Summary of Updated Model Calibration


After macrophytes and enhanced pH buffering were incorporated into the upper Klamath River model, the model was recalibrated by adjusting selected parameter values to obtain a good match between measured data and model output. Most parameters were unchanged from the original model calibration and remain as documented by Sullivan and others (2011). Those parameters that were adjusted are shown in table 5. The new macrophyte parameters, as used in the upper Klamath River model, are shown in table 2 and the enhanced buffering parameters are shown in table 6.


During macrophyte calibration, the macrophyte field data were used to provide insight into the general patterns of density, spatial and temporal distributions of biomass, and macrophyte species characteristics. The model was not calibrated directly to the macrophyte field data because the macrophyte field data were collected in a different year (2011) than the model years (2006–09). Examples of the modeled distribution of the three macrophyte groups through the model domain on July 2 and August 17 for model year 2007 are shown in figure 6. Although most constituents in CE-QUAL-W2 are laterally averaged across the channel, macrophytes are modeled in a quasi three-dimensional mode so that macrophytes can be modeled at different depths from bank to bank across a model segment (Berger, 2000; Berger and Wells, 2008; Cole and Wells, 2008). As in the field data, the modeled pondweed density was highest in early summer and diminished by August, whereas coontail and common waterweed were simulated with a higher density in late summer and were more common in the downstream reaches near Keno Dam.


Macrophytes affect the cycles of oxygen, nutrients, organic matter, and pH. In the updated model, the total annual production of dissolved oxygen by macrophytes is approximately the same order of magnitude as that by algae (fig. 7). Oxygen consumption by respiration is higher in macrophytes compared to algae, due in part to the fact that the macrophyte oxygen respiration parameter was adjusted to account for the large population of snails within the macrophytes. Dissolved oxygen production and consumption by macrophytes begins earlier in the year than oxygen production and consumption by algae, largely because the pondweed macrophytes are early season species, whereas the largest algal blooms occur in late June and early July.


pH buffering by ammonia, phosphates, and dissolved organic matter was enabled in the recalibrated model, but buffering by particulate organic matter was disabled (table 6). Particulate organic matter, much of it derived from dying algae, may have some acid/base properties, but being enclosed in particles likely inhibits its reactivity with the bulk river water. Algae and macrophyte material may also have acid/base properties, such as cell walls with carboxyl groups (Chojnacka, 2010), but this buffering was also assumed to be small relative to buffering by dissolved organic matter. Nonconservative alkalinity was turned on, and two monoprotic acids were used to simulate the dissolved organic matter buffering—one to represent carboxylic acids and one to represent phenolic and amine groups together.


The enhanced pH buffering algorithms had a notable effect on modeling pH in summer and fall. Without the enhanced buffering, the modeled pH was generally lower and subject to larger daily variability compared to measured pH (fig. 8). The addition of macrophytes also improved pH modeling, especially in the lower part of the reach near Keno Dam where macrophytes were more abundant (fig. 9). With both macrophytes and enhanced pH buffering implemented, the model was able to simulate pH well with mean absolute errors (MAE) of ≤0.34 for years 2006–09 (fig. 10; table 7). The inclusion and calibration of pH in the upper Klamath River model will now allow model scenarios to examine the effect of management and other system changes on pH, which is a water-quality limited constituent in this reach. 


The updates to the model affected the error statistics of other constituents modestly. The most improved error statistics were for dissolved oxygen, with an average MAE improvement of 0.1 mg/L, and particulate carbon, with an average MAE improvement of 0.2 mg/L. Average error statistics for other constituents were essentially the same compared to the original USGS model (Sullivan and others 2011). In addition to the model performance metrics included in table 7, measured and modeled pH and dissolved oxygen time series are shown in figures 10A–D and 11A–D, respectively, for calendar years 2006–09. Algae, nutrients, organic matter, and bottom sediment are shown in figures 12A–D for the calendar years 2006–09. Modeled temperature and specific conductance are not shown here; those plots are available in the original USGS Klamath River model report (Sullivan and others, 2011).


When interpreting measured/model comparison plots and error statistics, remember that model output and measured data are at the same time and location but not at the exact same scale. For example, measurements from a continuous monitor probe represent a small volume near the probe tip. Model output is from an entire model cell that might have dimensions of 1,000 ft × 500 ft × 2 ft near-surface. Measured data and model output should show the same large-scale temporal and spatial patterns, but some mismatch at smaller scales can be expected as a result of these different scales.


The USGS upper Klamath River model has already been used to analyze water-quality effects of system changes in several preliminary model scenarios (Sullivan and others, 2011, 2012). The predictive improvements provided by the updated model will benefit future analyses of water-quality scenarios and help improve the understanding of water-quality processes in this reach of the Klamath River. 


First posted March 1, 2013