Home / Chroniques / Will agriculture find its salvation in mathematics?
Harvesting a wheat field, dust clouds
π Digital π Planet

Will agriculture find its salvation in mathematics?

Jeremie Wainstain
Jérémie Wainstain
PhD in Physics and Founder of Thegreendata
Key takeaways
  • In agriculture, serious issues lay ahead: food security, agro-ecology, decarbonisation, soil restoration.
  • Nitrates and mechanisation may have previously allowed a spectacular increase in agricultural productivity, but they rely on the use of hydrocarbons.
  • Maintaining high productivity while greening agriculture is possible by reasoning at the scale of the food system and by improving coordination between producers, consumers, distributors, and financial investors.
  • The issue is therefore to model better: more data, better shared and better used, with more complex models, integrating agronomy, climate, and markets.
  • The mathematisation of agriculture will be possible through a European food platform, allowing data to be shared between all the

To think about tomorrow’s agriculture, you insist on the need to think in terms of the food system. Isn’t that forgetting the heart of this system, production?

No. It’s to come back to it and to open up spaces for reflec­tion. The “food sys­tem” is at the heart of dif­fer­ent process­es: dis­tri­b­u­tion, con­sump­tion, and pro­duc­tion. Pro­duc­tion used to be a direct func­tion of three fac­tors: soil, ani­mals, and humans. All of these have been made more pro­duc­tive with mech­a­ni­sa­tion and nitrates (i.e. hydro­car­bons). In a way, we are eat­ing oil! The ques­tion today is: can we main­tain this pro­duc­tiv­i­ty with­out oil? And it is very dif­fi­cult to answer if we look at pro­duc­tion alone.

If we think in terms of the farm itself, one pos­si­ble answer is low tech; a return to prac­tices that exist­ed before mech­a­ni­sa­tion. But the ques­tion of pro­duc­tiv­i­ty soon comes up, and with it the ques­tion of cost.

Anoth­er response is to increase tech­no­log­i­cal inten­si­ty. But this still pos­es many prob­lems: it is not easy to extract beet­root with a robot. And ener­gy remains the key issue, even if we can imag­ine a shift to electricity.

On the oth­er hand, if we broad­en our think­ing from the farm to its wider con­text, and in par­tic­u­lar the food sys­tem, oth­er pos­si­bil­i­ties appear. We can opti­mise organ­i­sa­tion, mak­ing sure that agri­cul­ture is bet­ter inte­grat­ed into food and finance chains. There is a lot of poten­tial here because we are talk­ing about a poor­ly organ­ised, frag­ment­ed and poor­ly mod­elled world, where many deci­sions are tak­en in a non-coop­er­a­tive way.

How can we improve cooperation: more state, more market? 

To opti­mise the way things are organ­ised, pub­lic poli­cies are need­ed. The issue of food secu­ri­ty has been an invis­i­ble prob­lem in the pub­lic are­na for sev­er­al decades, but just because the prob­lem has been solved does not mean that it will not arise again. Ques­tions of food sov­er­eign­ty will come up again soon, as cli­mate change will put a strain on pro­duc­tion systems.

How­ev­er, it is cer­tain­ly not a ques­tion of cen­tral­is­ing every­thing as we saw with Gos­plan, the dis­as­trous results of which are well known in agri­cul­ture. Rather, the chal­lenge is to bring about bet­ter coor­di­na­tion between play­ers whose inter­ests are not cur­rent­ly aligned.

Hence, one of the hori­zons is a plat­form for Euro­pean food, with a cer­tain amount of data shared between all the play­ers with­in the next thir­ty years. The plat­form will sup­ply the major dis­trib­u­tors and will also allow financiers to car­ry out risk analy­sis. Plat­formi­sa­tion allows the math­e­ma­ti­sa­tion of agri­cul­ture. It is the key to bet­ter organ­i­sa­tion. The chal­lenge is to make the dif­fer­ent chains (pro­duc­tion, dis­tri­b­u­tion, financ­ing) more col­lab­o­ra­tive and to have new deci­sion-mak­ing tools. To do this, it is impor­tant to mod­el these chains from end to end, from the farm to dis­tri­b­u­tion and invest­ment funds, and to equip the deci­sion-mak­ing process­es at all geo­graph­i­cal lev­els. We need to put maths at the ser­vice of agriculture.

Is this a return to the spirit of the agricultural cooperatives that marked the modernisation of European agriculture after 1945?

Yes, in the sense that the coop­er­a­tive mod­el in Europe com­pen­sat­ed for the fact that farms were too small by mak­ing it pos­si­ble, for exam­ple, to pool equip­ment: mod­erni­sa­tion and coop­er­a­tion went hand in hand and a new lev­el was reached. These coop­er­a­tives, some of which have become very pow­er­ful, were organ­ised as buy­ing and sell­ing groups, with some advice.

But the vir­tu­ous side of this mod­el was under­mined by two phe­nom­e­na: the first is that it belongs only to the farm­ers. The sec­ond is that Euro­pean agri­cul­ture has been organ­ised in silos, and that it has been organ­ised with­in the frame­work of an agri­cul­tur­al pro­duc­tion pol­i­cy. Pub­lic poli­cies have thus cre­at­ed a clear sep­a­ra­tion between pro­duc­tion and food, which can still be seen today in the Green Deal.

It is time to rec­on­cile them, and math­e­mat­i­cal mod­els are capa­ble of doing so. Plat­formi­sa­tion and mod­el­ling offer a way of man­ag­ing col­lec­tive deci­sions and intro­duc­ing more ratio­nal­i­ty into them.

The agri-food sec­tor opti­mis­es its logis­tics and indus­tri­al process­es and fore­casts its food sales using mod­els. But noth­ing is coordinated.

Yet many models are already used on farms today. 

Yes, but they too are marked by an incred­i­ble frag­men­ta­tion. Agri­cul­tur­al exper­tise is stored in thou­sands of small tools, spread­sheets, mini-sim­u­la­tors, small cal­cu­la­tors, devel­oped ad hoc by farm­ers, tech­ni­cal insti­tutes, asso­ci­a­tions, agri-food man­u­fac­tur­ers, coop­er­a­tives or lab­o­ra­to­ries, with­out any con­sol­i­da­tion. These tools are gen­er­al­ly “out­side the infor­ma­tion sys­tem”, i.e. they are not fed into any recur­rent data flow. They are also not very user-friend­ly, due to a lack of invest­ment, and are most­ly unused. Last but not least, what they lack is a sys­temic approach.

On the tech­ni­cal side, plant and ani­mal genet­ics are mod­elled, as well as fer­tilis­er and plant pro­tec­tion prod­uct pre­scrip­tions. On the finan­cial and agri­cul­tur­al account­ing side, risk scores, per­for­mance indi­ca­tors and busi­ness plans are mod­elled because of reg­u­la­to­ry mod­els. But agri­cul­ture has no sys­temic mod­el at the lev­el of indi­vid­ual farms. Deci­sion sup­port tools gen­er­al­ly mod­el only one facet of the liv­ing world: the con­trol of a par­tic­u­lar dis­ease, the imple­men­ta­tion of a par­tic­u­lar prac­tice, in short, a very small part of the over­all system.

The agri-food sec­tor opti­mis­es its logis­tics and indus­tri­al process­es and fore­casts its food sales using mod­els. But noth­ing is coordinated.

As for the mod­els used by pub­lic poli­cies, they are obso­lete and serve to dis­trib­ute sub­si­dies. They are mod­els of eco­nom­ic flows and bal­ances that look to the past and ignore the agro­nom­ic dimen­sion. We are thus deprived of the tools to steer the Green Deal, which explains why ide­ol­o­gy takes prece­dence over real­i­ty: the fig­ure of 30% less pes­ti­cides is thus a polit­i­cal fig­ure, which is not sup­port­ed by data. In terms of data and mod­el­ling, every­thing remains to be done.

Does the energy and environmental transition offer an opportunity to move to new models?

It leaves us no choice. The prob­lems of poor coor­di­na­tion, frag­men­ta­tion and the sep­a­ra­tion of pro­duc­tion and food large­ly explain the con­di­tion of Euro­pean agri­cul­ture. It is a sec­tor that is poor­ly financed by the pri­vate sec­tor and sup­port­ed by pub­lic funds, which is won­der­ing about its future and find­ing it hard to invest.

Yet the real prob­lems are ahead of us: food secu­ri­ty, agro-ecol­o­gy, decar­bon­i­sa­tion, soil restora­tion. Every­thing needs to be done, in a con­text marked by cli­mate change, ten­sions over raw mate­ri­als and prob­a­ble tur­bu­lence on the world mar­kets for agri­cul­tur­al and food products.

Agri­cul­ture, which had been sim­pli­fied at the cost of car­bon and sub­si­dies, has sud­den­ly become what it was: a com­plex activ­i­ty, because liv­ing things are com­plex. And food is a very com­pli­cat­ed sub­ject. The sec­tor now has to deal with con­tra­dic­to­ry injunc­tions that place it on the thresh­old of a major disruption.

From the new geopo­lit­i­cal sit­u­a­tion to the rise in the price of ener­gy and there­fore of inputs, all the ele­ments are present for a cri­sis, with domi­no effects. Even a minor ele­ment such as envi­ron­men­tal labelling con­tributes to desta­bil­is­ing the system.

This major dis­rup­tion opens up a field for cre­at­ing new mod­els, and the tech­nol­o­gy is there: with­out wait­ing for plat­formi­sa­tion, we are now capa­ble of cre­at­ing and run­ning mod­els that are sophis­ti­cat­ed and rich enough in data to allow math­e­ma­ti­sa­tion of pro­duc­tion, but also of con­sumer behav­iour and financ­ing – this last issue being cru­cial to accel­er­ate the transformation.

If we need to mod­el, it is because we are col­lec­tive­ly blind, and we look to the past when the chal­lenges are ahead. The actors are aware of what awaits them. The ques­tion for them is now to organ­ise them­selves to avoid tak­ing the wave head-on.

Interview by Richard Robert 

Our world explained with science. Every week, in your inbox.

Get the newsletter