Opinions Wouter de Heij
Current nitrogen calculations do not help anyone
Measuring is knowing. You would think so. The doctor can measure your temperature very precisely, but cannot immediately conclude from this whether you are ill or not. The same goes for nature and Aerius, the ecologist's thermometer. This is stated by Wouter de Heij, trained as a chemical engineer and active as an entrepreneur in the field of food innovation.
Suppose doctors decide to call someone sick if their body temperature rises above 37,0°C. Although this limit value seems theoretically very clear and precise, a problem immediately arises. This quickly becomes apparent when we look at the measuring equipment. A thermometer with a 0,2°C scale can only measure degrees in 0,2°C increments. The doctor cannot record temperature fluctuations with more decimal places.
The appearance of precision
The example shows how precision functions in practice in the medical world. Doctors understand that individuals' body temperatures vary within a certain range. That is why they do not use a hard limit but one range. In his daily practice, the doctor relies more on diagnostic criteria than on apparently 'rock-hard' precision to make his patient better. Because the body temperature of adults varies between 35,5°C and 37,2°C, setting a limit at 37,0°C is not useful in medical practice, but also not scientifically because you are misinterpreting the condition of people. also interprets.
We go one step deeper. Suppose someone has a clinical thermometer with a display with five decimal places and the doctor reads it as 37,23542°C. If this person is 0,03542°C above the typical upper limit of 37,2°C, you might think he is sick. However, the accuracy of a measuring instrument depends not only on the number of decimal places it can display, but also on the actual precision of the measurement made. If the thermometer has a maximum accuracy of one decimal place, this means that only the decimal place is considered significant. In this case, the measured temperature of 37,23542°C is actually only accurate to 37,2°C because the decimals are not significant. Neither for the measurement itself (which does not produce a 'fact' but does produce a number), nor for the diagnosis.
The two examples show that it is important to be clear about what accuracy means. In science, medicine and even the legal world, we need to be aware of the limitations and accuracy (which in technical terms is called noise) of measuring instruments, regulations and standards to make informed decisions and draw the right conclusions. Accuracy is not just a matter of numbers, but also of understanding the context in which they are measured and used.
Aerius, the ecologist's thermometer
There is also a similar need for accuracy in the nitrogen dossier. We will look for concepts in the nitrogen file that fulfill similar functions as fever as an indication of illness. Key question: which thermometer do we use in the nitrogen file to determine the health of nature?
For example, the standard for illness in the case of fever is a body temperature of 37,2°C or higher. The higher you are above this body temperature, the greater the chance that you are actually ill. But doctors know that this is just a chance and that they have to consider other aspects as well. Consider various things such as physical exertion, the ambient temperature or a high stress level. Such factors can also influence your temperature.
When monitoring the border where nature is endangered by nitrogen emissions, the Netherlands uses the critical deposition value (KDW). This value is comparable to the 37,2°C limit value on the thermometer used by the doctor. The KDW indicates that there is a chance that nature will deteriorate. If the KDW is exceeded, the chance that the quality of nature will deteriorate increases. However, if we remain below the KDW, the risks for nitrogen deposition are non-existent or negligible.
The KDW has been determined by ecologists. They did this mainly on the basis of practical research in which manure (nitrogen) is applied to the soil. This happens in more or less precise steps. It can be done with 5, 10, 20 or even 50 kilos of nitrogen per hectare. In practice, however, research is not conducted per hectare, but rather on a much smaller surface area, such as one square meter. In fact, it mainly happens in flower pots because they allow more precise measurements. And the latter indicates that the research is done outside nature in a laboratory setup.
In case of fever, the doctor uses a clinical thermometer as a measuring instrument real life situation. A real measurement takes place. However, in ecology we do not have a 'meter' that can determine the total deposition as easily. Even academic researchers can hardly measure deposition accurately in practice, let alone to values after the decimal point. That is why the government, which is responsible for monitoring the state of nature in the Netherlands, has chosen to use the Aerius computer model. This is calibrated based on some actual measurements of nitrogen concentration in the air; Aerius itself is a model used to perform calculations. It does not measure, but calculates based on assumptions what the total nitrogen deposition in an area is per hectare per year. What the thermometer is in the hands of the doctor, Aerius is in the hands of ecologists.
Precision and correctness
Even though the doctor can measure your temperature very precisely, we saw that the doctor has more to diagnostic criteria than a hard and precise limit value to estimate the health status of his patient. What about the accuracy of the KDW and that of Aerius as an ecological thermometer in the form of a calculation model?
First something about the concept of 'correctness'. In statistics, accuracy is the degree of agreement between the (average) value of a series of observations and the true value. The accuracy of a value that you use as a starting point for a diagnosis is determined by the so-called systematic error. The more the values systematically deviate from the true value, the less accurate the observations are. The magnitude of systematic errors depends on the number of unknown factors - and their mutual influence on each other - that play a role as noise around the factor under investigation. If you do not identify the systematic error, you will end up with models that run away from reality.
Another concept from statistics: precision. That indicates something completely different. Precision is the extent to which measurements or calculations differ, or how consistent measurements and the algorithms applied to them always produce the same results. Precision is determined by random error. The smaller the random deviations, the greater the precision. The impact of the two concepts on your understanding of reality can be understood using the example of shooting a bull's eye.
Source: Wikipedia
In the left image you can see that the precision is high because the points are close together. At the same time you see that the accuracy is low because they are far from the target. This indicates a systematic error. In the right image you can see that the accuracy is better, but the precision is lower. In that case, there is less deviation in the instrument we use to record reality in a value.
We can expect that the KDW can be determined with an accuracy of +/- 35 to 70 moles per hectare per year. This corresponds to the actual measurement accuracy of the manure sampling. See the box for the substantiation of these numbers The critical deposition value Below.
Roland Bobbink was recently heard on The New World. He is considered an authoritative ecologist and contributed to the preparation of the KDWs for Dutch nature reserves. The video gives a good impression of his enthusiastic attitude and gives an impression of the energy that fuels his thinking. He has contributed to many studies and reviews of studies. In Bobbinks Review and revision of empirical critical loads of nitrogen for Europe (110/2022) per nature type (habitat) you can find what the established KDW is according to experts. Expert committees play a very important role in the determinations. LNV requested translations of experimental values for Dutch nature reserves in 2008, 2012 and 2023. Bobbink's authority gave it weight.
I take Appendix 2023 from the latest edition (1):
Suppose that the KDW of area H3110 is indeed 6 kilos per hectare per year. Then this number also indicates the accuracy. In any case, there is only one significant figure; rightly, no number after the decimal point has been given. By multiplying these 6 kilos by 71,4 (with three significant figures), you arrive at 428,4 moles per hectare per year with the help of the calculator. But that would be a scientifically incorrect representation that suggests too high an accuracy. Actually, you should write this number with one significant figure: 4 x 10^2 mol; the notation 429 mol is scientifically completely 'wrong'.
Another way to look at this is to realize that fertilization series are used in practice when determining the KDW. Often per 1 kilo or per 2 kilo resolution (so 3, 4, 5, 6, 7 or 4, 6, 8, 10 kilos per hectare), but sometimes also in steps of 10 kilos. It is therefore safe to assume that this mass determination has a precision of at least +/- 0,5 kilo (but probably even +/- 1 kilo or more). Calculated back to moles: +/- 0,5 x 71,4 = +/- 40 mol or +/- 1 kilo x 71,4 = +/- 70 mol.
The KDWs for very sensitive nature are then approximately 5 x 10^2 +/- 70 mol. They could never be more precise. However, I intuitively think that we have to take at least +/- 100 moles into account. In the political debate about nitrogen, it is about a tiny fraction of it, namely 0,0005 mole and now also about 1 mole or 10 mole. The relationship between those numbers is rather unclear and certainly does not argue in favor of the more precise variants in the debate.
Computers can calculate thousands of decimal places if necessary. Modelers who spend too much time behind a screen and do too little in the field, or who are less statistically skilled, can happily let their processors simmer away to an awful false precision. However, they are just numbers with no actual object in the outside world. Nevertheless, we can try to estimate the precision and accuracy of the numbers that the model calculates for Aerius.
To start with the latter: measurements of dry deposition are scarce. Dry deposition models have been compared in scientific studies (and show considerable differences), but each of these model calculations also has its own accuracy. In this earlier text I discussed quantitatively how Aerius works. However, there is an inaccuracy in each of the steps emission > dispersion > deposition. If we assume for the sake of convenience that the deposition is linearly dependent on the local concentration of nitrogen (total ammonia and NOx), then the question is: how accurately do we know what the average annual nitrogen concentration is, and how accurately do we know the amount of rainfall? ?" Accuracy refers to the number of significant figures that are scientifically correct. I must admit that I have not carried out an extensive analysis, but the rain is expressed in millimeters (and not with a decimal place). Therefore, it is safe to say that the maximum there are two significant figures. The concentration of ammonia is often reported with one or two decimal places. However, I suspect that in that case too only two figures are significant.
This means that wet deposition can never be measured better than with two significant figures. The determination of dry deposition is even more difficult. I find differences between models of a factor of 2 to 4. I think the relative error is at least 50%. Due to the addition of wet and dry deposition, the accuracy of the total deposition is very much determined by the least accurate deposition (I think the dry one). If the measurements are already so inaccurate, how can we think that the overall derived Aerius model can be?
If you follow through with this reasoning consistently, you can hardly help but come to the conclusion that Aerius has a scientific precision of approximately +/- 25 to 50 moles or more. I also think that the dry deposition is greatly overestimated in the model. This means that Aerius is not only not very precise, but that there is probably also a systematic error in the deposition model. This systematic error means that the accuracy is also probably incorrect. The accuracy of the deposition calculated by Aerius could turn out to be tens to perhaps a hundred moles higher than is actually the case. Aerius therefore misses the target quite a bit.
As for thermometer Aerius, the computer model with which the government monitors the health of nature, we can assume that the total deposition calculated by the model has a comparable degree of accuracy as the measurements on which the model is based. This means that despite its apparent ability to calculate with many decimal places, Aerius probably has an accuracy of +/- 25 to 50 moles per hectare per year. See the box Accuracy, in terms of precision and correctness below to substantiate this estimate.
The concepts of precision and correctness play a crucial role in various aspects of our lives, but often require several aspects to capture a complex process in a theoretical framework that is useful in practice. In the nitrogen file, the KDW is only determined with one significant figure. That number also gives the expected precision of the determination: +/- 40 mol or more, I think. For the accuracy of Aerius (i.e. the model as a measuring instrument), the precision is +/- 25 to 50 moles or more; there is probably also a systematic error, which means that it is not yet possible to say anything about the correctness of the model.
In addition to the Hordijk Committee, TNO also looked at the accuracy. On April 28, 2022, TNO sent a note to the Ministry of Agriculture, Nature and Food Quality entitled Demarcation in the modeling of deposition contribution of individual project contributions (Phase 2) Version 3. TNO concludes that the noise in the calculated deposition is between 1 and 10 mol/ hectare/year. TNO mainly relied on the meteorological uncertainties in OPS (TNO speaks of 'a consideration based on the physics chosen arithmetic lower limit'). TNO did not delve deeper into dry deposition and wet deposition measurements and the associated measurement inaccuracy. It is precisely this measurement inaccuracy that ultimately determines the accuracy of the (calibrated) model. It is physically pointless - and scientifically harmful - to include more significant figures in a model calculation than the original measurements on which the model was validated.
If TNO were to look at these practical measurements, it is likely that TNO would also have to conclude that the inaccuracy of Aerius is on the order of tens of moles, and that TNO's lower noise limit of 1 mole is the very conservative (optimistic) side. On page 26 of the relevant note, TNO also discusses the inaccuracy of the KDW provision. Abbot et.al (2003) is cited as a source. Abbot assumes an uncertainty for the KDW of 20%. Van Dobben et al (2012) is also cited and states that there is an uncertainty of several kilograms per hectare per year. I think this uncertainty is justified. However, TNO assumes that the KDW is measured with a precision of 0,1 kilo and therefore concludes that the precision is 7 mol. Van Dobben is probably right in his observation that +/- 1 kilo is really the best precision for the KDW (theoretical +/- 0,5). TNO's 7 mole for KDW is therefore a friendly scientific guess. It is therefore more realistic to assume at best +/- 35 moles or even +/- 70 moles per hectare per year for the precision of the KDW.
There is now a heated debate in parliament about figures of 0,005 mol per hectare as the basis for the permit requirement for construction projects for companies and infrastructure of the government itself and their replacement with a limit value of 1 mol. The responsible minister Van der Wal talks about the great care with which she wants to deal with law and reality. Why if the correctness of the model is so gross? It can be established that policymakers and parliament do not take into account the limitations and inaccuracies of the measuring instruments and models used.
For example, German companies are allowed to emit more nitrogen and tax nearby nature reserves more than Dutch companies. It appears that Germany may be taking a more scientifically correct approach to assessing nitrogen deposition. Our eastern neighbor allows 21 mol (300 grams) deposition as a limit value. It should immediately be noted that better science and making nature better are not necessarily the same. However, it does mean that cutting nitrogen very hard is unlikely to help due to the limited accuracy of the models the Dutch government works with. This indicates that unknown but determining factors other than nitrogen alone are at play. In plain English: if you keep a heart patient with COPD at a temperature strictly below 37,23542°C, but forget to give them enough oxygen, water, pills, a ventilator and a puff, the patient will still die.
The whole of the Netherlands suffers from bad science
It is important that in the Netherlands we approach the concept of accuracy in a scientifically correct manner, as previously concluded by the Hordijk Committee. Aerius should not be used for permits as it is not suitable as a legal instrument on scientific grounds. In addition, it is only suitable as a thermometer for monitoring the health of Dutch nature if a correct assessment of the systematic error is made.
If we do not delve into the precision and correctness (accuracy) and do not determine the systematic error of the algorithm-based nitrogen thermometer Aerius, then the whole of the Netherlands will suffer. It does not help nature, it makes life impossible for companies and farmers and can cost citizens an incredible amount of wasted money. Rutte IV has allocated almost €25 billion for very inaccurately determined nature goals. Investing is fine, but not in castles in the air.
This article is part of the content collaboration between Boerenbusiness en foodlog.
Wouter de Heij
© DCA Market Intelligence. This market information is subject to copyright. It is not permitted to reproduce, distribute, disseminate or make the content available to third parties for compensation, in any form, without the express written permission of DCA Market Intelligence.
