Stream discharge in gaged watersheds at the Andrews Experimental Forest, 1949 to present

ORIGINATOR: Jack S. Rothacher
OTHER RESEARCHER: Julia A. Jones, Jeffrey J. McDonnell
ABSTRACTOR: Donald L. Henshaw
METHOD CONTACT: Craig Creel, Greg Downing
FORMER INVESTIGATOR: R. Dennis Harr, Gordon E. Grant, David A. Post, Richard L. Fredriksen, Alfred B. Levno
All work was carried out by the U.S. Forest Service Pacific Northwest Research Station, Watershed Project, Corvallis, Oregon, from 1952 until 1980. Since 1980 the PNW Research Station has worked in collaboration with the NSF Long-Term Ecological Research (LTER) program in the collection and processing of this data.
28 Nov 2005
6 Mar 2013
Experimental watershed studies, Floods, Forest ecosystems, Harvest methods, Hydrologic processes, Hydrology data, Hydrology/Water, Long-term studies, Radio telemetry, Streamflow, Timber harvesting, Disturbance
Originally streamflow from sets of adjacent small watersheds was monitored to examine changes in streamflow from management practices, primarily clearcut logging and other harvest methods, with one experimental watershed in each set serving as a control (Andrews WS 1, 2, 3; WS 6, 7, 8; and WS 9, 10). Undisturbed watershed with old growth forest are WS 2, WS 8, WS 9, and Mack Creek. Treated watersheds are WS 1 (clearcut 1962-66, burn 1967), WS 3 (roads 1959, patch clearcut 1963), WS 6 (clearcut and burn 1974), WS 7 (shelterwood cut 1974, remaining overstory removed 1984), and WS 10 (clearcut 1975). Additional objectives include providing baseline data for affiliated nutrient, water chemistry, and sediment transport studies, and more recently stream cross-section, hyporheic, succession and ecophysiology studies. Models have also been extensively used with this data to characterize the hydrologic regime of old-growth forests at different elevations. Larger order basins are also monitored within the Andrews: Mack Creek and Lookout Creek (USGS).
Experimental Design - HF004:

The H.J.Andrews Experimental Forest comprises the Lookout Creek Basin. Lookout Creek as well as nine smaller watersheds within the basin have gauging stations. The gauged sites consist of three sets of watershed experiments (WS 1, 2, 3; 6, 7, 8; and 9, 10). Drainage areas for the adjacent (paired) watersheds range from 9 to 100 ha and elevations range from 460 to 960 m. Mack Creek, a larger (580 ha), old-growth watershed and Lookout Creek are used in conjunction with the small watersheds to characterize the hydrologic regime of old-growth forests at higher stream order and different elevations.

The original design used a paired watershed technique to detect changes in streamflow due to watershed treatments. Treatments include 100% clearcut (WS 1, 6, and 10), patch cut with roads (WS 3), and shelterwood cut with follow-up overstory harvest and eventual thinning (WS 7). Watersheds 2, 8, and 9 are forested controls. With this technique a hydrologic variable in one watershed is compared with that in another watershed during a pretreatment or calibration preiod to establish a relationship between watersheds for that variable over a range of conditions. Then one watershed of a pair is treated or altered in some way by timber harvest activities, while the other remains an undisturbed control. Post-treatment measurements of the hydrologic variable are compared with a prediction of the variable based on the pretreatment relationship to evaluate changes caused by treatment. A control is paired with each logged watershed for sets WS 1, 2, 3 and WS 6, 7, 8 to form two pairs of watersheds.

Field Methods - HF004:

Nine watersheds within the Andrews are gauged to measure stream height or stage. The records of stage are converted to stream discharge by applying station-specific discharge rating curves. Historically, sites have been visited on a 3-week basis and "check sheets" guide the data collection and maintenance activities for the visit. Beginning with wateryear 1993, sites are visited on a weekly basis for a quick check, with the full check every 3 weeks. The instrumentation used to measure and record stream stage has changed through the years in response to changes in data collection technology, data management considerations, and end-user needs including the need for real-time distribution of data. Leopold-Stevens A-35 chart recorders were the primary collection mechanism from the beginning of record through WY 1998, although Fischer-Porter punched tape recorders were used in lieu of charts from 1967 to 1975. Stage height is currently measured with a Model 2 Stevens Instruments Position Analog Transmitter (PAT) recorders controlled by a data logger. Charts are still maintained to provide a visual backup record. The discharge measurement is re-initialized every visit to a reference measurement of stream height taken by a hook gage located in a stilling well, or in the case of Mack Creek, a tape measurement of water height in the flume and fish ladder. Radio telemetry is employed at WS 1,2,8,10, and Mack Creek to relay information to the base station at Andrews Headquarters on an hourly basis. Provisional data is available in near real-time on the Andrews webpage for the radio telemetry sites. Additionally, the Lookout Creek gaging station is maintained by the U.S. Geological Survey and satellite telemetry provides near real-time data on the USGS webpage. Records are maintained by the USGS but are also stored and available through the Andrews LTER.

The regular and frequent system of checks indicate any problems with the recorders. Every week a hookgage measurement is taken and compared to the recorder readings. Significant variations from the hookgage would indicate changes in recorder operation, but factors such as chart spooling, or stilling well response time during rapidly changing stage levels must also be considered. The hookgage position/elevation in relation to the flume is surveyed every year. Measurements of the water level in the flume also serve as a rough check of hookgage accuracy.

Streamflow is measured in broad-crested concrete trapezoidal flumes designed to withstand movement of large wood and located to minimize subsurface water leakage. Beginning in 1997, compound weirs are used at our gauging stations in the summer (all except Mack Creek). They are metal v-notch weir plates that are mounted to the existing trapezoidal flumes/weirs to increase measurement sensitivity during the low flow period. We remove the weirs in the winter to allow passage of debris and to measure flows beyond the range of the v-notch weirs.

During the summer of 1995 a fish passage was built around the flume at the Mack Creek station. A compound weir (120 degree v-notch at the bottom of the weir plate with a rectangular area on top.) was installed in the passage to measure the water level. The fish passage stage measurements began on October 5, 1995. Mack Creek streamflow is the sum of the flows from the regular flume and the fish passage compound weir from this time forward.

A rating curve has been specifically developed for each weir/flume from on-site rating point collection. Various methods have been used over time, but currently a velocity meter is typically used across the cross-sectional flume area to calculate flow for a given stage height. Rating curves change with the configuration of the flume, e.g., summer v-notches have their own curves. Rating points are routinely collected to check accuracy of existing rating curves.

Currently, stage height is measured with a Model 2 Stevens Instruments Position Analog Transmitter (PAT) recorder controlled by a data logger (Campbell Scientific CR10X or equivalent). This is a float pulley potentiometer device, modified to read the potentiometer and not the current. Radio telemetry is employed at WS 1,2,8,10, and Mack Creek to relay information to the base station at Andrews Headquarters on an hourly basis through a Campbell Scientific radio telemetry unit. Prior to 1996 streamflow records were derived from stage data collected by Stevens Type A (A-35 or A-71) recorders. The Stevens Type A recorder is a float- operated recorder that provides a permanent and continuous graphic record of stream stage fluctuations. A clock movement controls the rate at which a strip chart is advanced. The rise and fall of the float moves a marking stylus laterally across the chart. These recorders are still operative and the A-35 charts serve as the backup media.

The Type A recorder has a precision of .001 feet. Accuracy of the Type A recorders is influenced by small measurement errors due to float lag and line shift. These errors have been minimized by the selection of proper floats and counterweights. The estimated error value over a range of 5 feet for line shift is .001 feet, and the estimated error due to float lag is .004 feet (Error tables provided by Stevens). Accuracy is also influenced by hook gage or tape reference height measurement variations. Rapidly fluctuating water levels or observer bias can cause variations up to +/- .001 feet during repeated measurements. The PAT potentiometer can malfunction resulting in electronic spikes or dead zones in the range of resistance, and we inspect annually to assure full range of resistance. The potentiometer has no adjustments, so it must be replaced if problems occur.

The v-notch plates were fabricated by a local metal shop. We provided rough drawings of what we wanted and very detailed specifications for the v-notch openings and their position within the existing flumes/weirs. We then worked with the fabricator to design the mounting system. Once the basic v-notch plates were complete, the fabricator then did a custom fit of each one in the field to ensure all specifications were met. Since we were confined to the existing weir structure, some of the elements of a 'true' v-notch weir could not be met (example: proper pool depth for expected head, and plate not being exactly vertical.). Because our weirs were different we could not use the theoretical discharge formulas, so we had to develop our own discharge curves for each site.

For the Mack Creek fish ladder a MTS Level Plus model LT420 magnetostrictive transmitter mounted in a stilling well is used to measure the stage. The transmitter consists of a ferromagnetic waveguide protected by a solid rod, a electronic assembly that determines the water level based on waveguide behavior, and a float containing a set of magnets that "ride" the outside of the waveguide protective rod. The transmitter has a precision of .001 feet, and an accuracy of .002 feet over a range of four feet. A tape reading of the water level in the stilling well is used as the stage reference.

Instrumentation: Model 2 Stevens Instruments Position Analog Transmitter (PAT) recorder controlled by a data logger (Campbell Scientific CR10X or equivalent); Stevens Type A (A-35 or A-71) recorders; MTS Level Plus model LT420 magnetostrictive transmitter; Fischer-Porter punched tape recorders
Citation: Rothacher, Jack; Miner, Norman. 1966. Accuracy of measurement of runoff from experimental watersheds. In: International symposium on forest hydrology: Proceedings of National Science Foundation advanced science seminar; 1965 August 29-September 10; Pennsylvania State University, PA. New York, NY: Pergamon Press: 705-713.
Algorithms - HF004:

Before Wateryear 2012, the set of collected raw gage height data are essentially "key (turning) points" that is, points where the slope or trajectory of streamflow height changes. This set of key points have historically been identified programatically from high temporal resolution data collections and through the process of hand digitizing. Points are saved when a new point does not fall on the same slope as the previous two saved points. More recently points are collected when the stage changes by .001 ft. A critical aspect of this approach is that the raw gage height data are a set of points spaced irregularly in time, e.g. more points for periods of more rapid change such as during storms or during wet periods.

The FLOW program (Entity 5) allows the user to output this data at any specified time interval. This program uses linear interpolation on gage height (stage) to calculate stage at the requested time step, and the data is run through the rating equations with proper calculus to calculate flow for each interval. Beginning in 2012, data is collected at 5 minute intervals and the notion of finding and outputting these "key turning point" values is abandoned.

The raw data are corrected by a program that adjusts measured gage height to be consistent with a set of concurrent, independent observations from the hook gages located adjacent to each gage house. When hook gages are not installed (currently only Mack Creek flume and fish ladder), direct measurement of gage height is made in the flume with a rule. The corrected gage heights are converted to discharge using rating curves with a Fortran program WEIRD01. Throughout the historic digitizing process of A-35 charts (until 1997), time is corrected to the nearest one-half hour (essentially 15 minute resolution). Since 1997 time is reset on data loggers when it differs from true time by more than 5 minutes. Missing periods have generally been estimated using regression equations with a paired watershed. All data is graphically displayed and corrected for general noise, logs caught in flume, etc. Range and date checks are performed on final data. Current processing and graphical display is conducted using Visual Foxpro programs (STRMCHK), along with the historic Fortran program.

Since data is stored as a series of time and stage values reduced to a series of points representing straight line segments, data can be aggregated at any time interval. Discharge data desired for a particular period (e.g. 15 minute, hourly, etc.) can be determined by interpolating between the stored stage height values for the times of interest and calculating flow through the rating equations. A C-language computer program that allows the gage height to be identified at pre-determined time periods, and calculates the discharge directly from the rating equations for these points was developed in 1994 by JinFan Duan. The program was later rewritten in Perl by Duan (2000) and customized and revised for web use by Kyle Kotwica in 2003 (FLOW). The program is currently interactive and available on the Andrews website to dynamically produce even-step high temporal resolution data, e.g., 15 minute, hourly, etc. for any set of dates for all watersheds (except the USGS maintained Lookout Creek gauge).

Background on rating curves: Each rating curve is fitted to a set of calibration points. Some flumes (e.g. pre-built H-flumes) came with a manufacturer's rating curve; other flumes (e.g. the concrete trapezoidal flumes in use today at all Andrews gages) require the user to develop the curve. The technology for taking calibration points, and hence, possibly, their accuracy and precision, has changed over time at the Andrews. Calibration points consist of synchronous measures of stage height and velocity. These measurements were taken by a velocity-head rod in the early years (1953-1973). While hook gages had been installed, only the flume measurements had been recorded with velocity measurements from 1953-1957. The dye dilution method was used to develop the WS 9 & 10 rating curves from 1975-1977. In more recent years (1983-present) velocity measurements have been taken with a velocity meter. Volumetric or "bucket" samples have also been used to measure velocity during low flows since 1973. The technology for fitting curves also has changed over time. In the early days (prior to computerized statistical packages) curves were fitted by eye and rating tables were developed (WS 1, 2, and 3). Subsequently curves have been fitted using regression techniques, typically with a log-linear or piece-wise log-linear function. However, in some cases other functional forms were used (e.g. a reverse sigmoidal (cubic) curve for Watersheds 9 and 10 in the 1970s). Piece-wise functions have been used to account for bends in an otherwise straight log-linear relationship inferred from the calibration points; some rating curves have as many as seven segments.

Instrumentation: A Model 200 Flo-Mate velocity meter is used to collect rating points for the development and checking of rating equations.
Processing Procedure - HF004:
Description: Beginning with WY 1960, all A-35 charts from the Stevens recorder were digitized at the Coweeta Hydrological Lab. Beginning with WY 1986, the process was moved to the Corvallis FSL using a Summagraphics MM1812 digitizer. Early chart data from 1953 to 1959 that had originally been summarized daily by manual scanning techniques were redigitized in 1988. Consistent digitizing practices have been maintained for all chart data. Fischer-Porter punched tape recorders were used in lieu of charts from 1967 to 1975, and punched tape output was used as the raw record instead of the A-35 charts. This resulted in 15 minute data output during this period. A-35 chart digitizing resumed after WY 1975. Since 1997, Campbell scientific data loggers replace chart digitizing as the primary source of streamflow data, and the charts are still collected and maintained as a backup when electronic data collection fails.
Instrumentation: SummaGraphics 1812 digitizer is able to record up to 40 linear points per mm at highest precision. In practice the digitizer was able to produce raw data points of stage height within .003 ft.; Fischer-Porter punched tape recorders
Processing Procedures - HF004:

An hourly record of Lookout Creek streamflow has been reconstructed from WY 1950 (10-01-1949) through WY 1986 (09-30-1986). Prior to 2013, Lookout Creek streamflow data at less than a daily time-step had only been available through the USGS in 30 minute steps beginning with WY 1987 (1 Oct 1986) and 15 minute steps beginning in WY 2010. The hourly record for this early period (1950-1986) was reconstructed and placed online in Feb 2013. Stage height was gathered from the following sources:

  • WY 1950 – WY 1955: USGS maintained A-35 charts (Stevens Type A stage recorder)
  • WY 1956 – WY 1963: USFS PNW maintained A-35 charts (Stevens Type A stage recorder)
  • WY 1964 – WY 1965: USGS maintained A-35 charts (Stevens Type A stage recorder)
  • WY 1966 – WY 1974: USGS maintained Fischer-Porter paper punch-tape stage recorder at 2-hour intervals
  • WY 1975 – WY 1986: USGS maintained Fischer-Porter paper punch-tape stage recorder at 1-hour intervals

Chart data was digitized using a Summagraphics 1812 Digitizer with a precision of .03 feet (1949-1965). Punch tape data stored on computer printouts was entered (1965-1986). All 24 USGS Lookout Creek rating tables used over the period of 1950-1986 by the USGS to generate stored daily values and seasonal peak flow were gathered and entered. These table data were used as the basis for fitting 24 sets of rating equations that calculated flow in cubic feet per second (cfs) from gage height (stage) in feet on an hourly basis. Each set of rating equations is a series of piecemeal linear power curve equations in the form of: ln y=ln a +b ln x, where y is flow in cfs and x is stage in feet. Using this equation form allows the gage height to be linear interpolated on an hourly basis. This method is necessary for reducing the 16 years of digitized chart data, but was also used to linearly interpolate every hour when 2- hour interval punch tape data was collected (1966-1974).

As a means of checking the accuracy of the rating curves and digitization procedures, a new daily record is generated from the new rating curves and compared to the published USGS daily values. All of the new generated daily values fall within 5 percent of the USGS daily value when the flow in cfs is greater than 30, and fall with 8 percent of the USGS daily value when the flow is less than 30. Eighty-three percent of all new daily values fall within 3 percent of the USGS published value. Some differences between the newly constructed record and the USGS published values were likely caused by post-processing corrections or shifts in the original gage height data. Gage height data were adjusted accordingly to better match the USGS flow record for noted periods where a previous correction was evident. Note that no USGS record exists for WY 1956-1963 and this type of comparison checking was not possible.

The daily values published with this data set are in fact the USGS published values, however, daily max and min values are added to the daily record representing the maximum and minimum hourly values for the day. Also note that no USGS published data is available for Lookout Creek for WY 1956-1953, and the daily record for this period is only available with this data set. Also, hourly data before WY 1987 is only available with this data set (and not available through the USGS).

Caveats: 1) The 1982 wateryear data are estimated as the primary data was missing and the hourly data values are estimated from published USGS daily max-min values for this year. 2) While the Forest Service was operating the gage from 1956-1963, only one stream calibration was made to create a rating table in 1961. The 1955 USGS rating table was used in 1956, but water years 1957 and early 1958 seem to not fit very well in either the 1955 or subsequent 1961 rating table, although the 1961 table was used and seems a better match. 3) Water years 1958, 59, 60, and 62 were originally hand scanned and hourly data is only approximate, but all storm peak flow points are accurately noted.

Instrumentation: Stevens Type A stage recorder; Fischer-Porter paper punch-tape stage recorder; Summagraphics 1812 Digitizer
Citation: Carter, R. W., W. L. Anderson, W. L. Isherwood, K. W. Rolfe, C. R. Showen, W. Smith. 1963. Automation of Streamflow Records. U. S. Geological Survey Circular 474. U.S. Government Printing Office, Washington, D. C. 22p.

USGS data: The USGS maintains the Lookout Creek Gauging Station (14161500). High resolution 30 minute data is currently available beginning WY 1987 and 15-minute resolution data beginning in WY 2010 though the USGS. The Lookout Creek station was maintained by the U.S.Forest Service PNW from WY 1956 through WY 1963 (and this data is not available through the USGS), while the USGS has maintained the station at all other times beginning in October, 1949. However, all Lookout Creek streamflow data is available through this HF004 data set. Hourly data has been reconstructed from WY 1950-WY 1986 and daily max-mins have been integrated into the daily record, Entity 2 (see description under processing procedure method).

The USGS also maintains a station at Blue River near Tidbits Creek (14161100). Note that a Blue River near Blue River USGS site (14162000) was discontinued in Dec, 1963, due to the building of the dam. The data provided for upper Blue River here includes the Blue River site (14162000) through WY1963 and the Tidbits site (14161100) beginning in WY 1964.

Cautions: The Lookout Creek station was maintained by the U.S.Forest Service from WY 1956 through WY 1963 Three separate calibrations (rating tables) were developed during the seven years. The USGS prefers that this be done every year, and does not acknowledge any data during the Forest Service maintained period.

Continuous streamflow data have been collected over varying periods of record in eight first to second-order watersheds, one third-order watershed, and the fifth-order watershed in the H. J. Andrews. Watersheds 1, 2, and 3 comprise a low-elevation (442 m to 1082 m) set of experimental basins with Watershed 2 serving as the control. Watersheds 6, 7, and 8 comprise a high-elevation (863 m to 1190 m) set of experimental basins with Watershed 8 as the control. Watersheds 9 and 10 are a low-elevation pair of small watersheds with Watershed 9 as the control. Mack Creek is a control basin and has not been harvested. The gauged portion of 5th-order Lookout Creek comprises nearly all of the Andrews Forest.
Andrews Experimental Forest and adjacent small watersheds (WS 9 and 10), Willamette National Forest, Western Cascade Range of Oregon
Ground condition