Pesticide residues are the single most common reason a frozen vegetable lot is rejected at an EU border or at a retailer’s intake, and almost never because someone sprayed something exotic. The usual causes are mundane: the limit that was checked was the wrong one, the sample that was taken did not represent the lot, or the certificate that was accepted covered forty active substances when the crop’s realistic risk list contained sixty. This guide walks the chain from the legal text to the piece of paper you actually receive, and shows where the decisions are really made.
What the regulation does and does not do
Regulation (EC) No 396/2005 is the framework instrument for maximum residue levels of pesticides in and on food and feed of plant and animal origin. Three features of it matter operationally.
First, it is substance-by-commodity. There is no single “vegetable MRL”. There is a value for a given active substance in a given commodity group, and the same substance can carry very different values in peas, in spinach and in beetroot.
Second, where no specific MRL has been set, a default value applies, generally 0.01 mg/kg, which in practice is at or near the limit of quantification for many methods. A substance that was never evaluated for your crop is therefore not unregulated; it is regulated at the strictest end.
Third, MRLs are not toxicological safety thresholds. They are trade and good-agricultural-practice ceilings derived from supervised trials. Exceeding one is a legal non-compliance, which is a commercial catastrophe, but it is not automatically a health incident. Keeping these two ideas apart is what allows a calm, evidence-based conversation with a supplier when a result comes back at 0.014 mg/kg against a 0.01 default.
The consolidated values are published and searchable in the European Commission’s EU Pesticides Database, and that database, not a supplier’s PDF summary, is the reference you cite in a specification.
The limit is set on the raw commodity, not on the frozen bag
This is the point most frequently misread. Annex I of Regulation (EC) No 396/2005 lists products in their raw agricultural state. Green peas, spinach leaves and broccoli heads have entries. “IQF green peas, 10 kg carton” does not.
For a processed product, compliance is assessed by relating the residue found in the processed food back to the raw commodity limit, taking account of what processing did to concentration. The regulation provides for processing factors to be applied where they are established. In practice that means one of two situations.
Where the process concentrates the commodity, residues can concentrate with it. Dried and pressed products are the obvious cases.
Where the process washes, blanches and hydrates, residues usually fall. Blanching in particular removes water-soluble residues into the blanch water, so a frozen vegetable typically carries less residue per kilogram than the field-fresh crop it came from.
The trap is that a favourable processing factor is not a defence you can invent at the time of a dispute. If your specification says nothing, an authority or a retailer will compare the measured value in the frozen product to the raw commodity MRL directly. That comparison is conservative and it will not usually harm you, because processing reduces rather than raises residues in this category. But you should still write down which basis of assessment applies, because the reverse case, a borderline result that is only compliant after a factor is applied, needs a documented and referenced factor rather than an argument.
| Step in a frozen vegetable line | Usual direction of residue change | Practical consequence |
|---|---|---|
| Field to reception, dry trim | No meaningful change | Field-level control is what decides the outcome |
| Wash and flume | Reduction, surface residues | Water hygiene and turnover matter |
| Blanching | Reduction, water-soluble residues | Largest single reduction step in the line |
| Cooling and freezing | No meaningful change | Freezing does not degrade residues |
| Sorting, packing, storage | No meaningful change | Cross-contamination risk only |
Sampling is where most disputes are actually decided
Two laboratories testing the same lot will agree far more often than two people sampling the same lot. Commission Directive 2002/63/EC establishes Community methods of sampling for the official control of pesticide residues in and on products of plant and animal origin, and it is the reference your specification should name.
What it gives you is a structure rather than a single number: a defined number of incremental samples drawn from across the lot, combined into a bulk sample of a defined minimum mass, from which the laboratory sample is prepared. For most frozen vegetables that means multiple grabs taken from different pallets and different cartons, not one bag taken from the pallet nearest the door.
Three practical rules follow.
Draw incremental samples across the whole lot, including from cartons that are inconvenient to reach, and record which pallets were sampled. A residue result is only as representative as the sampling record behind it.
Keep frozen samples frozen. A sample that thaws and refreezes before analysis invalidates the microbiological work you may want on the same material and can affect residue distribution in leachate.
Define lot size before you define sampling. A “lot” that means one day’s production is a different risk unit from a “lot” that means one contract. Your specification should say which.
Ordering the analysis you actually need
A multi-residue screen is defined by three things, and a certificate that omits any of them is close to useless.
Scope. How many active substances the method covers, and whether the list includes the substances realistically used on that crop in that origin. A 300-substance GC-MS/MS and LC-MS/MS panel that omits the two products actually applied to the crop tells you nothing. Ask for the substance list, and compare it against the crop protection programme the grower declares.
Limit of quantification. For substances at a 0.01 mg/kg default, an LOQ of 0.01 is the minimum useful sensitivity. An LOQ of 0.05 cannot demonstrate compliance with a 0.01 limit, and a “not detected” against that LOQ is not evidence of compliance.
Accreditation and uncertainty. Testing to ISO/IEC 17025 by an accredited laboratory, with the measurement uncertainty stated. Official enforcement decisions in the EU are normally taken by deducting the stated uncertainty from the measured value, which is why a result of 0.014 mg/kg against a 0.01 limit may or may not be an enforceable exceedance depending on the method. Know which convention your customer applies before you argue.
Add to that a stated basis: whether the reported figure refers to the sample as received or is recalculated to the raw commodity, and which residue definition is used where a substance is regulated as the sum of parent and metabolites.
Retail percentage-of-MRL clauses are trade practice, not law
Several European retailers require residues below a share of the legal MRL, commonly in the region of one third to one half, and some also cap the total number of detected substances or the sum of residues expressed as a percentage of their respective MRLs. This is a private commercial requirement. It has no basis in Regulation (EC) No 396/2005 and it varies by retailer and by year.
Treat it accordingly. It belongs in the customer’s specification, not in your legal compliance register, and it should be priced, because meeting it can require a different grower base and a tighter crop protection programme rather than merely a different laboratory.
Building a residue plan that survives an audit
A defensible plan is risk-tiered, written down before the season, and executed with records. The frequencies below are trade practice; the structure is what auditors look for under BRCGS and IFS Food raw material risk assessment clauses.
| Tier | Typical trigger | Usual approach |
|---|---|---|
| Baseline | Established grower, low-residue crop history | Screen per grower per season, plus finished-product screens per campaign |
| Elevated | New grower, new region, crop with a known residue history | Pre-harvest screen on standing crop plus per-delivery screening early in the season |
| Reactive | A detection above an internal action level, a customer complaint, an alert notification | Lot hold, extended scope screen, grower programme review, root cause and corrective action |
Two components are frequently missing. The first is a pre-harvest screen, which is the only test whose result arrives in time to change a decision; a finished-product screen tells you what you already froze. The second is an internal action level set below the legal limit, so that a trend is caught before a breach.
Also keep the GLOBALG.A.P. IFA certificates and the crop protection records of the growing base in the same file as the laboratory reports. An audit question about residues is usually answered by the agronomy record, not the certificate.
Organic does not mean zero
Organic production under Regulation (EU) 2018/848 restricts which plant protection products may be used and is verified by certification and inspection. It is a production-method standard. It is not a residue guarantee, and trace findings can arise from drift, from soil persistence of substances applied years earlier, or from cross-contamination in shared handling. Organic lots are still assessed against the same MRLs, and organic customers typically also apply their own internal action levels, often at analytical detection level, with an investigation procedure rather than automatic rejection.
Imported product and official controls
Consignments entering the EU are subject to official controls under Regulation (EU) 2017/625, and Commission Implementing Regulation (EU) 2019/1793 sets out a list of goods from certain third countries subject to a temporary increase in official controls or to emergency measures, with named commodity and origin combinations and defined check frequencies. That list changes. Before a season, check whether your commodity and origin appear on it, because a listed combination means border testing, delay and cost that must be built into the delivery plan rather than discovered at the border post.
Common mistakes
Accepting a certificate without its substance list, so nobody notices which actives were outside scope.
Comparing a result to an MRL found on a third-party website rather than to the current entry in the EU Pesticides Database.
Testing finished product only, which documents the problem instead of preventing it.
Treating a retailer’s percentage-of-MRL rule as a legal limit, and then over-specifying every customer to the strictest one.
Sampling one carton from the front of one pallet and calling it a lot sample.
Assuming organic status removes the need to test.
A worked example
A buyer contracts 240 tonnes of chopped frozen spinach across a season, from two growing regions, for a retail private-label pack.
Before the season, the supplier’s grower list, GLOBALG.A.P. certificates and declared crop protection programmes are reviewed. Two substances on the programme have a 0.01 mg/kg default for spinach, so the analytical scope is checked to confirm both are covered at an LOQ of 0.01.
Ten to fourteen days before the first harvest window, a pre-harvest screen is run on standing crop from each region. One region returns a detection at 40 per cent of the legal MRL for one substance, below the legal limit but above the retailer’s one-third clause. The harvest window for that block is deferred and re-screened rather than frozen and argued about later.
At production, one composite sample per campaign day is drawn using incremental samples across pallets, held frozen, and screened. Results are recorded against the internal action level of 50 per cent of MRL, and the two results that approach it trigger a grower conversation rather than a rejection.
At intake, the buyer’s own verification screen is run on one lot in five, sampled to the Directive 2002/63/EC structure, and reconciled against the supplier’s certificate for the same lot. Where the two differ by more than the combined measurement uncertainty, the retained sample is retested before anyone writes a claim.
The cost of that programme is a known number per tonne. The cost of one rejected container, including return freight, destruction and a delisting conversation, is not.
FAQ
Does freezing reduce pesticide residues?
Freezing itself does not. The reduction in a frozen vegetable line comes from washing and above all from blanching, where water-soluble residues transfer into the blanch water. Freezing preserves whatever the product carried when it entered the freezer.
Which MRL applies to a frozen vegetable, the raw or a processed value?
The limits in Annex I of Regulation (EC) No 396/2005 are set on the raw agricultural commodity. Compliance for a processed food is assessed by relating the result back to that commodity, applying an established processing factor where one exists. State the basis of assessment in the specification so that both parties compute the same answer.
How many substances should a screen cover?
Enough to cover the crop’s realistic use pattern in that origin, which is a question about agronomy rather than about laboratory marketing. Broad panels of several hundred substances are normal, but the useful test is whether the declared crop protection programme is fully inside the scope at an adequate LOQ.
Is a supplier certificate sufficient, or do we need our own testing?
Both, in different roles. The supplier certificate is the control record; your own verification screen on a sampled share of lots is what makes the certificate credible. Retailers and third-party audit schemes generally expect a documented verification element rather than reliance on supplier declarations alone.