Hydrologic network metrics based on functional distance and stream discharge David Theobald & Mary Kneeland Natural Resource Ecology Lab Dept of Recreation & Tourism Colorado State University Fort Collins, CO USA May 16, 2003
Goal: develop approaches for spatio-temporal design and modeling in order to further our understanding of aquatic resources Objectives, to develop: 1. spatio-temporal models for a continuous response, 2. spatio-temporal models for count and/or categorical data, 3. design and analysis methods for data collected at different scales.
STARMAP Projects 1. Combining environmental datasets (Hoeting) 2. Local inferences (Briedt) 3. Development and evaluation of landscape indicators (Theobald) 4. Extension and outreach (Urquhart) 5. Integration and coordination (Urquhart)
Big questions: Broad-scale processes (e.g., acid deposition in Mid-Atlantic region) to watershed processes Probability-based sampling for state compliance to CWA Sampling perennial/intermittent streams (I.e. flow all year for most years) –What is perennial and shouldn’t be? (~24%) –What is not included and should be? (~18%) Fragmentation of hydrologic regime on biodiversity
Goals of indicator development Develop and evaluate landscape-level indicators suitable for spatial and temporal analyses of EMAP data Investigate limitations of currently- available data and offer new, robust methodologies
Overview of presentation Link watershed and hydrologic network: “…in every respect, the valley rules the stream.” – Hynes 1975 From surrogates to direct measures Towards network-based metrics
Indicators that measure watershed characteristics and aquatic ecology: Reviews 1. Land use in entire watershed vs. riparian buffer (IBI): -watershed better: Richards et al buffer better: Arya (1999); Lammert and Allan (1999) 2. Other indicators: - road density (Bolstad and Swank) - dam density (Moyle and Randall 1998) - amount of roads near streams (Moyle & Randall) and Arya (1999)
Key: measuring watershed- stream linkage? 1. Lumped measures - %, #, density 2. Spatially-explicit - Euclidean distance 3. Network-based (directional, cumulative) - Strahler stream order - Length of stream line - Watershed area 4. Direct network-based - Discharge?!
1. Lumped % agricultural, % urban Ave road density Dam density (Moyle and Randall 1998) # mines Road length w/in riparian zone EPA An ecological assessment of the US Mid-Atlantic Region: A landscape atlas. Southern Rockies Ecosystem Project
1. Lumped (cont.) ArcINFO, Basinsoft (Harvey and Eash 1996): –Drainage area, shape, relief –# O 1 streams, main channel length, stream density
2. Spatially-explicit, Distance: As the crow flies (Euclidean)
3. Network-based Distance: As the seed floats (downstream)
Distance: As the fish swims (down & up stream)
Distance: Upstream length - mainstem (2) - arbolate ( )
Upstream 66 km Downstream 298 km Mainstem Upstream 37 km Network 16 km (down) 6 km (up) RWTools ArcView v3 extension
Direct measures Surrogate, e.g. Strahler order: The usefulness of stream order assumes, with a sufficiently large sample, that order is proportional to stream discharge – Strahler 1957 Ordinal data Not robust to data artifacts
Link watershed and network 1 to 1 relationship between stream reach and catchment Need robust method of delineation for large extents
Pilot area: Colorado, Yampa
“Smart bump” delineation 1. Reach catchment - flowdirection 30 m DEM - watershed from buffered hydrology (USGS NHD 1:100K) 2. Differentiate local ridges (artifacts) from true catchment boundary - “smart bump” using ZONALMIN 3. Remove conversion slivers at shared boundaries - regiongroup - if <10 cells, NIBBLE Currently, 1-2 days processing time per basin
Comparison of automated vs. hand-delineated 1.Randomly selected 111 (out of 2151 watersheds) 2.Computed area of automated vs. hand-delineated (“truth”) 3.RMSE = (in ha) 4.Mean error 2.4% 5.Challenges in defining commensurate watersheds
Hand- delineated “truth” watersheds 11% error
Reaches are linked to catchments 1 to 1 relationship Properties of the watershed can be linked to network for accumulation and networking operations Ordinal value (order) to real value (length, area, etc.)
Networking Import into ArcGIS Geometric Network Use networking tools, e.g. 1. Set flag 2. Trace upstream 3. Trace downstream
4. Direct metric: stream discharge Physical-based model: Q = Precipitation – Evapotranspiration Q is VMAD (Virgin Mean Annual Discharge)
USGS Stream Gauges
R 2 = P-value=3.407e-006
Surrogate Direct metric Order Area Discharge
Fragmentation and flow regulation Deynesius and Nilsson, Science (1994) – 77% of upper 1/3 of northern hemisphere rivers are strongly or moderately affected - F = regulated/total channel length - R = % of VMAD (cumulative reservoir live, gross capacity) RCL TCL
Alteration of natural flow regime Accumulation of dam storage Tributaries below dams mediating flow modification?
Flow modification How to measure relative modification of hydrologic regime? 1. Degree of modification to flow = cumulative annual flow – cum. dam max. storage: Q’ = Q-S 2. Proportion of modified to VMAD ( “natural”) flow: F = Q’/Q
High Dam “shadow” Reservoirs
Or/CO Table of output data Expand this to other factors: e.g., geology, vegetation, etc. Linked to rest of data
EMAP sites
Oregon + dam accumulation + overlap of catchment area
Within catchment hydrologic distance Moved from basins, HUCs and watersheds to stream reach catchments Within catchment: –Distance along hydro network distance (distance along the network upstream of pour point) –Allocation (using flat weight surface) 1
Challenges Data NHD 1:100K Dams – NID Processes natural flow diversions ET
Data: attribute errors Irrigation canals and pipelines incorrectly attributed as river/stream
Data: positional error Spatial location of dam locations is imprecise ? ?
Data: duplicates Stagecoach reservoir is duplicated – Challenges of understanding diverse datasets
Data: missing data? ?
Scale Dam on tributary that is not in 1:100K network
NID dams (red) > 50’ high, many other dams (in yellow) and other structures!
Dam data
Western Water Assessment, Figure 7
Network metrics Have foundation – direct measure Build on/refine existing metrics: –# first order streams –Main-channel length –Total stream length –Drainage density = stream length / catchment area Examine location within network and make available to statistical models
EMAP sites
Euclidean distance 1 2 Use x,y to create distance matrix Reasonable for broad-scale processes
Hydrologic distance 1 2 Follows stream network 3 4
Spatial weights W =
Functional distance Reflect distance weighted by: -Stream gradient -Geology -Land use -Etc A B C
Functional weighting W = 6 7 E.g., downstream hydrology
Connectivity matrix To/ from
Functional spatial weights StationDischarge (kacft) StationOrderArea overlap (%, km2) Length (m) Discharge 1 35 5 98%=2900/ % 1 45 5 11%=2900/ % 1 75 5 11%=2900/ % 2 31 5 0.4%=14/ % 2 41 %=14/ % 2 71 5 11%=14/ % 3 45 5 11%=2952/ % 3 75 5 11%=2952/ % 4 75 5 97%=25316/ % 5 72 5 0.5%=145/ % 6 74 5 0.5%=140/ %
Incorporate watershed conditions? W = 6 7 E.g., macroinvertebrates
Challenges Generating spatial weights matrix –O(n 2 ) O(n)? Functional (cost-weighted) spatial weights table
Products Watershed-reach network database GIS-based tool to develop functional spatial weights matrix ArcGIS extension for hydrologic network metrics
Thanks! Comments? Questions? Work funded by: US-EPA STAR Cooperative agreement CR awarded to CSU STARMAP: RWTools: