Can open data show farming’s share of the nitrogen?
The usual figure is that about 70 % of the nitrogen reaching Danish coastal waters comes from human activity on land, mostly farming. We tried to check it with nothing but open measurements, using a method locked before we looked at a single value. The direction is clear. The size is not.
More farmland, more nitrogen
Each dot is a stream: its farmland share against its nitrogen, with treatment plants and sewer outlets taken out.
Nitrogen is the flow-weighted mean of total nitrogen over each station’s valid years, after removing the 2024 discharges from treatment plants and rain-dependent outlets upstream. Farmland is the share of the catchment covered by registered field blocks. Hollow dots are stations upstream of another station in the analysis.
No stream with little farmland
The streams that passed every rule have between 41 % and 84 % farmland in their catchments. What a stream carries with no farmland at all, the natural background, is where the line meets 0 %. That is a long way from the data.
The split, and how sure it is
Shares of the nitrogen passing the outermost stations, with 95 % ranges from 10,000 resamples of whole river systems.
Cited split: about 70 % human-caused diffuse (mostly farming), 20 % background, 10 % point sources (Aarhus University, DCE, for 2010). Farmland here is the remainder: what background and registered point sources don’t explain.
Three reasons
- The background is extrapolated. Small changes in the line move the meeting point at 0 % a lot.
- Few catchments could be drawn. Station catchments are not open, so we drew them from the elevation model. Only 83 of 197 matched their stated area within 10 %.
- Big rivers dropped out. Flat land, drains and road embankments defeat a terrain model. The largest catchment that passed is 762 km².
The same question, asked eight ways
All fixed in advance. R6 could not be run: catchments were only drawn for stations with a stated area. Shares above 100 % mean the fitted background is below zero, which is not physical: one more sign the line can’t pin it down.
Direction yes, size no
Streams with more farmland carry clearly more nitrogen, and point sources are a small part at these stations. Whether farming’s share is 60 % or 80 % cannot be settled with open measurements alone. That takes catchments and models that aren’t public, which is what the official split rests on.
Every station
The streams in the analysis. Nothing here is a verdict on any stream, farm or municipality.
Locked before we looked
The method, every rule and the period were committed before the first nitrogen value was read. The reading “Inconclusive” was one of four outcomes written down in advance.
Method, data and code
Every rule, every change with its reason, the data with SHA-256 and the script that produces this page. CI recomputes it on every change.
The repository →Stream chemistry
Danmarks Miljøportal, VanDa “Vandkemi Vandløb”. Flow from the open VanDa hydrometry API.
The dataset →The figure we checked
Aarhus University, DCE: nutrient inputs to Danish waters, 2010, table 5a.
The note (PDF) →What this can’t show
These are shares at measuring stations, not at the coast: nitrogen is also removed further downstream. Farmland is a remainder, so houses outside the sewer and nitrogen from the air sit in it. Point sources are from 2024, three years after the period. A relation across streams is not an experiment.
Better catchments
We have asked Aarhus University for their station catchments. If they can be published, the analysis is run again with them, the change is logged with its reason, and this result stays next to the new one.