Testing for Fipronil Alone Misses One in Four Detections in US Streams

The question a single-compound test cannot answer

Fipronil was used in 1,122,487 structural pest control applications in California in 2022, the sixth most of any active ingredient, according to our analysis of California structural pesticide use. Our report on insecticides in US streams found fipronil itself in 17.0 percent of stream samples since 2015, once hidden non-detects are read correctly.

Once applied, fipronil changes. Its breakdown products, including fipronil sulfone, fipronil sulfide and fipronil desulfinyl, are separate chemicals that laboratories measure separately. The US Environmental Protection Agency publishes aquatic life benchmarks for three of them alongside the parent. A monitoring result that reports only “fipronil” therefore answers a narrower question than it appears to.

We matched every stream sample in the national record that was analysed for fipronil and for at least one breakdown product, and asked how often the breakdown products turn up where the parent does not.

How often each form turns up

Source: Water Quality Portal, stream sites, results from 1 January 2015, retrieved 2 October 2026, with hidden non-detects corrected as described in our monitoring audit. ng/L is nanograms per litre. Author calculation.

What it means. Three breakdown products are each found in roughly one stream sample in eight, not far below the parent’s one in six. They are not trace curiosities. They are routine findings in the same samples, at concentrations of the same order as the parent.

What limits it. Each compound was analysed in a different number of samples, so the rates in this table are not directly comparable with each other. The matched comparison below fixes that by using only samples analysed for both.

Samples a parent-only test would have cleared

Restricting to the 20,676 samples analysed for fipronil and at least one breakdown product, at 2,221 stream sites, gives a like-for-like comparison.

Source: Water Quality Portal, stream sites, 2015 to 2025. Author calculation.

What it means. Of the 5,102 samples with any fipronil-related compound, 1,320, or 25.9 percent, would have been recorded as clean by a test for fipronil alone. Fipronil sulfone was the breakdown product most often found on its own, in 759 samples, followed by fipronil desulfinyl in 734.

What limits it. “Not detected” means below the laboratory’s reporting limit, not zero. Some of the 1,320 samples probably contained fipronil below that limit. The finding is about what a parent-only result would report, which is the number most readers see.

226 sites that only the breakdown products reveal

Counting stream sites instead of samples, fipronil was detected at least once at 674 of the 2,221 matched sites. Fipronil or a breakdown product was detected at 900. That leaves 226 sites where a breakdown product was found and the parent never was, in any matched sample.

What it means. At those 226 sites, 25.1 percent of all sites with any fipronil-related detection, a monitoring summary based on the parent compound would classify the stream as fipronil-free.

What limits it. Sites sampled only once or twice are more likely to show an all-or-nothing pattern by chance. We did not weight sites by sampling effort.

Is it one programme’s quirk?

Our audit of national monitoring data found that how programmes code non-detects can distort national rates. We therefore also repeated the matched comparison inside each of the four largest contributing programmes.

Source: Water Quality Portal, stream sites, 2015 to 2025. Increase is the relative rise in detected samples when breakdown products are counted. Author calculation.

What it means. In each of the four programmes that detected anything, counting breakdown products raises the number of samples with a fipronil-related detection, by between 12 and 96 percent. The effect is largest where the parent is rarest. In Washington, the breakdown products nearly double the detection count.

What limits it. The National Park Service samples, 791 of them, detected neither the parent nor any breakdown product. We did not examine their reporting limits, which may be higher than those of the USGS laboratories.

The gap year by year

What it means. The breakdown products add detections in every year from 2015 to 2023, by between 3.9 and 8.3 percentage points; 7.9 points in 2015 and 7.1 in 2023.

What limits it. Both lines drift up after 2019. We do not read that as rising contamination, because laboratory methods and the mix of contributing programmes changed over the period. The years 2024 and 2025 are excluded from the chart because they contain few matched samples, 332 and 188. They remain in the published CSV.

Imidacloprid is the counter-example

The same matching for imidacloprid, the most frequently detected insecticide in our stream analysis, gives the opposite answer.

Source: Water Quality Portal, stream sites, 2017 to 2025. 3,414 of the 4,219 matched samples, 80.9 percent, come from the Minnesota Department of Agriculture. Author calculation.

What it means. For imidacloprid, a test for the parent compound captured every fipronil-style “hidden” detection except one. Measuring its breakdown products added almost nothing to the count of contaminated samples. The one-in-four gap is specific to fipronil, not a general property of insecticide monitoring.

What limits it. The imidacloprid comparison rests on far fewer samples, 4,219 at 376 sites, and is dominated by one state programme. It shows that the fipronil result does not generalise automatically. It does not show that imidacloprid breakdown products never occur without the parent elsewhere.

What the federal benchmarks cover

EPA’s Aquatic Life Benchmarks, last updated 2 February 2026, list fipronil at 0.11 micrograms per litre for acute effects on freshwater invertebrates and 0.011 micrograms per litre, or 11 ng/L, for chronic effects. The same table lists separate benchmarks for fipronil desulfinyl, fipronil sulfide and fipronil sulfone, each updated in 2021.

Of the 3,994 parent fipronil results in our data carrying a measured concentration, 427, or 10.7 percent, exceeded the 11 ng/L chronic freshwater invertebrate benchmark, and 12 exceeded the 110 ng/L acute benchmark.

What it means. Federal risk assessment already treats the breakdown products as substances with their own toxicity. The 1,320 samples a parent-only result would clear are not samples EPA considers irrelevant.

What limits it. We compared only the parent compound against its own benchmark. We did not assess the breakdown products against theirs, and adding concentrations of different compounds against a single benchmark would not be valid. A benchmark exceedance indicates potential for effects on sensitive invertebrates. It is not a measured effect.

What this does not show

• It does not show where the fipronil came from. Structural pest control is one use among several, and these data cannot separate sources.
• It does not show harm to any stream. Detection rates and benchmark comparisons describe exposure potential.
• It does not explain the chemistry of why fipronil’s breakdown products persist without the parent. That is a question for laboratory and field studies, not for a monitoring database.

Checks on the result

1. Like-for-like matching

Definition. Comparing parent and breakdown products only within samples analysed for both.

Method. Samples were matched by the Water Quality Portal activity identifier, which is shared by all results from one water sample.

Result. 20,676 matched fipronil samples and 4,219 matched imidacloprid samples.

Limitations. A sample whose results were split across activity identifiers would not be matched and is excluded.

2. Programme robustness

Definition. Testing whether the gap survives inside individual monitoring programmes.

Method. Repeated the matched comparison for each of the five largest contributors.

Result. The gap appears in all four programmes that detected anything, from +12 to +96 percent.

Limitations. Smaller programmes were not tested individually.

3. A control compound

Definition. Running the identical method on a second insecticide family to see whether the gap is an artifact of the method.

Method. Same matching and detection rules applied to imidacloprid and four of its breakdown products.

Result. One hidden detection in 4,219 samples, against 1,320 in 20,676 for fipronil. The method does not manufacture a gap where none exists.

Limitations. The control dataset is smaller and less geographically spread.

4. Download completeness

Definition. Confirming every result the database reported was received.

Method. Each download was checked against the Total-Result-Count header returned by the portal, and downloads were fetched one at a time.

Result. All nine breakdown-product downloads matched their reported counts exactly, for example 24,742 rows for fipronil sulfone.

Limitations. The portal changes as programmes submit data. A later download will differ.

Download the data

The tables in this report are published as CSV files under a Creative Commons Attribution 4.0 licence. You may republish, remix and build on them, including commercially, with attribution to The Pest Control Brief and a link to this page.

• Detection statistics for fipronil, imidacloprid and nine breakdown products in US streams (CSV, 11 rows)
• Matched co-occurrence of each breakdown product with its parent (CSV, 9 rows)
• Fipronil parent and family detection rates by year, 2015 to 2025 (CSV, 11 rows)

How to cite this report

The Pest Control Brief, “Testing for Fipronil Alone Misses One in Four Detections in US Streams”, 2 October 2026. https://thepestcontrolbrief.com/fipronil-breakdown-products-in-streams/

References

1. Water Quality Portal, Result search for Fipronil, stream sites, from 1 January 2015. waterqualitydata.us Result search, Fipronil. Retrieved 25 September 2026.
2. Water Quality Portal, Result search for Fipronil Sulfone, Fipronil Sulfide, Fipronil Desulfinyl, Desulfinylfipronil amide and Dechlorofipronil, stream sites, from 1 January 2015. waterqualitydata.us Result search, Fipronil Sulfone (other compounds by the same query with the name changed). Retrieved 2 October 2026.
3. Water Quality Portal, Result search for Imidacloprid and Imidacloprid-urea, Imidacloprid-olefin, 5-Hydroxyimidacloprid and Imidacloprid desnitro, stream sites, from 1 January 2015. waterqualitydata.us Result search, Imidacloprid-urea. Retrieved 2 October 2026.
4. US Environmental Protection Agency. Aquatic Life Benchmarks and Ecological Risk Assessments for Registered Pesticides. Last updated 2 February 2026. epa.gov aquatic life benchmarks. Accessed 2 October 2026.
5. The Pest Control Brief. Insecticides in US streams. thepestcontrolbrief.com/insecticides-in-us-streams. Source of the parent-compound detection rates.
6. Florida Department of Environmental Protection. Rule 62-160.700, Table 1: Data Qualifier Codes. floridadep.gov (PDF). Accessed 3 October 2026. Definition of the U qualifier.

How the samples were matched

Approach. We downloaded every stream-site result since 1 January 2015 for fipronil, imidacloprid and nine breakdown products from the Water Quality Portal, grouped results by sample using the activity identifier, and compared detection of the parent with detection of each breakdown product in the same sample.

Detection rule. Results flagged “Not Detected”, “Not Detected at Reporting Limit”, “Not Detected at Detection Limit” or “Below Reporting Limit” are non-detections, as are numeric results whose qualifier code contains the WQX code U or whose comment text carries Florida’s U flag. Results flagged “Present Below Quantification Limit” or “Trace”, and other positive numeric results, are detections. Programmes that never report a non-detect for a compound are excluded for that compound.

Exclusions. Laboratory quality-control recoveries, passive-sampler units, sediment and tissue units, and results with no usable date or unit.

A sample counts as a family detection if the parent or any breakdown product it was analysed for was detected.

Limitations. Reporting limits differ by laboratory and year. Not every sample was analysed for every breakdown product, so the five fipronil products have different matched totals. The data describe where programmes chose to sample, which is not a random draw of US streams.

Correction, 3 October 2026. The first version of this report read numeric results that were flagged as not detected, mainly by Florida programmes, as detections. Correcting them lowered every detection rate and raised the share of fipronil-related detections that a parent-only test misses, from one in seven (943 of 6,632) to one in four (1,320 of 5,102), and the number of sites revealed only by breakdown products from 98 to 226. The imidacloprid control result is unchanged. Full explanation in our monitoring audit.

Last reviewed: 3 October 2026.

For journalists

Most quotable statistics

1. One in four fipronil-related detections in US streams, 1,320 of 5,102, involved only a breakdown product, with no detectable fipronil.
2. At 226 stream sites, fipronil breakdown products were detected and the parent never was.
3. Fipronil sulfone, desulfinyl and sulfide were each detected in roughly one stream sample in eight.
4. For imidacloprid, the same test found one hidden detection in 4,219 samples.
5. 10.7 percent of measured fipronil concentrations exceeded EPA’s chronic freshwater invertebrate benchmark of 11 ng/L.

Caveats to carry with them

“Not detected” means below a reporting limit. Detection does not show harm. The data cannot separate structural pest control from other fipronil uses.

Fields in the published CSV files

Family; compound; samples analysed; samples detected; detection rate; sites; sites with a detection; quantified values; median; largest contributing programme and its share; years. For the matched file: breakdown product; matched samples; parent detected; product detected; product without parent; parent without product.

Press summary

Monitoring that tests only for fipronil misses about one in four fipronil-related detections in US streams, according to an analysis of the national water-quality record by The Pest Control Brief. Across 20,676 stream samples analysed for both fipronil and its breakdown products between 2015 and 2025, 1,320 contained a breakdown product but no detectable fipronil. At 226 stream sites, the breakdown products were found and the parent never was. The gap appears in every large monitoring programme that detected fipronil, and it does not appear for imidacloprid, where the same method found one hidden detection in 4,219 samples. EPA publishes separate aquatic life benchmarks for three fipronil breakdown products.

Headlines editors could use

1. Testing for fipronil alone misses one in four detections
2. 226 US stream sites look fipronil-free until you test for what it becomes
3. The termite insecticide that outlives its own test
4. Fipronil’s breakdown products turn up in one stream sample in eight
5. Why imidacloprid tests catch everything and fipronil tests do not

Questions people ask

What are fipronil breakdown products?

Chemicals that fipronil turns into after application, including fipronil sulfone, fipronil sulfide and fipronil desulfinyl. Laboratories measure each one separately, and EPA publishes separate aquatic life benchmarks for these three.

How often are fipronil breakdown products found in US streams?

Fipronil sulfone was detected in 13.6 percent of stream samples analysed for it since 2015, fipronil desulfinyl in 13.3 percent and fipronil sulfide in 11.1 percent.

Does a test for fipronil alone miss contamination?

In this data, yes, about one time in four. Of 5,102 samples with any fipronil-related compound, 1,320 had a breakdown product and no detectable fipronil.

Is the same true for imidacloprid?

No. Across 4,219 matched samples, imidacloprid breakdown products appeared without the parent only once.

Are the breakdown products toxic?

EPA treats three of them as having their own aquatic toxicity and lists benchmarks for each. This report does not compare their concentrations with those benchmarks.

Does this mean fipronil pollution is rising?

These data cannot show a trend. Detection rates drift upward after 2019, but laboratory methods and the mix of monitoring programmes changed over the same period.

Where does fipronil in streams come from?

These data cannot say. Structural pest control is one use among several, and monitoring results do not record the source.

What should a water-monitoring programme do with this?

Report fipronil results together with its breakdown products where the method measures them. A parent-only summary understates how often fipronil-related compounds are present.

Why did the National Park Service find nothing?

Its 791 matched samples detected neither fipronil nor any breakdown product. We did not examine its reporting limits, which is the most likely explanation to check first.

Can I reproduce this?

Yes. Download the stream results for each compound from the Water Quality Portal, group by activity identifier, and apply the detection rule in the methods section. Our CSV files contain the summary tables.