The Puget Sound area is currently experiencing a population boom, which is only expected to continue into the near future. The development associated with this increasing population has started to encroach into less developed regions to the east and north. As this urbanization increases in the Seattle area, the water quality of the lakes in the region is at greater risk of eutrophication from increased phosphorus exports off the land to the water.

Increased development means more impermeable surfaces that water cannot infiltrate. This water would normally be absorbed into the ground allowing the phosphorus it contains to be bound within the soil, but instead, the water is transported directly into streams and lakes supplying phosphorus into the system faster than would occur in a non-urbanized region. Why is phosphorus an important parameter? The lakes in this region are phosphorus limited and so increased phosphorus loading to these systems will increase the likelihood of nuisance algal blooms. These algal blooms are unsightly and odorous, and can even be toxic.

Increased urbanization also means more wastewater to treat and more clean water to provide. Water reuse is being considered to increase the freshwater supply to the growing population of this region. Wastewater that undergoes tertiary treatment would be discharged into Lake Washington to recharge the water lost due to water consumption. But care must be taken to ensure that the water entering the lakes will not upset the phosphorus balance since this could lead to nuisance algal blooms. In order to determine this, we must know how much phosphorus is already entering the lakes on an annual and seasonal basis.

One of the main objectives of this course was to develop a spatially-explicit nutrient loading model for the Lake Washington and Lake Sammamish watersheds, which could be incorporated into PRISM (Puget Sound Regional Synthesis Model). PRISMs focus has been to model the Puget Sound region on a number of different physical and social scales. The PRISM model will be used to assess future impacts to the region from such things as population growth, land use, and water utilization.

A key feature of the spatially explicit nutrient loading model was the prediction of the phosphorous load which washes into the surface waters of the two watersheds, based on land-use and land-type. Our analysis of stream phosphorous export focused on calculating actual phosphorous loads into Lake Washington and Lake Sammamish from historical data sets, and comparing these results to the spatially explicit model.

King County’s Department of Natural Resources Sammamish-Washington Assessment Modeling Program (SWAMP) provided much of the necessary information for our analysis. SWAMP has several main objectives, which include the monitoring of lake and stream quality, identification of short- and long-term trends, model development for water quality predictions, and support of other programs. Most of the lake and stream data was obtained from SWAMP. This data included information on monthly stream flows, total phosphorus, and soluble reactive phosphorous for 32 stream locations in the watersheds (see Figure 1 for stream locations and identification). We have estimated annual phosphorus concentrations from each of the streams into Lake Washington and Lake Sammamish and ultimately modeled the P-loading from a daily flow regime.

Our primary objective was to predict yearly phosphorous concentrations and subsequent yearly loading values within the watersheds. Initially, the resolution to this problem appeared uncomplicated: perform a simple linear regression between phosphorous and flow. However, it quickly became apparent that such a solution would not be feasible, since a direct correlation between TP concentrations and flow did not exist. The majority of the data showed high TP during low flow and low TP in high flow periods, which is contrary to our understanding of flow vs. concentration relationships. Also the TP vs. flow data was widely scattered on log-log plots, which demonstrates little correlation. Due to these two conditions, an alternative analysis had to be resolved.

The next course of action appears to be development of a time series analysis, which incorporates a time lag into the prediction of TP. This time series should be able to determine TP based on previous hydrologic conditions. In other words, if there has been a period of dry weather (low flow), followed by a storm event, the time series would predict a high TP value. Likewise, after a period of repeated storm events, the next storm event would not correlate to a high TP value.

Due to time constraints, the development of this time series analysis was beyond the scope of this course. In order to provide daily and annual phosphorous loading values, which are needed by the other groups, we used the geometric mean of TP/SRP concentrations over time. The use of a geometric mean as opposed to an average will diminish the impact of periodic high TP concentrations. These mean concentrations will then be applied to a Daily Flow Index for each stream, resulting in a daily load value in kg/day (see Stream Index section below). A drawback to this procedure is that the values calculated could be overestimated or underestimated. This will add another aspect of inaccuracy that could be reduced in future projects.

The SWAMP project data is sparse in places. Some sites have consistent monthly readings with storm events interspersed. The inclusion of these storm events throws off the possible interpolation of the data into daily flows, because these measurements are not random. To avoid this problem, United States Geological Survey (USGS) gauging stations were investigated for potential daily flow data. Only four USGS stations corresponded directly with SWAMP data collection sites: Issaquah Creek, Mercer Creek, Cedar River and Sammamish River.

Since daily flows are not available for most of the streams we are investigating, we decided to develop a stream flow Index for the Lake Washington watershed (Brett). Although all of the streams in the watershed have slightly different hydrologic regimes, we are assuming that they all behave similarly to form this Index. This assumption is not necessarily true, but allows us to approximate daily flow values for any stream in the watershed. Of the four SWAMP sites, only the gauges at Mercer Creek and Issaquah Creek can be used towards the Index since a dam and Lake Sammamish affect the hydrology along the Cedar River and the Sammamish River respectively.

USGS historical hydrologic data was used to create the Index because it contains long-term, daily flow measurements. The hydrologic data for Rex Creek, a relatively undisturbed river feeding Chester Morse Lake, was added to that of Mercer Creek and Issaquah Creek to form the Index. Issaquah Creek represents a large stream that contributes approximately 70% of the flow to Lake Sammamish (Perkins, 1996). Mercer Creek is a small, urban stream flowing into Lake Washington while Rex Creek is in the undisturbed watershed of the Upper Cedar River Basin.

There is error associated with this Indexing process, but by choosing streams representative of those found in the watershed, we hope to minimize this error. An error analysis was not done on this data set. This Indexed-flow is multiplied by the geometric mean of the phosphorous concentrations to find the daily TP and SRP loading.

The daily hydrographs of the three indexing streams matched which shows that these streams are synchronous with each other. (Figure 2: Indexed flow 1992-1993) Although each stream showed major increases in flow at similar times, Mercer Creek exhibits a spikier profile, which is typical of an urban stream. The Index incorporates historical daily flow values from October 1979 through September 1996.

The index was made by first finding the geometric mean daily flow for each stream. The average of these geometric means is the log flow of our Index stream.

Index flow = average of the geometric means of the three streams.

The next step is to compare this Index with the King County data to create a scaling factor for each stream. Assuming that the Index represents the flow rate trends of the streams within the King County data, an average difference between the Index and the stream data, for any particular day, could be used to scale the Index to approximate the stream magnitudes.

The King County flow data must first be formatted to place them on the same daily time interval as the USGS flow data. This was done by using the computer program Transform™. The scaling factor is found by subtracting the LOG flow values for the Index from that of the transformed King County stream flow data. The average of this difference is the scaling factor, which describes the size of the stream compared to the Index.

Scaling factor = average [log (measured flow) – log (Index flow)].

Log ("Indexed" flow) = scaling factor + log (Index flow).

The Index stream has a log flow value of one: this correlates to a mean flow of 10 cubic feet/second (cfs). The scaling factor is added back to the daily Index flow to develop a daily flow regime for the desired stream. The inverse log of these values was taken to get the "indexed" flow in terms of cfs.

Two concerns arose with the use of this technique; "What effect does the inclusion of the storm events have on the final loading value?" and "What is the effect of not accounting for seasonal flushing variations?". The final output of this project does not deal with these concerns for two reasons. First, due to the time constraints and limited data availability, including them would be unfeasible. Second, the effects of both "storm chasing" and "seasonal variations" were investigated and their results seem to justify our final approach, as is shown in the following.

To examine the effects of including "storm chasing" TP/SRP concentrations (those that fall outside of the regular sampling routine), the known storm values were eliminated from four stream data sets. King County provided a breakdown of which recorded TP/SRP values were storm related. These new sets of data were then used to obtain Average Yearly Loading of TP/SRP for these streams. See Tables 2 and 3 below. Table 2 includes the TP/SRP concentration variations and Table 3 includes the loading differences of TP/SRP. The four streams are all from the Lake Sammamish watershed since storm data for Lake Washington was unavailable.

Examination of the TP vs. flow data seemed to indicate some seasonal pattern to TP. Elevated TP concentrations were seen during early fall, when increased precipitation causes a "first-flush" effect from urban watersheds. During late fall and winter, TP values are low due to increased flow and stream dilution. This same routine also seemed to occur with SRP. To examine this trend, the TP/SRP values from September to November were extracted from the data sets of four streams. It is assumed that this time span would encompass the "flush" concentrations, and that the remaining months would encompass the "normal" concentrations. These now separate data sets were each used to compute average values of TP/SRP for the "flush" and "normal" phases. These average values are shown in Table 4 below.

Note that the Flush and Normal concentration values do vary from the original total Average Concentration, though the Flush values are sometimes not all that much higher. These values were then applied to the corresponding months within the Index and the same steps were used to obtain a yearly average. Of the four streams used, two are streams flowing into Lake Washington and two flow into Lake Sammamish. The final TP/SRP loading values are compared with the Total Averaged values in Table 5.

From this comparison, you can see that there is no discernable pattern to the difference in loading values. For Juanita Creek the TP/SRP loading increased substantially, but Issaquah TP loading dropped over 100 Kg., and Mercer TP only changed by 7 Kg. The concentration values are more consistent, however they also do not necessarily fit a pattern. This seems to show that this technique for dealing with flushing effects is not perfect, and a new approach needs to be found. The ideal next step would be to catch and record some actual flushing concentrations and loads to get a better idea of the system.

The results from our modeled output have been applied in several formats. 1) A breakdown of TP flowing through the watershed (given as an average monthly load) was compared to a GIS database. This comparison gave an indication of how the land use and land type has affected the watershed, as well as the change in TP loading due to urbanization. 2) A measurement of external SRP loads to Lake Sammamish and Lake Washington, from their respective tributaries, was used in phytoplankton growth models. These models were used to predict phytoplankton dynamics and assess any significant impacts to the lakes.

Time Series Analysis. From our analysis and the problems that arose, we can recommend to King County that continuing investigation of TP/SRP prediction models needs to occur for future understanding of phosphorus loading. A time series analysis would be helpful in modeling the TP/SRP concentrations most of the year vs. higher summer flushing concentrations due to the first major storm of the season. Also this model could help eliminate inconsistencies in our analysis due to storm chasing events.

Daily Data Sets. Identify a few sites to be sampled for both flow and phosphorous concentrations for at least a year. This would help refine the TP/SRP prediction models.

Index Formulation. Work on improving index formulation. Since the Index was formed with streams from the lower portion of the watershed, a northern site would be recommended for daily sampling.

Organization of data. Streamline coordination between King County and USGS sampling sites. Make all available information, usable and compatible.

Brett, M. T., Personal Communication

Perkins, W. W., et al. (1997) A zero degree of freedom total phosphorous model: 2. Application to Lake Sammamish. J. Lake and Reservoir Management. Vol. 13, no. 2. Pp. 131-141.