This data set contains variables characterizing the magnitude, variability and predictiability of flow at 72 USGS gauge stations on Iowa rivers. Using the methods of Poff & Ward (1989) and Poff & Allan (1995), these variables were created using mean daily discharge and partial peak data. Data from the entire period of record at individual USGS gauge stations were used to create the following variables: catchment area (area), daily mean flow (qmean), mean annual runoff (mar) - calculated as mean daily flow/catchment area, predictability of flow (daypred), coefficient of variation of daily mean flow (daycv), mean flood duration (flddur), predictability of flooding (fldpred), frequency of flood events (fldfreq), predictability of non-flooding (fldfree), predictability of low flow events (lowpred), period of year when low flow events failed to occur (lowfree), and extent of intermittency (zeroday). Detailed descriptions of the variables are given below.
Catchment area (AREA, km2)--The surface area of the catchment draining to the gage elevation.
Daily mean flow (QMEAN, m3/sec)--The average daily flow at the site over all years in the record.
Mean annual runoff (MAR, mm/yr)--Ratio of QMEAN/AREA. MAR represents the difference between annual evaporation and precipitation.
Coefficient of variation of daily mean flow (DAYCV, %)--This dimensionless index represents the average of the ratios between the annual mean daily flow and the standard deviation of the daily flows, multiplied by 100 and expressed as a percent. DAYCV describes overall flow variability without considering the temporal sequence of flow variation.
Predictability of flow (DAYPRED, %)--DAYPRED was determined using a index developed by Colwell (1974) which is based on information theory. When expressed as a percent, this index ranges in value from 0 to 100 and is composed of two independent, additive components: constancy (C), a measure of temporal invariance, and contingency (M), a measure of periodicity. The index can be used to express the degree to which flow states (i.e. quantity of discharge) are predictably distributed across specified time intervals. Predictability values are sensitive to definition of flow states, which are ultimately arbitrary. In this analysis, 11 categories were defined with a log2 series with boundaries at 2-3, 2-2, 2-1, 20, 21, 23, 24, 25, and 26 times modularized mean flow. Thus, the 11 flow states ranged from <12.5% of the mean flow to >640% of mean flow. Predictability measures are also sensitive to length of record.
Mean flood duration (FLDDUR, d)--The average number of days that flow remains above the flood threshold for a site.
Frequency of flooding events (FLDFREQ, floods/yr)--The average number of discrete flood events per year having a magnitude equaling or exceeding that associated with the 1.67 yr return-interval flood. The number of days that separate independent flood events may vary geographically; therefore, a 10-d period separating individual bankfull events was used as a criterion to identify separate spates.
Predictability of flooding (FLDPRED, dimensionless)--Maximum proportion of all floods over the period of record that fall in any 60-day seasonal window. This index ranges from 0.167 (random flooding) to 1.0 (perfectly seasonally predictable). For this metric the partial duration series was used. All instantaneous flows > 1.67 yr return-interval flow in the period of record were ordered according to day of the year they on which occurred, and the temporal distribution of this collapsed data set was analyzed for seasonal patterns. High flows occurring within 60 days of the beginning or end of the water year were considered to fall within the same season. In addition, the day of the water year marking the beginning of the 60-d period when FLDPRED was highest was recorded by the variable FLDTIME. This variable was not used as a primary classification variable, but was used to evaluate the range of timing of flood-onset within groups of hydrologically similar streams as identified by the cluster analysis.
Predictability of non-flooding (FLDFREE, dimensionless)--Maximum proportion of year (#days/365) during which no floods have ever occurred over the period of record. Again, the partial series was used and no-flood periods were allowed to pass through the end of one water year into the next.
Extent of intermittency (ZERODAY, d)--Average annual number of days having zero discharge.
Predictability of low flow events (LOWPRED, dimensionless)--Proportion of low flow events > 5-yr magnitude falling in a 60-d Seasonal window (as described above for flood predictability). Also the variable LOWTIME was derived to evaluate within-cluster variation in timing of lowflow-onset.
Period of year when low flow events failed to occur (LOWFREE, dimensionless)--Maximum proportion of year (#days/365) during which no 5-yr+ low flows have ever occurred over the period of record.
Poff, N.L. & Allan, J.D. (1995) Functional organization of stream fish assemblages in relation to hydrological variability. Ecology, 76,
606-627.
Poff, N.L. & Ward, J.V. (1989) Implications of streamflow variability and predictability for lotic community structure: a regional analysis of
streamflow patterns. Canadian Journal of Fisheries and Aquatic Science, 46, 1805-1818.