Temporal Variation

The effect of time on water quality data

For 15 years, volunteers have dutifully collected and processed lake samples spring through fall, and then waited patiently for LMVP staff to report back with water quality data. Among the various graphs and tables, the annual data reports always list mean values for the monitored parameters. These values are used to summarize lake water quality.

Have you ever wondered just how well these numbers represent lake conditions? The answer to this question depends somewhat on you, the volunteer. The mean values that we report are nothing more than estimates, and the accuracy of the estimates hinges on a number of factors, starting with proper sample collection. The LMVP sampling protocol is designed to account for natural variation in nutrient, chlorophyll and suspended solids concentrations over time. In this article we consider the different scales of temporal variation and review what you, the volunteer, can do to ensure data quality. In the next issue of the Waterline we’ll look at spatial variation in water quality.

DAY-TO-DAY VARIATION

Nutrient and algal chlorophyll concentrations are constantly changing within a lake. Conventional wisdom suggests that short-term shifts (day-to-day) would be fairly small, but actually these changes can be considerable (Figure 1). In 2004 the MU limnology lab collected samples from Little Dixie Lake on 108 consecutive days during summer (this lake was also monitored by LMVP). Most day-to-day fluctuations in phosphorus, nitrogen and algal chlorophyll were small, but occasionally changes were substantial. The largest 24-hour fluctuation in phosphorus was ±15 μg/L, a one day shift that equals 28% of the average phosphorus concentration (54 μg/L) during the project. The maximum daily nitrogen shift was ±200 μg/L, 22% of the overall average of 900 μg/L. Algal chlorophyll was even more variable, averaging 42 μg/L and having a few daily fluctuations larger than 30 μg/L (71% of the average). Again, these were the exceptions and not the rule. Most of these large changes occurred as a result of heavy rain fall which transported nutrients into the lake as nonpoint source pollution; quick sedimentation of the nutrients right after the rain fall (nutrients bound to soil particles settle out fairly quickly); and in the case of chlorophyll, an algal bloom that formed and dissipated quickly.

Because daily sampling is not feasible for the volunteer program, we rely on random sample collection to account for variability in water quality. That is, we do not want to collect our samples only after heavy rains when the lake looks its worst or only on days when the lake is clear. By collecting samples on a pre-set schedule of every three weeks, we hope to avoid the potential problem of selectively targeting the worst or best of conditions. While having a pre-set sampling schedule to ensure randomness seems counter-intuitive, please remember the events that impact water quality (such as rainfall) are generally random. The pre-set schedule also assures that samples are not all collected during a short period of time (e.g. eight samples collected in one month). By scheduling sample collections evenly over the spring-fall period, we ensure data quality by monitoring the lake over a variety of conditions.

Figure 1. Day-to-day variations in phosphorus concentration in Little Dixie Lake during 108 days of monitoring in 2004.

Figure 2. Comparison of volunteer and MU phosphorus data from Little Dixie Lake during 2004.

SEASONAL VARIATION

Lake water quality is dynamic, so we have to collect samples throughout the season to accurately estimate nutrient and algal chlorophyll concentrations. The exact number of samples depends on various factors, including: stability of water quality in a given lake (because of hydrology, some lakes are more variable than others); length of the sample period (e.g. spring-fall vs. summer only); and the desired precision of our estimates. In general, the more samples collected, the stronger our estimate of water quality.

Even sampling just once every three weeks, LMVP is able to detect much of the variation in water quality that occurs during the sample season (Figure 2). Revisiting Little Dixie Lake, which has been part of the LMVP for nine years, we find that the maximum measured nutrient values are generally twice the minimum values. On average, phosphorus ranges from 40 to 80 μg/L and nitrogen ranges 620 to 1200 μg/L. During some years the ranges are larger than this 2-fold range. Chlorophyll is more variable, with the maximum value averaging about five times the minimum (average range 12.4 to 46.8 μg/L). To account for the variable nature of water quality, we need to collect a sufficient number of samples to allow for a strong estimate of average conditions.

By collecting the eight scheduled samples each season, volunteers ensure sufficient data points to allow an accurate estimate of water quality. We could go into the statistical explanation of why more is good, but that would make for a boring article. Trust me, more data result in a more accurate estimate of conditions. While there is no such thing as too much data, we find the gains in accuracy associated with collecting more than eight samples per year are relatively small.

YEAR-TO-YEAR VARIATION

Monitoring a lake for one sample season informs us only about water quality for that period. Even if a large number of samples were collected and we were confident about the generated mean values, one season of data tells us little about the overall water quality of a lake. Much like the need to collect multiple samples during an individual season to estimate annual conditions, we need multiple years of data to allow a long-term estimate of water quality. In Little Dixie Lake we find that annual geometric means for phosphorus range between 44 and 69 μg/L, nitrogen ranges from 603 to 1159 μg/L, and chlorophyll ranges between 10.4 and 57.4 μg/L (Table 1). Variation among annual geometric means from the nine years of sampling is comparable to the variation observed within most individual seasons. Differences among the seasonal geometric means relate mostly to climate (e.g. wet versus dry conditions).

How influential is climate on lake water quality? During summer 1989, after a long period of drought, the average phosphorus concentration in Mark Twain Lake was 18 μg/L. Following an exceptionally wet spring, phosphorus concentrations in this lake during summer 1990 averaged 163 μg/L. This example, while going from one extreme to another, shows the obvious influence that climate has on lake water quality.

Research on Missouri’s lakes indicates that at least four years of data are needed to begin generalizing lake water quality, with more data being better (especially if the lake in question tends to be more variable than the average Missouri lake). Not only are multiple years of data necessary to describe general conditions, but even more data are needed to identify long-term shifts in water quality taking place within a given lake.

To achieve the LMVP’s goals of describing current water quality of Missouri lakes and monitor for changes over time, volunteers need to be diligent in collecting enough samples over the spring-fall period to allow an accurate estimate of lake conditions, and lakes must be monitored for multiple years to allow long term evaluation.

Year	Phosphorus (ug/L)	Nitrogen (ug/L)	Chlorophyll (ug/L)
1999	44	603	10.4
2000	49	700	16.8
2001	50	798	16.9
2002	49	846	17.8
2003	51	690	16.2
2004	60	1159	28.2
2005	53	844	25.4
2006	69	945	57.4
2007	66	1120	37.7
Range:	44-69	603-1159	10.4-57.4

Table 1. Annual geometric mean values for trophic parameters monitored in Little Dixie Lake.

---