Modeling Source-Water Contributions to Streamflow and Characterizing Watersheds in a Mountain Lake District in Idaho

Christopher D. Arp

Department of Biology and The Ecology Center

Old Main Hill 5305

Utah State University

Logan, UT 84322-5305

My primary interest in developing this term project was to explore the applications of ARC-GIS and related tools to assist in modeling hydrologic controls on stream and lake biogeochemistry, along with a corresponding watershed characterization. I prepared a digital elevation model (DEM) using TauDEM for the intended use in TOPMODEL, a distributed hydrologic modeling framework. I also acquired and prepared hydrologic and climate data for model input and validation. This work is an important step towards using TOPMODEL to predict source-water contributions from catchments at specific locations along a stream-lake-stream couplet, which we are studying in the Sawtooth Mountains in Idaho. Modeling results may ultimately be used in the analysis of watershed biogeochemical functioning in mountain lake districts.

A view of the stream inflow to Bull Trout Lake, our primary study site in the Sawtooth Mountains, ID.This picture shows half of a stream-lake-stream couplet, which is a stream reach that flows into a lake, the lake, and the outflow reach and is our main unit of study to understand stream-lake interactions. 
The red color is a dye (rhodamine WT) used for tracing water movement.

Table of Contents

Introduction

Term Project Objectives

Subject Review

Basemap Construction

Preparation of Digital Elevation Model

Hydrologic and Climate Datasets

Watershed Characterization

Conclusions and the Next Steps

Acknowledgements

Literature Cited

Introduction

Research Focus and Stream-Lake Interactions

A National Science Foundation study initiated in 2002 by principle investigators Wayne Wurtsbaugh (limnologist), Michelle Baker (aquatic biogeochemist), James Haefner (biological modeler) at Utah State University and Robert Hall (stream ecologist) at the University of Wyoming.The study area is the Sawtooth Mountains in Idaho and the goal is interdisciplinary understanding of lake-stream interactions.

The focus of my dissertation research is hydrologic and landscape controls on nutrient fluxes, specifically nitrogen, in headwater catchments possessing lakes. This is part of a broader research effort, the Stream-Lake Interactions (SLI) study funded by the National Science Foundation, to better understand and predict nutrient cycling and temperature regimes in aquatic ecosystems of mountain lake districts. The main premise of this study is that streams and lakes have traditionally been studied in isolation, however in many watersheds lakes are linked to stream networks, yet little is know about how these very different, yet coupled ecosystems interact. Investigations focused on the linkages between streams and lakes will help improve our understanding of the hierarchical structure and function of watershed ecosystems and their management. For my Ph.D. research I hope to contribute to this project by investigating hydrologic-biogeochemical linkages at several pertinent scales including: (1) hillslope to channel, (2) stream reach, and (3) catchment scale.

A conceptual diagram of hypothesized stream-lake interactions, including: 1) plunging inflows due to colder stream water, 2) lower N loads in outflows, and 3) higher organic N in the outflow than inflow due to transformations in the lake. I hope to contribute to this conceptualization by studying source-water and N contributions to streams from the watershed.

Investigations at the watershed scale will require analysis of water and solute fluxes from drainage areas at locations along stream-lake-stream couplets (monitoring stations located above and below stream reaches that flow into and out-of lakes). An important aspect of this analysis will be modeling source-water contributions (i.e. overland flow, soil-water, and groundwater) to streamflow generation at these locations and using this information to model and understand solute contributions from these water sources (a mixing model approach). We have proposed using TOPMODEL (Beven et al. 1995), a spatially distributed hydrologic model, to determine source-water contributions for this approach.

Return to Table of Contents

TERM PROJECT OBJECTIVES

1. Prepare a general literature review concerning hydrologic controls on nutrient export and the potential use of TOPMODEL (and associated models) that will be important to the

goals of my research.

2. Develop a GIS basemap for the Bull Trout Lake study area to be used for site description, and data storage and query.

3. Acquire a digital elevation model and prepare it for use in TOPMODEL using TauDEM.

4. Acquire, format, and present hydrologic and climate data to be used for model input and validation.

5. Analyze watershed attributes of Bull Trout and other catchments using TauDEM and ArcGIS/Arc Hydro tools for description and comparison, and to guide hypothesis generation

concerning hydrochemical response.

Return to Table of Contents

SUBJECT REVIEW

Hydrologic Controls on Nitrogen Export in Mountain Catchments

Nitrogen cycling and export in watersheds is a complex process driven by inputs (e.g. atmospheric deposition), biotic uptake and transformations in upland soils and aquatic ecosystems, and the mobility of N in hydrologic fluxes. In mountain environments, snowmelt runoff exports high loads of N accumulated in winter snowpack and flushes N stored in soils and groundwater (Brooks and Williams 1999, Vanderbilt and Lajtha 2000). This N pulse typically proceeds or is coincident with stream peakflow and diminishes substantially over the summer as flows recede and biota utilize available N. Smaller N pulses often occur with storm events or changes in forest primary productivity (e.g. forest leaf senescence).

The contributions to stream N loads also vary spatially within a catchment due distributed attributes such as topography, soils, and vegetation and how these relate to N reactions (e.g. nitrification and mineralization) and hydrologic mobilization and transport (McHale 2000, Band 2001).For example, soil water on unvegetated slopes may have high NO₃^- and low dissolved organic nitrogen (DON), while groundwater at toeslopes may be very different in chemical composition. These water-sources contribute to streamflow in different proportion over time and thus source-water dynamics affect stream N load regimes as well. Therefore an improved understanding of flowpath dynamics and source-water contributions of a catchment can help predict the form and amount of N export to downstream ecosystems (Robson 1992, Band 2001).

Source-Water Contributions and Solute Mixing Models

Streamflow is generated from the sum of water types in a catchment, referred to as endmembers, that vary in amount and proportion contributed. For example, during stream baseflow groundwater is the dominant contributor to stream flow, while during a stormevent the proportion of water from soils and overland flow often increases greatly.Studies of source-water contributions typically discretize two endmembers as ‘old’ (groundwater) and ‘new’ (soil- and event-water) or three endmembers as groundwater, soil-water, and event-water (overland flow) (Robson et al.1992, Hooper 2001).In these studies it is assumed that endmembers have distinct chemical signatures, due to different sources, flowpaths, residence times, etc., often determined by conservative solutes (e.g. Ca²⁺, Cl^-, SO4^-2) or isotope ratios (e.g ¹⁸O or deuterium) (Hooper 2001, Chanat 2002).Mixing models are then used to predict the volume of each endmember source from the generalized equation:

C_TotalV_Total = C_soilV_soil + C_groundwaterV_groundwater + C_eventwaterV_eventwater,

where C is concentration and V is volume. Mixing model approaches generally require measurements of stream discharge and solute concentrations, along with the analysis of endmember solute concentrations. Such methods have been applied successfully in a number of situations, however the costs of chemical analyses and fieldwork, along with inherit model assumptions, often limit their applicability.

TOPMODEL and Hydrochemical Prediction

Another approach to modeling source-water contributions to streamflow is through hydrologic modeling predictions. Spatially distributed hydrologic models are most appropriate for making these types of predictions because they account for catchment heterogeneity, such as slope, contributing area, soils, and vegetation, that contributes water and solutes to streams in temporally varying amounts and proportions. TOPMODEL is a framework for spatially distributed modeling that is widely used for various applications and is flexible in allowing the user to specify hydrologic behavior for particular conditions (Beven 1995). The incorporation of geographic information systems (GIS), such as ArcMap, in spatially distributed models allows more efficient preparation of input data and analysis of model results.

TOPMODEL has been successfully used for modeling source-water contributions and solute loads in the United Kingdom (Robson et al. 1992) and small mountain catchments in Colorado (Hornberger et al. 1994, Boyer et al. 2000).These studies used concurrent mixing model approaches to validate and calibrate TOPMODEL estimations of source-water and partitioned solute loads and suggest that TOPMODEL predictions may provide valid alternative to more intensive mixing model approaches in studying catchment hydro-biogeochemistry.

A number of versions of TOPMODEL are available for particular applications (e.g. snowmelt modeling).I have not determined yet, which model may be most appropriate for the site and application for my study. However for this term project, I have focused on the model used by David Tarboton and colleges at Utah State University and the corresponding input data for this version. For this version of TOPMODEL, processed DEM files and climate data are prepared first in TOPSETUP.

Return to Table of Contents

BASEMAP CONSTRUCTION

Steps

1. Acquire Datasets

· Hydrologic Cataloging Unit (HUC) for 2-digit water resource region 17 (Columbia River Basin)

http://water.usgs.gov/lookup/getspatial?huc250k

• Enhanced River Reach File 1 (RF1) for region 17

http://water.usgs.gov/pub/dsdl

• Other digital datasets (DEM and NHD data) were also utilized for basemap construction, however will be detailed later.

2. Create a Geodatabase in ArcCatalog and Import HUC and RF1 Coverages

3. Add Monitoring Station Points 

• Create shapefile in ArcToolBox (click Edit to inherit coordinate system of the baselayer) and add it to Geodatabase
• Use ‘Editor’ toolbar to create points at correct locations (monitoring stations located using a Global Positioning System this summer)

Results

Site map of the Upper Warm Springs Creek watershed surrounding Bull Trout Lake in the Sawtooth Mountains of Idaho.

The stream-lake-stream couplet is indicated by monitoring stations and corresponding catchment areas.

Note that other spatial data was utilized to construct this particular layout, however will be detailed in later sections.

Return to Table of Contents

PREPARATION OF DIGITAL ELEVATION MODEL

Steps

1. Obtain Datasets

· Seamless Digital Elevation Model (30 m) that encompassed the entire study area of interest (44.15°N, 115.15°W to 44.50°N, 115.50°W)

http://edcnts12.cr.usgs.gov/ned/

· National Hydrography Dataset for HUC 17050120 (South Payette River)

http://nhd.usgs.gov/data.html

2. Project Datasets

· In ArcToolBox use ‘Project Wizard (coverages and grids)’ in Data Management Tools/Projections

· Project both coverages in Universal Transverse Mercador (UTM) units for zone 11 (longitude -115) and 0X shift and Y shift

· Change ‘Save as type’ from coverage to grid and specify ‘Cubic’ for resampling method

3. Add Data to ArcMap

A view of the raw DEM and NHD network for the Bull Trout Lake study site.

4. Use TauDEM (to obtain TauDEM and learn about its application go to

http://moose.cee.usu.edu/taudem/taudem.html) to process DEM for input to TOPSETUP.

· Start with ‘Basic Grid Analysis’ and ‘Fill Pits’ in order to raise the elevation of any DEM raster cells to that of surrounding cells to allow continuous flow directions to be calculated.

· Proceed sequentially with ‘D8 Flow Directions’, ‘Dinf Flow Directions’ and so on through each ‘Basic Grid Analysis Step’.

· Upon completion check ‘src’ delineated channel raster network against the NHD network. I found major discrepancies and decided that the NHD network was much closer to what I

had observed form field observation and topographic maps.

  A view of the stream raster network (‘scr’) overlain by the NHD network indicating 

a large difference in channels and flow directions. The latter network is correct,
so the processed network needs to be edited and ‘burned-in’ .

· The following general directions in TauDEM are provided to deal with such issues:

A grid giving flow directions used to impose existing streams into the system. This should use the same encoding as D8, i.e. 1 - east, 2 - North east, 3 - North, 4 - North west, 5 - West, 6 - South West, 7 - South, 8 - South east. No data values should indicate off stream locations. These flow directions are given precedence over the flow directions determined from the DEM, so this approach should only be used when the stream data source is deemed to be better than the DEM. This can be created by the network editor in MapWindow or in ArcGIS by burning in a stream feature dataset using the following steps.

1. Convert features to raster retaining the same cell size and extent as the target DEM. Call the resulting grid strgrd.

2. Use raster calculator to subtract a large number from each elevation value that corresponds to a stream. This results in a temporary DEM with deep canyons along the streams. Call

the resulting grid demcanyon.

3. Use "Fill Pits" and "D8 flow directions" to calculate flow directions on demcanyon. The flow directions calculated will be demcanyonp.

4. Use raster calculator to evaluate demcanyonp/strgrd. This will result in no data values off the stream raster due to a divide by 0, but will retain flow directions calculated on the stream raster. The convention for naming the result is to use the suffix fdr. This is the grid input to the "Fill pits function" to enforce stream flow directions.

· Before step 1, I also needed to add lines and nodes (channels) using ‘Editor’ to correct the drainage network before it was rasterized.

A view of the corrected drainage network overlain by the original NHD network.

• Add monitoring point shapefile (from ‘Basemap Construction’ step 3) and confirm that points are located exactly within a ‘scr’ raster cell.
• Using TauDEM ‘Network Delineation’ select ‘Do All Network and Watershed Delineation Steps’ and specify to delineate above the monitoring point shapefile. For ‘Stream Delineation’, I set the accumulation threshold at 50 cells, as I was unable to perform a constant drop test to the size of the catchment being analyzed.
• During this process, the step ‘Stream Reach and Watershed Grid’ is important because two data files (nedtree.dat and nedcoord.dat) are generated that will be used in TOPSETUP.
• The final steps are to calculate the ‘Wetness Index’ and ‘Flow Path Distance’ in TauDEM’s ‘Specialized Grid Analysis’, which will be input in TOPSETUP and are the primary raster datasets used to run TOPMODEL.

Results

A view of the processed DEM datasets that will be input into TOPSETUP and used in TOPMODEL to model

saturation excess thresholds and hillslope flow distances to stream channels.

Return to Table of Contents

Hydrologic and Climate Datasets 

Precipitation

The water input to TOPMODEL is precipitation, both in the form of rainfall and snow. Obtaining accurate rainfall for mountain regions is often difficult as weather patterns and rainfall intensities can vary greatly by elevation and position within a mountain range. Fortunately for the Bull Trout Lake study site, a National Resource Conservation Service SNOTEL station (Banner Summit #312, http://www.wcc.nrcs.usda.gov/snotel/snotel.pl?sitenum=312&state=id) is located < 2 km (44.30°N, 115.23°W, ~2150 m) from the center of the study catchment and at a similar elevation. SNOTEL stations provide daily precipitation and snow-water equivalents data. This data should be verified with on site measurements and may need to be standardized by elevation.

Graphs showing rainfall (left) and snow water equivalent (right) during the spring and summer of 2002 from the

NRCS Banner Summit SNOTEL station near Bull Trout Lake, ID.

Most versions of TOPMODEL do not specifically model snowmelt, so I may use a method by Hornberger et al. (1994) and Boyer et al. (2000) to model snowmelt based on temperature and stratification of the catchment into north and south aspects and high and low elevations. Tools in ‘Spatial Analyst’ of ArcGIS readily facilitates such determinations.

A layout stratified sub-basins at Bull Trout Lake watershed that can be used in modeling snowmelt.

Temperature

Another important TOPMODEL input parameter is temperature, which can be used to model evapotranspiration (Hamon 1961) and snowmelt (Hornberger et al. 1994).We measured hourly temperature from a weather station located at the center of the study catchment using a thermocouple (HOBO).

A graph showing mean daily temperature recording at the Bull Trout Lake

weather station (SLI study) during 2002.

Streamflow

Stream reaches above and below Bull Trout Lake were instrumented with at pressure transducers (Global Water) to gage water levels at lake inflows and outflows, along with 2000 m above and below the lake. Stream discharge was measured 12 times at each station throughout the snow-free season to establish a stage-discharge relationship and convert hourly stage measurements to discharge. Having accurate discharge data is important for comparing TOPMODEL predictions and for model calibration. Additionally, monitoring stations were also instrumented with automatic samplers to collect water for chemical analyses. Discharge data combined with nutrient concentrations (i.e. NO₃, PO₄, TN, and TP) will be utilized to calculated nutrient loads exported from the Bull Trout Lake watershed along the stream-lake-stream couplet.

.
A hydrograph from locations along the stream-lake-stream couplet at Bull Trout Lake watershed in 2002.This shows peakflow occurring

around June-1 due to snowmelt runoff, with several storm events in mid-June and early July, and flows gradually receding to

baseflow conditions in August. Flow increases are noted from high to low stations in the watershed,

though not easily recognized from this display.

Return to Table of Contents

WATERSHED CHARACTERIZATION

In this section I present some basic descriptive analyses of the Bull Trout Lake study catchment, along with data from 5 other local watersheds for comparison that are part of the Stream-Lake Interactions study. These catchments were selected because they have different extents and configurations (positions) of lakes within their watersheds. I did this comparison to better understand whether watersheds with and without lakes were otherwise similar or differ by other characteristics as well (e.g. form and land cover). These measurements were made from raw and processed DEMs, a USGS National Land Cover Dataset (http://edcwww.cr.usgs.gov/programs/lccp/nationallandcover.html) raster grid using tools in ArcGIS (primarily ‘Raster Calculator’ and ‘Reclassify’ in Spatial Analyst).It is noted that these measurements are a general and first attempt to compare these watersheds and no statistical analyses or thorough interpretation of these metrics are given. Additionally many data are presented as means, which may not be the best way to describe how these parameters represent watershed morphology and hydrologic and biogeochemical behavior.

A view of the Bull Trout and other selected watersheds in the Sawtooth Mountains. These watersheds represent

varying extents and configurations of lakes and are thought to be otherwise similar.

Steps (for most measurements)

1. Specify a watershed mask in ‘Spatial Analyst / Options’.

2. Use ‘Spatial Analyst / Raster Calculator’ to delineate the raster dataset to the watershed mask (i.e.specify the dataset and click ‘Evaluate’).

3. Use ‘Spatial Analyst / Reclassify’ to isolate the data of interest (e.g. specific landcover classes or to stratify elevations) and then view the attribute table and use ‘Summary Statistics’ if necessary to compute means or other data summaries.

Other tools / methods in ArcGIS could probably be used to accomplish these same computations; however this is what I found to work.

Results

Watershed	Drainage Area (km^2)	Outlet Elevation (m)	Maximum Elevation (m)	Mean Aspect	Lakes
Bull Trout	11.32	2116	2611	NE	terminal lake only
Stanley	41.77	1985	3004	NW	few tarns and terminal lake
Yellow Belly	30.85	2151	3249	NW	lake chain w/ terminal lake
Pettit	28.85	2091	3169	NW	lake chain w/ terminal lake
North Alturus	24.78	2164	3236	W	many tarns, no terminal lake
Beaver Creek	39.15	2168	3117	NW	no lakes

General characteristics including the extent and configuration lakes of Bull Trout watershed and other select watersheds in the Sawtooth Mountains, ID.

These descriptive data and information generally show that these watershed are similar, except for the extent and configuration of lakes,

though the Bull Trout watershed is smaller and has a lower divide than the others analyzed. 

Watershed	Stream Order (outlet)	Mean Channel Gradient (m/m)	Outlet Reach Channel Gradient (m/m)	Drainage Density (km/km^2)	Mean Wetness Index*	Hypsometric Index**
Bull Trout	2	0.169	0.0007	0.94	0.0034	0.37
Stanley	3	0.099	0.0059	0.63	0.0041	0.34
Yellow Belly	2	0.067	0.0032	0.63	0.0046	0.36
Pettit	2	0.064	0.0253	0.63	0.0051	0.42
North Alturus	3	0.087	0.0311	0.71	0.0063	0.42
Beaver Creek	2	0.077	0.0154	0.72	0.0037	0.36

*Wetness Index = Slope / Specific Contributing Area,

**Hypsometric Index = (mean elevation - min. elevation) / (max. elevation - min. elevation)

Bull Trout watershed appears somewhat distinct in channel gradient metrics and drainage density, which may relate to its smaller size and/or location

and aspect that probably reflect a different glacial history than watersheds with N-NW aspects on the opposite side of the divide.

A smaller wetness index describes a shallower water table depth (more wet) and can also be considered as a measure of topographic

steepness (larger number is more steep).Comparison of wetness index indicates that lakeless watersheds generally have greater extent

of low gradient, moist surfaces (i.e. wetlands), though the NLCD data (below) does not represent this well among all sites.

The higher wetness index (steeper topography) of watersheds with many large lakes may be due to the glaciated features that form lakes,

such as steep walled cirques.

A graph of hypsometric curves (cumulative area per elevation) plotted for each watershed. Standardized curves represent a means to compare the general morphology of watersheds, though more detailed analyses of curve types can be used. A comparison of curve shapes indicates that Bull Trout, Stanley, and Beaver Creek (watersheds with < 2 % lake coverage) are somewhat similar, while the other watersheds with > 3 % lakes (Yellow Belly and Pettit) and
many high elevation lakes (North Alturus) may be different in form.

Watershed	% Lakes	% Wetlands	% Forest	% Bare Rock / Unvegetated	% Alpine
Bull Trout	1.57	0.12	62.4	3.7	0
Stanley	1.61	0.18	49.3	18.7	2.1
Yellow Belly	3.38	0.05	34.1	36.2	17.3
Pettit	5.48	0.06	38.2	28.9	21.2
North Alturus	0.22	0.02	31.1	41.1	18.0
Beaver Creek	0	0.24	47.8	14.5	12.5

Land cover (%) characteristics of Bull Trout and other select watersheds in the Sawtooth Mountains, ID (from USGS National Land Cover Dataset 1992, alpine is land above 2800 m from a DEM).This comparison shows the intended differences in lakes among watersheds, but also notable differences forest, extend above treeline (alpine), and unvegetated, rocky areas. Though these data are interesting, however measurement accuracy and representation may not be reliable enough to dra any conclusion (data is produced from remotely sensed images and correlation of spectra to broad cover types).

Return to Table of Contents

CONCLUSIONS AND NEXT STEPS

I was able to obtain and prepare most of the datasets for use in TOPMODEL, with the exception of finding adequate soils data. Learning how to use TOPMODEL and design the model to appropriately represent Bull Trout Lake watershed will be more difficult and I am still unsure of how to treat the presence of lake using this approach (i.e. all depressions, such as lakes, are ‘filled’ during the DEM processing).This may be an important limitation to using TOPMODEL. Hydrologic modeling is a challenging and labor intensive process with no guarantees of successful results and I will need to weight this against alternative approaches to studying hydrologic controls on nitrogen biogeochemistry. I found the descriptive and analytical capabilities of tools in ArcGIS to be very useful for simple, and possibly more specialized, analyses that I intend to pursue in the future.

Potential Next Steps

1. Continue to explore the applications of spatial datasets and ArcGIS in analysis of hydrologic and geomorphic features in our study sites.

2. Determine whether hydrologic modeling and TOPMODEL approaches to hydro-biogeochemical studies are the appropriate methods for answering the questions I am interested in, and if so proceed with steps 3-6.

3. Locate and obtain a soils dataset to represent soil porosity in TOPMODEL (SSURGO data was not available for this area).

4. Enter datasets into TOPSETUP to prepare them for TOPMODEL.

5. Run TOPMODEL and compare results to measured streamflow data.

6. Continue with iterative modeling process and begin exploring nitrogen source modeling methods.

Return to Table of Contents

Acknowledgements

I thank Dr. Michelle Baker, my advisor, for her initial ideas in developing this term project and for the providing opportunity to work on the Stream-Lake Interactions study, and Dr. David Tarboton, course instructor, for his time and considerable patience in helping me to learn ArcGIS and TauDEM. Support for my graduate education is provided by a fellowship from the Subsurface Science Graduate Program of the Inland National Research Alliance and support for my research is from the National Science Foundation, which I appreciate greatly.

Literature Cited

Band, LE, CL Tague, P Groffman, and K Belt.2001.Forest ecosystem processes at the watershed scale:

hydrological and ecological controls of nitrogen export.Hydrological Processes 15:2013-2028.

Beven, L, R Lamb, P Quinn, R Romanowicz, and J Freer.1995.TOPMODEL.In: Computer Models of

Watershed Hydrology, Singh (ed.), Water Resources Publications, Colorado.

Brooks, PD and MW Williams. 1999.Snowpack controls on nitrogen cycling and export in seasonally

snow-covered catchments.Hydrologic Processes 13:2177-2190.

Chanat, JG, KC Rice, and GM Hornberger. 2002.Consistency of patterns in concentration-discharge

plots.Water Resources Research 38(8):22.1-22.10.

Hamon, WR.1961.Estimating potential evapotranspiration. Journal of Hydraulics Divisions, ASCE

87:107-120.

Hooper, RP.2001.Applying the scientific method to small catchment studies: a review of the Panola

Mountain experience.Hydrological Processes 15:2039-2050.

Hornberger, GM, KE Bencala, and DM McKnight.1994.Hydrological controls on dissolved organic

carbon during snowmelt in the Snake River near Montezuma, Colorado.Biogeochemistry 25:147-

165.

McDonnell, JJ and T Tanaka.2001.On the future of forest hydrology and biogeochemistry.

Hydrological Processes 15:2053-2055.

McHale, MR, MJ Mitchell, JJ McDonnell, CP Cirmo.2000.Nitrogen solutes in an Adirondack forested

watershed: importance of dissolved organic nitrogen.Biogeochemistry 48:165-184.

Robson, A, K Beven, and C Neal.1992.Towards identifying sources of subsurface flow: a comparison

of components identified by a physically based runoff model and those determined by chemical

mixing techniques.Hydrological Processes 6:199-214.

Return to Start