Lesson 15f: Stream Flow Estimates in NHDPlus HR

This lesson will cover Stream Flow Estimates in NHDPlus High Resolution.

Enhanced Runoff Method, or EROM is the method used to compute estimates of the mean annual flow for the NHDPlus flowline features in the NHDPlus HR network. The flow, in cubic feet per second, reflects the average annual flow from 1971 to 2000 time period.

Flow and velocity estimates are important parts of the NHDPlus HR package, and are included for all network flowlines. This allows the NHDPlus HR to be used in riverine fate and transport models—models which simulate the movement of contaminants as they move through the river network.

This presentation covers the basics of how flow rate estimates are calculated.

The NHDPlus HR packages are delivered as a geodatabase. Each geodatabase represents a vector processing unit, or VPU. NHDPlus HR flow rates are calculated for all network flowlines in a VPU.

There are many tables included in the geodatabase. We will consider the following tables in this presentation:

The first table is the NHDPlusEROM MA, or EROM Mean Annual table. This is the table that contains flow estimates for all NHDFlowline features in the NHDPlus HR network.

The second table is the NHDPlus EROM QA MA, or EROM Quality Assurance Mean Annual table. This table contains the gage and EROM flows for users to do their own analyses. The file layout is designed to facilitate graphical and statistical analyses. It’s useful for users who want to look at graphs or additional statistics about the stream gages used in the flow estimates.

The third table is the NHDPlus EROM QA Rpt, or the EROM Quality Assurance Report. This report contains comparisons of the EROM flow estimates and the observed flows from real stream gages. The report is stored as a table.  Note: No example shown.

The first table we will discuss is the NHDPlus EROM Mean Annual table. As you can see, the table has many attributes. We will cover just the five highlighted fields.

The first is “QAMA”. The alias for this field is “Flow Est A Runoff MA”
or “Flow Estimate A, Runoff, Mean Annual”. This is the initial estimate of the flow.

Second is “QBMA”. The alias for this field is “Flow Est B Excess ET MA”,
or “Flow Estimate B, Excess Evapotranspiration, Mean Annual”.

“Q” is a variable in hydrology equations that stand for “discharge”, which is why you are seeing Q in the field names.

The third relevant field is “QCMA”, alias “Flow Est C Ref Gage Regression MA”,
or “Flow Estimate C, Reference Gage Regression, Mean Annual”.

Fourth is “QDMA”, alias: “Flow Est D Addition Removal MA”,
or “Flow Estimate D, Addition Removal, Mean Annual”.

Finally, “QEME”, alias: “Flow Est E Gage Adjusted MA”,
or “Flow Estimate E, Gage Adjusted, Mean Annual”.

Each of these five fields represents an iteration to refine the accuracy of the estimated flow for the flowline. Let’s walk through a real world example to illustrate how each of the five fields refines the flow estimate.

The explanation for ‘Estimates of the flow’ is broken down into steps A through F:

Steps A and B compute “natural flow.”

Steps C, D, and E adjust “natural flow” values based on observed gage data and known flow transfers.

Step F performs QA/QC analysis.

The image on the slide shows the drainage basin of Pitkin Creek, a creek about 5 miles west of Vail Ski resort in Colorado.

Note the catchments of all the flowlines have been aggregated, so you can see the entire drainage of Pitkin creek in pink.

Locations of stream gages are shown as red dots.

In step A, initial estimates of mean annual flow are based on a grid from a flow balance model of surface runoff, published by McCabe and Wolock in 2011.

Each 900 m x 900 m grid cell symbolizes natural runoff, or mean water-year runoff in mm.

Zooming in to the Pitkin creek drainage, you can see the individual runoff grid cells. The flow rates for all flowlines are captured in QAMA, “Flow Estimate A, Runoff, Mean Annual”. In this, and following screenshots, only the record for the most downstream flowline of Pitkin Creek is shown. The QAMA attribute is shown in yellow on the far left: 6.1 cubic feet per second.

In step A, the runoff raster is overlaid with the NHDPlus HR catchments to compute runoff within each catchment. The catchment runoff values are conservatively routed downstream to arrive at the first estimate of stream flows for each networked NHDFlowline feature.

The flow estimate represents the flow at the bottom of each flowline.

Step B represents adjustments for losses due to excessive evapotranspiration.

This method, developed by McCabe and Wolock, considers the total available water in a given catchment to compute additional losses due to evapotranspiration.

Evapotranspiration losses can exceed the total water available in a catchment, resulting in a net loss in stream flow.

This loss of instream flow is a significant observed phenomenon, especially in arid areas west of the Mississippi River.

Estimates of the loss made in this step are subtracted from the flow estimates
and are stored in QBMA, “Flow Estimate B, Excess Evapotranspiration, Mean Annual”

This estimate gets us a little bit closer to “natural flow.”

QBMA is shown in yellow in the second column from the right.

Notice that the first record in the attribute table shows 6.1 cubic feet per second. This means that there was no appreciable loss of flow due to evapotranspiration in this area.

This is expected because Pitkin creek is in a mountainous area at higher elevation. Just 150 miles west of Denver, near the town of Grand Junction, Colorado.

Step C is a regression analysis using “reference stream gages” to provide a further adjustment to the flow estimates calculated in step B.

A “reference gage” is a stream gage that is minimally affected by human activities. For example, there are no dams or other flow alterations upstream.

Gages used in EROM, including Reference gages, are screened based on two additional criteria:

First, the NHDPlus HR drainage area for the gage must be within a certain percentage of the NWIS-reported drainage area. NWIS is “National Water Information System” 

Second, the gage must have a minimum of 10-years record in the 1971 to 2000 time period.

The reference gages used for each Vector Processing Unit (VPU) are shown in the “EROM QA MA” table, which we will discuss in one of the following slides.

A log-log regression is calculated to further refine the flow estimate to represent natural conditions.

The gage adjusted natural flow value is captured in QCMA (FlowEstCRefGageRegressMA) attribute.

In this example, the stream gage shown here is the USGS stream gage at Pitkin Creek.

Data and descriptive information for the gage can be found on the NWIS web site.

If we look up this stream gage at NWIS, we see its Gage ID number, and it’s official name: “Pitkin Creek Near Minturn Colorado”.

We also see that NWIS records the Total Upstream Drainage for this gage at 5.32 square miles.

Finally, the gage has more than 10 years of continuous record.

Therefore, the gage at Pitkin creek is suitable as a reference gage, and can be used in the regression equation to estimate flow

We can also calculate the upstream drainage area from this gage point using the NHDPlus HR in a geographic information system, or GIS.

GIS calculates the sum of the areas of all the catchments upstream of this point as 5.35 square miles. The calculated value is within 20% of the upstream drainage recorded for this stream gage in NWIS.

A log-log regression is calculated using the reference gage flows to further refine the flow estimate to represent natural conditions.

The gage adjusted natural flow value is captured in QCMA, “Flow Estimate C, Reference Gage Regression, Mean Annual” attribute.

This is shown in the center yellow column of the attribute table.

Notice the flow for the first record is 5.7 cubic feet per second.

So the reference gage regression estimates the flow at the gage site to be a bit lower than in Step C.

The QCMA values are the best estimate of “natural” flows.

Step D adjusts the streamflow for flow transfers, withdraws, and augmentation.  Manmade additions, removals and transfers are found in the NHDPlusFlowAR table.This table is empty in the Beta version of NHDPlus HR, and will be built over time, based largely on user input.

Adjustments in this step are made to QCMA.

The newly adjusted values are stored in QDMA, alias “FlowEstDAdditionRemovalMA.

If we look at the data in Pitkin Creek, we will find that the adjusted values stored in QDMA are the same as the values calculated in step C for the QCMA attribute. The reason for the unchanged values is the absence of any manmade streamflow adjustments for Pitkin Creek in the NHDPlusFlowAR table.

Step E further adjusts flow estimates based on observed gaging station data. In this step the network features that are upstream from the gages are adjusted for observed gage-base flow. The flow for the downstream features, in turn, are calculated using the adjusted incremental flows for affected catchment areas.

Thus, step E impacts estimates downstream from the gages to better reflect flow alterations not taken into account in the first four steps.

It is possible for the flow values to decrease downstream at flowlines where a gage is present.

Gage selection criteria in this step is the same as in step C: drainage areas must be within 20% and must have continuous record of 10 years. Only gaging stations linked to the NHDPlus High Resolution network are used to adjust flows.

The flow estimates after adjustments in step E are considered the best NHDPlus High Resolution flow estimates of actual flows.

The value is stored in QEMA, alias “FlowEstEGageAdjustedMA.”

This slide shows the final flow estimate, QEMA, of 6.1 cfs for Pitkin Creek in the last column.

Note this estimate is very close to the initial estimate based on runoff only, but this is not always the case. Here are the final QEMA values in the Pitkin Creek area. You can symbolize QEMA values in class intervals. In the example of Pitkin Creek, we used 20 classes. As you can see, such classification gives you an effective map display where major and minor streams are displayed differently on the map based on their flow estimates.”

Step F is a quality assurance step to measure the accuracy of gage adjusted flow estimates on ungaged NHDFlowline features.”

Because Step D uses all gages, the flow estimates at the gage locations will always match the gaged flow values. This means that any statistical analyses on the Step E flows compared to gage flows will always be a “perfect” match.”

Step F randomly removes about 20% of gages from the gage adjustment process and repeats the gage adjustment process on the remaining 80% of gages. Step F then compares the adjusted estimates to the actual gage flows.

The “EROM QAMA” table captures which gages were used in the QA analysis. The attribute “GageRef” contains the value of “Yes” for the used gages.

The EROM QA MA table and EROM QA Report may be used to evaluate the accuracy of the reference gage regression and gage-adjusted flow estimates. 

Using values found in the EROM QA MA table, the graph on this slide shows the comparison of the actual Log10 mean flow at the gages and the Log10 of the EROM mean flow estimates. Gage Flows are on the x-axis and EROM flow estimates are on the y-axis.

The blue diamonds are the runoff flow estimates. The pink squares are the flows adjusted with the reference gage regression. The red line is where the gage and EROM flows would be equal. Note how in this region the runoff estimates consistently under-estimate the gaged flows.

Stream flow QA information is included in two tables.

The first output is EROM QA MA table. It contains statistical description of initial estimate of streamflow from runoff in a table form. The file layout is designed to facilitate graphical and statistical analyses. It is useful for users who want to look at graphs or additional statistics for only the reference gages.

The second output is the   report. The report contains comparisons of the EROM flow estimates and the observed gage flows. The report is stored as a table.

To summarize this lesson, we learned how Stream Flow Estimates get calculated in NHDPlus High Resolution. It’s a global method to assign flow rates for the entire United States.

Many other methods for flow estimation exist, especially on a more local level. For example, USGS StreamStats program calculates flow estimates using smaller regional regressions, usually presented by state.

There are many potential uses for EROM flow estimates. For example, flow, in combination with the flow network, allows for time of travel studies. Flow, in combination with catchments or watershed boundaries, can be used to estimate water budgets for certain areas.

Finally, flow estimates can be symbolized to create effective cartographic products, as we have seen in the example of Pitkin Creek.

For further questions related to NHDPlus High Resolution or any NHD products, please contact USGS support email: nhd@usgs.gov.