Stream_Classification_and_EFlows_(Tarver,_Mead)

Thoughts on Stream Classification and EFlows
Fred Tarver (NC Div. of Water Resources) & Jim Mead (Environ. Defense Fund)
9/25/2012

Session Law (SL) 2010-143 states “The Department shall characterize the ecology in the different river
basins and identify the flow necessary to maintain ecological integrity.” The purpose of characterizing
or classifying streams is to separate them into categories with each having a distinct algorithm for
determining the ecological flow (EFlow). We interpret the law to mean that classifying is a means to the
end of determining EFlows, not an end in itself.

A useful classification system should be:
1. Meaningful - groups streams according to discernible key characteristics producing an
algorithm(s) for determining the EFlows. One would expect the groupings to correspond to
different assemblages of organisms that respond differently to changes in flow, or have
different flow needs;
2. Reliable – produces consistent results;
3. Relatively simple and explainable; and
4. Applied using desktop information for a given stream, not requiring field investigations each
time a classification is needed.

The Biofidelity Project performed by the Research Triangle Institute (RTI) for the Environmental Defense
Fund (EDF) was intended to evaluate Item #1 above. Item #2 was first checked by comparing the
classification results generated using the Environmental Flow Specialists, Inc. (EFS) software for actual
U.S. Geological Survey (USGS) gage data to “virtual gage records” simulated with RTI’s WaterFALL
program. Because of problems encountered with Item #2, RTI did not proceed to completion with the
project (Item #1).

In retrospect, EFS was not given the appropriate specifications for developing a classification system.
The stakeholders should have stipulated that the classification system use only variables that are
typically simulated well by hydrologic models and focus more on the central and lower end of the range
of flows.

After some discussion, we are suggesting an alternative approach to stream classification for
determining EFlows for streams in North Carolina. This approach would rely heavily on the following
map layers:
1. Physio- or geographic Location – using traditional boundaries, for example: mountains, foothills,
piedmont, sandhills, inner coastal plain, outer coastal plain, or using an alternative approach.
2. Drainage Area – divide streams into 2- or 3-size groupings.
3. Elevation – a surrogate for temperature to subdivide mountain and foothill streams into cold
and cool water ecosystems.
Thoughts on Stream Classification and EFlows
Fred Tarver (NC Div. of Water Resources) & Jim Mead (Environ. Defense Fund)
2

4. Gradient – The Division of Water Resources’ (DWR) habitat modeling work indicates this has a
significant influence on the physical habitat conditions and how a stream responds to changes in
flow.

Physio- or geographic location (Variable #1) could be delineated with additional information regarding
flow yields per square mile to establish the demarcation between the different geographic units. We
suggest using mean annual flow per square mile, 30Q2 (analogous to September median) per square
mile, and 7Q10 per square mile. These flow statistics are commonly used by DWR and routinely
determined by USGS. They also focus on the lower and central range of flows – which are most affected
by removals of water, and which in turn is the impetus behind the modeling and planning efforts called
for by SL2010-143. High flows are ecologically important, but are only influenced by large-scale
reservoirs that have their own environmental review process for determining EFlow regimes. DWR has
always said that the EFlow algorithms developed in consultation with the Science Advisory Board (SAB)
are not intended to replace such project-specific studies.

The inner and outer coastal plain regions could be delineated based on: the “fall line” (piedmont vs.
inner coastal plain); and the limit of tidal influence (inner vs. outer coastal plain, downstream boundary
of hydrologic models).

If resources permit, we recommend that RTI develop these map layers for consideration of this
classification approach, and then perform the biofidelity test.1 It may be that the biological data and
fidelity test suggest modifications or partitions to the classification system, for example where to draw
the line between different ranges of drainage area, or between different physio- or geographic regions.
Perhaps a random forest type of analysis may provide some clarification. The final result would be a
map with each stream color-coded according to classification.

Some other points we considered during our deliberations that we wanted to note:

A. DWR’s habitat modeling approach to determine EFlows is probably not significantly influenced by
short, infrequent high flow events in the hydrologic record. With this in mind, the use of simulated
records from OASIS or WaterFALL to perform habitat modeling time series analysis should NOT have the
same difficulties with USGS versus virtual gage data as encountered with the EFS-generated stream
classes. To test this, it would desirable to obtain WaterFALL output for the nine habitat model sites
DWR has evaluated to date – the full, unregulated flow record with 1970 land cover. Habitat model
results for OASIS and WaterFALL could then be compared.

B. One type of ecological flow regime being considered by DWR is the “percentage of inflow” approach.
This was the right-hand grouping on the bar chart displays produced for the SAB, comparing the various
EFlow approaches. The middle grouping was the Tennant method approach (different percentage of
-----
1 Note: The Nature Conservancy is pursuing a classification scheme in the southeast similar to their effort in the northeast
(Olivero and Anderson 2008), as presented to the SAB on 02-21-12, which should provide useful map layers.
Thoughts on Stream Classification and EFlows
Fred Tarver (NC Div. of Water Resources) & Jim Mead (Environ. Defense Fund)
3

average annual flow), and the left-hand grouping was the minimum flow approach (7Q10 and monthly
median flows). The percentage of inflow approach is an attempt to recommend a “natural-like” flow
regime for the EFlow that limits flow alterations to a degree that maintains ecological integrity. The
other two groups of methods, on the other hand, start at the low end and add enough water until
ecological integrity is barely maintained. They also do not maintain any degree of natural flow
variability, which is ecologically desirable, and typically not evaluated by the available habitat indices.

Given the uncertainties that we are dealing with – uncertainties about hydrologic simulation models,
stream classification, and ecological response to flow changes – the percent of inflow approach is
desirable from the standpoint of being more conservative and protective. Taking a conservative
approach is an appropriate screening tool for planning purposes because it prevents overestimating
water availability for offstream users during the basinwide planning process. If users do not realize that
adequate water availability for offstream needs will not extend beyond their planning horizon, then they
could end up being short on time and funding and a site-specific study may create challenges toward
addressing future demands.

By being conservative, a percentage of inflow approach might also help address uncertainties about how
coastal stream ecosystems respond to changes in flow.

C. Lastly, the SAB has focused on water removals accomplished by simple withdrawals, not large
reservoirs that alter a large portion of the downstream hydrograph. This is justified by the extensive
project-specific studies required of large reservoir projects, and the fact that each reservoir operation is
unique and does not lend itself to a generic ecological flow algorithm. That said, there is a possibility
that a percentage of inflow approach for determining ecological flows could be applied to potential
reservoirs for basinwide planning purposes, not just to simple water withdrawals.