in order to investigate relationships between topography and erosion rates. As basin size increases, the range of basin mean slope decreases. Lithology 10 Be concentrations from single lithology basins can be robustly interpreted as erosion rates, due to uniformity of quartz distribution. The number of available single-lithology basins increases as basin scale decreases. Basin scale Basins should have well developed streams that serve to mix sediment from the basin. Sampling small basins helps to achieve goals of single lithology basins across a wide range of mean basin slopes. We decided upon a target basin area of 3 to 10 km 2.