Trends from volunteer monitoring

Imagine tracking a child’s growth if your measuring tape had no inch markings. 1 ft, 2 ft, 2ft, 3 ft, 3 ft, 3ft

Not very helpful is it? The same thing could happen if you were trying to measure a gradual decline in nitrate with a Hach test strip, which measures in very course increments: 10 mg/L, 10 mg/L, 10mg/L, 5 mg/L, 5mg/L

For this reason, I’ve always assumed that trend monitoring is not a realistic use for volunteer monitoring data. However, that wrongly assumed that any year-to-year trends would be small and gradual. Instead, using weekly lab data, I’ve found several big swings in nitrate over the past 20 years. It turns out volunteers with the Squaw Creek Watershed Coalition tracked some of the same trends. These six sites were sampled monthly from 2002 to 2008. Trends at three sites were even statistically significant, at the 90% confidence level.

Nitrate trends as measured by volunteers with the IOWATER program.
Weather whiplash

Weather whiplash

cover crops near Nevada
Nice looking rye cover crop near Nevada, IA
November 2020

A big thank you to farmers who planted cover crops after this challenging year. Cover crops will hold soil and nutrients in place through the winter and early spring. That could be especially important this year.

After a drought, nitrogen that might normally have been taken up by a high-yielding crop or flushed away by rainfall remains in the soil. That leftover nitrogen could be available for next year’s crop, but only if:

A) we have a dry spring, or

B) farmers have made use of practices like cover crops or nitrification inhibitors that prevent nitrogen losses during the fallow season.

A wet year in 2013 following a dry year in 2012 (dark blue) caused nitrate concentrations in many central iowa streams to jump as much as 10 mg/L higher than usual (red). Figure from Van Metre et al. 2008

A drought in 2012 following by a wet spring in 2013 led to nitrate concentrations in excess of 20 mg/L in many rivers in Central Iowa. If we have a wet spring in 2021, we could see this happen again. As one scientific paper put it, “weather whiplash drives deterioration of water quality.”

“Weather whiplash” can also help explain the long-term trends I’ve been seeing in the South Skunk River and its tributaries: a decline in nitrate concentrations from 2005-2012, a big jump in 2013, and another decline over the past 7 years. I’ll walk you through my analysis.

Explaining nitrate concentrations in the South Skunk River

Technical details, feel free to skip: This data was collected by the City of Ames just upstream of wastewater treatment plant. The City has monitored the South Skunk River above and below its wastewater treatment plant almost every week since 2003! Flow is measured continuously at a few miles upstream USGS gage near S. 16th St. I’ve summarized nitrate concentrations and streamflow by season (Jan-Mar, Apr-Jun, Jul-Sep, Oct-Dec). At each step of the way, I apply a linear regression equation and graph the model residuals. Taken together, these three factors explain 59% of the variation. The effects of “weather whiplash” may extend beyond one year, since nitrate from some parts of the field may travel more slowly to streams via groundwater.

In a given quarter, nitrate concentrations in the South Skunk can be up to 10 mg/L higher than the long-term average, or as much as 10 mg/L lower. The following graphs show how much variation is left to explain after correcting for current weather, last year’s weather, and season.

Nitrate in the South Skunk is again declining after a sudden increase in 2013.
Low-flow conditions in 2011-2013 explain unusually low nitrate concentrations.
Wet springs following dry years explain unusually high nitrate concentrations.
Seasonal patterns explain some of the remaining variation.
  1. The lowest nitrate concentrations can be explained by streamflow: when the weather is dry and tiles aren’t flowing, nitrate levels in rivers taper off to the background levels seen in groundwater.
  2. The highest nitrate concentrations can be explained by weather over the previous 12 months: a wet period following a dry period will flush out nitrate that’s accumulated in the soil.
  3. After that, there’s still a seasonal pattern independent of rainfall: nitrogen is most susceptible to loss in spring when soils are bare and microbial activity picks up (April-June) and least susceptible when the maturing crop is using up the available nitrogen (July-Sept).
  4. Can some of the remaining pattern be explained by greater adoption of conservation practices in the watershed in the past 5 years? We hope so, but let’s see what happens next spring!

Too close to call

Too close to call

Disclaimer: This article is not about politics. PRI is a non-partisan organization and does not want to get drawn into a discussion about the election. My intent here is to use an example that’s fresh in our minds to illustrate a challenge for progress tracking in water quality.

Polls are not always accurate. If you didn’t know that before November of 2020, you do now.

There are plenty of parallels with water, so if you’re looking to water quality monitoring to tell you whether or not conservation efforts in your watershed or your state are succeeding, read on.

Bottom line: Short-term water quality trends are usually too close to call.

Quinnipiac University can’t talk to every eligible voter in Florida, so they surveyed 1,657 people the week before the election.

Similarly, it’s not practical to test water quality in a stream 365 days a year, so we often make do with just 12 days a year. There are sensors that can test nitrate or turbidity continuously during the ice-free months, but they’re not cheap.

The voting patterns of those 1,657 people won’t be perfectly representative of Florida. How close do they expect to be? Based on the sample size, Quinnipiac calculates a margin of error: in this case 2.4 percentage points. Talking to more people would reduce the margin of error, but not enough to be worth the cost. In this case, Biden’s lead appears to be outside the margin of error.

Support for candidatePoint estimate95% confidence interval
Biden47%44.6% to 49.4%
Trump42%39.6% to 44.4%

Similarly, the 12 days we test water quality won’t be perfectly representative of the year. How close can we expect to be? We can calculate a margin of error (here, the 95% confidence interval) around our water quality average. Did phosphorus decline in 2019? Too close to call!

Phosphorus in South Skunk RiverPoint estimate95% confidence interval
2019 average0.30 mg/L0.20 mg/L to 0.40 mg/L
2018 average0.36 mg/L0.26 mg/L to 0.46 mg/L

Oops! This poll missed the mark, and by more than the margin of error. Trump actually won Florida with 51.2 percent of the vote. Well, some errors are unavoidable. If sampling error were completely random, we’d expect about 5% of polls to miss by more than the margin of error. That’s what “95% confidence” means. However, according to a new study of 1,400 polls from presidential primaries and general elections, 40% of polls conducted a week before the election missed the mark by that much.

That’s because sampling error isn’t all random. People who ultimately voted one way may have been less likely to appear on the list of phone numbers, less willing to respond, less likely to say what they truly intended, or were more likely to have changed their mind in the final week. And then there’s undecided voters, who don’t always split evenly between candidates. Any of these things can skew the results. Pollsters try to correct for some of these things by weighting various demographic groups, but it doesn’t always work. For “margin of error” to mean what we think it means, according to Kotak and Moore, it would need to be at least two times wider.

Similarly, if monthly samples are not collected on a fixed day of the month, you might underestimate phosphorus or sediment load by planning your week to avoid getting wet, and over-estimate it by going out of your way to capture a sample during a storm. This challenge is well understood in environmental science and there are sampling strategies and equipment to get around it.

What’s not widely appreciated is just how big purely random sampling errors can be. I’ve read the literature and run my own numbers. Even with a 3 years of sampling, you’re lucky to get below a 20 percent margin of error. The problem is more severe for phosphorus and sediment (which move with runoff and vary a lot within a month) than nitrate (which moves with groundwater and drainage water, and is less variable).

Knowing this can help us set realistic expectations. As much as we’d like to know whether conservation efforts on the land are translating into improvements in water quality in the river, we’re not going to be able to tell the difference between a modest improvement and no improvement unless we sample often enough or long enough to bring down the margin of error. Figuring out whether this is worth doing, how to sustain it, and what other things we can learn from water monitoring has been the task of PRI and our partners around Story County working on a ten-year monitoring plan.

A final note. Have some sympathy for pollsters and scientists who are doing their best to base their findings on data, acknowledge the uncertainty in their conclusions, and strive to be less wrong. There are plenty of people on both sides of the aisle who confidently make predictions based on anecdotes, are wrong more often than not, and never admit it. 😉

Fall 2020 Water Quality Snapshot finds sensitive critters

Fall 2020 Water Quality Snapshot finds sensitive critters

Thirteen volunteers braved the cold on October 24 to test water quality in Squaw Creek, the South Skunk River, and their tributaries. For some, this was their 14th Fall Water Quality Snapshot. For others it was their first time doing stream monitoring. What we found defies easy categorization.

volunteers collecting invertebrates in Squaw Creek with a kick-net
Volunteers Diane Birt, Paul Readhead, Liz Calhoun, and Rick Dietz take turns brushing off rocks to capture bugs in our net. The water was freezing!

By some indicators, water quality in Squaw Creek was good. Since I wasn’t sure how many creeks would be flowing when I planned this event, I added bug collection to the agenda to keep us busy. Excuse me. Benthic macroinvertebrate sampling. We were pleased to find tiny stoneflies and mayflies. They’re good fish food, ask any fly fisherman! Excuse me. An important part of the aquatic food web. These insects also act as a sort of canary in the coal mine. They need water with a lot of dissolved oxygen so will be rare or missing in streams with too much pollution, murky water, or not much in the way of habitat.

Fun fact: while the adult mayfly is notorious for living only 24 hours, the juvenile form (naiad) lives in the stream for several years. If you’re curious what adult mayflies and stoneflies look like, I found some photos from our neighbors in Missouri.

Vials containing tiny mayflies and stoneflies

By some indicators, water quality in Squaw Creek and it’s tributaries was bad. As in, there’s poop in the water. Excuse me, fecal indicator bacteria. This month, E. coli bacteria in Squaw Creek continued to exceed the primary contact recreation standard, and College Creek jumped above secondary contact standard. I wondered if this spike might be due to accumulated… debris… being washed out of the storm sewers and off the landscape by the 1.25 inch rain we received Thursday and Friday, but the lab samples were actually collected on Wednesday Oct 21, so I’m not sure. Anyway, covid-19 is not the only reason I bring hand sanitizer to these events!

By some indicators, water quality was unusually dry this fall. Nitrate was too low to detect at 13 of 16 sites we tested. Under wetter conditions, as we had this spring or last fall, nitrate in these same streams was higher and differences due to landuse or conservation practices in the watershed become more apparent.

SiteWatershedFall 2020Spring 2020Fall 2019
Squaw Creek @ Duff AveRural and urban0105
Bluestem Creek @150th StRural2105
Glacial Creek @ U AveRural (with a constructed wetland)022
College Creek @ UniversityUrban022
Nitrate-N concentrations, in mg/L

Water quality is rarely all good news or all bad news. Citizen science can us a more complete picture.

Thanks to all our intrepid volunteers!

E. coli in context

E. coli in context

Edited 2020-10-16 to clarify a few sections, add references as links, and correct statements about trends. It has gotten worse

People swim, paddle, and wade in Iowa lakes and rivers. I’ve seen a lot of that this summer. We’d like those waters to be clean enough that it doesn’t pose a health risk. But see enough E. coli data, and it’s tempting to give up. There’s raccoons in the storm sewers, and geese on the lakeshore, and livestock on the farms, so therefore there will be poop in the rivers and lakes. How on earth would you ever figure out where it’s coming from, let alone improve the situation? I’ve said this once or twice, and I’ve heard it from experts I’ve consulted.

So hat tip to Chris Jones for putting our numbers into context. I love context! Let me add some more! (Check out his blog post to for some excellent background on what this bacteria is, and why we measure it).

It’s bad, and don’t blame the geese

cow in a stream
Cow in an Iowa stream. Photo credit: Rick Dietz

I was inspired by Chris’s graph, which compares E. coli in the Raccoon and Des Moines Rivers in Des Moines to the Mississippi River in Minneapolis. That river has plenty of geese and raccoons, as I can attest to from having lived in the Twin Cities. So clearly there’s some additional sources of fecal contamination in these Iowa waters that we could do something about, if we had the will.

graph by Chris Jones. Raccoon River in Des Moines exceeds the standard 50%, while Mississippi in Minneapolis exceeds 10% of time
Figure by Chris Jones. Percent of samples collected from 2010-2020 during recreational season that exceeded 235 CFU/100mL.

It’s very bad some places, fine in others

I love this metric: percent of days in the recreation season that E. coli exceeded the standard. Nobody knows what a geometric mean is, or how many samples you need before you can apply it. This is a simpler way of acknowledging that sometimes there will be poop in waters where people wade or swim (say because there are racoons in the storm sewers and it just rained) but we’d like that to be less often. Using my new powers of R scripting, I computed this metric for every water body in an Iowa DNR database that had at least 50 E. coli samples during April-Oct of 2010-2020, and when I couldn’t fit 160 bars on a graph, I pulled out an arbitrary selection to give you a sense of the range.

As I feared, the South Skunk River below Ames is even worse than the Racoon River, exceeding the standard 75% of the time, as are several rivers in central, western, and northeastern Iowa. As Chris mentioned, E. coli is lower downstream of reservoirs, which allow sediment to settle and light to disinfect. Some of the cleanest waters aren’t on here because DNR rarely finds anything, and thus has scaled back testing. This includes destination spots like Okoboji Lake, and my favorite local beach, the former gravel pit at Peterson Park, which is mostly groundwater fed.

It hasn’t gotten worse, has it?

I have fond childhood memories of wading and canoeing and fishing in Iowa rivers and lakes, including the South Skunk River. How bad was it then? I don’t know. I do have data from 2000-2004, when I was an ISU student hauling trash out of the river and planting willow stakes with Jim Colbert’s “Skunk River Navy.” At that time, the South Skunk upstream of Ames met (today’s) primary contact recreation standard, but started to exceed it in 2006. The South Skunk River downstream of Ames and below the confluence with Squaw Creek has exceeded the standard in every year it’s been measured, but was twice as high the last two years as it was in 2000-2004.

Here I’m using the geometric mean for April-October and the 126 CFU/100 mL standard that applies to it, since I already had that calculated. Apologies if you have trouble keeping track of the various standards.

The DNR was monitoring the South Skunk upstream of Ames at Riverside Drive for a special project, which ended in 2014. However, we’ve been monitoring that site and several others in Story County this year. So far, it’s higher there than it was in 2000-2004, barely meeting the standard, but I have one more month left in the “recreational season.” Every other site we monitored (that didn’t dry up this summer) was 2 to 9 times higher than the standard.

Confession: I am unwilling to give up playing in rivers or deny that experience to my kids because of a fecal bacteria problem that seems unlikely to be addressed. As a whitewater paddler, I’ve always been willing to take some risks to enjoy nature, and am not sure how the risk of waterborne illnesses compares to the more obvious risks of drowning and injury. I now make sure to pack hand sanitizer on river outings, and try to keep my head out of the water, but hey, sometimes your boat tips over and that’s part of the fun!

Can you keep your head above water?

Primary contact recreation uses “involve full body immersion… such as swimming and water skiing.” The same standard applies to smaller creeks “where children’s activities are common, like wading or playing in the water.” Small kids tend to splash and put their hands in their mouths. Water designated for secondary contact recreational uses, “such as fishing and shoreline activities” have a less stringent single sample E. coli criteria of 2,880 CFU/100mL. The South Skunk River exceeds the secondary contact standard less than 10% of the time, so if fishing is all you’re doing, you can worry less.

Graph comparing streams to secondary contact standard

Wait, can I do that? Apply secondary contact criteria to a water body designated for primary use? The DNR doesn’t, even when it would make sense. Long Dick Creek is on the Impaired Waters List for E. coli levels above 235 but below 2,880. In reality, it’s too small to float a canoe and has no public access. If this decision affected anyone’s permit, someone would have asked for an assessment to rebut the presumptive designated use, but it doesn’t, so it will remain on the Impaired Waters List for the foreseeable future, last in line for a cleanup plan (TMDL). This underscores a frustrating truth about the Impaired Waters List and the broader Clean Water Act. It was set up to regulate point source polluters like wastewater treatment plants and industry. It’s not a very good framework for educating the public about risks, or for cleaning up waters affected by other (non-point) sources of pollution.

How risky is manure?

Which brings us to risk. Not only is it hard for most people to evaluate risks when they have the numbers, we don’t have numbers that are relevant for Iowa.

What does EPA mean by “protective of human health” when considering E. coli and contact recreation in a lake or stream, where immersion and ingestion of the water is likely? In this case, a threshold of 235 CFU/100ml (or a geometric mean of 126) would be expected to produce illness in no more than 36 people in 1000 (i.e. 3.6%). So meeting the standard does not equate to zero risk.

Chris Jones, “Dream On

I’ll have to dig up the references, but as I recall, these risk calculations were based on epidemiological studies of GI infections at swimming beaches that were sometimes affected by human waste. Part of the debate in the scientific literature is whether those same rates risks apply when the source of E. coli is animal waste.

It’s a question of relatedness. Assuming there’s poop in the water and assuming you swallowed some water, and assuming the animal in question was sick, you’re less likely to contract an illness from poultry and geese manure than swine manure, which is less risky than cow manure, which is less risky than untreated human waste. On the other hand, here’s some research linking pathogens in central Iowa streams to swine manure.

Also, a single threshold is not that helpful for evaluating risks in the many waters that don’t meet the standard. If E. coli a stream is 10 times the standard, is my risk ten times as high? Probably not. Twice as high? The original EPA studies are no help here because not only is the risk model hard to interpret (I’ve tried), the numbers we have here in Iowa are well outside the range used to calibrate the model!

What are we gonna do about it?

As part of Story County’s 10-year Monitoring Plan, we’ve been monitoring E. coli in several local streams. Our partners at ISU, Story County, and City of Ames will be exploring the use of optical brighteners to track down wastewater discharge. We know there are some bad septic systems and aging sewer pipes contributing to high E. coli levels in the South Skunk River and Squaw Creek, and since those pose a higher health risk and are perhaps easier to address than livestock manure, that’s where we’ll start. But I’ve love it if we could talk about manure too, without it being seen as controversial.

I appreciate that Chris Jones brought up E. coli. Amid all the talk about nutrients, we often lose sight of another pollutant whose impact is easier to see locally.