Project aims to unlock nitrogen savings through soil biology

Why can two fields managed in exactly the same way differ so dramatically in nutrient efficiency, crop resilience or fertiliser response? 

That’s what NutriGrow, a three-year Innovate-UK funded project launched in January 2025 is aiming to find out.

See also: Potato trials show little benefit from NUE products

The missing link, according to the project’s lead partner Elaniti, could be soil biology. On most farms, nutrient planning relies mainly on soil chemistry – N, P, K and pH – and basic physical assessments.

But the soil’s living organisms regulate nitrogen, phosphorus and sulphur cycling, support root functioning and interact with soil pathogens, affecting nutrient availability. 

NutriGrow is aiming to turn this hidden biological activity into usable information that helps farmers make smarter nutrient decisions.

Rather than looking at chemistry or even biology in isolation, the project uses a systems-based approach to examine how biology, chemistry and physical attributes interact in the soil to understand how soil functions and how it can be influenced, explains Elaniti chief executive officer and co-founder Scott Jarrett.

First year

In its first year, the project, which also involves the University of Lincoln, data integrator AgAnalyst, and the Nature Friendly Farming Network, is focused on baselining soil and farm management data from 10 participating farms growing wheat. 

“We are collecting approximately 250 soil samples from the farms in each year of the project,” Scott explains.

Each sample is analysed by the University of Lincoln for biological, chemical and physical properties, and then combined with various other data sets, including farm management data and in some cases grain nutrient analysis, before being evaluated using Elaniti’s machine learning and AI models.

 “These models provide information on soil functional performance,” Scott says. “One of those functions is nutrient supply efficiency, and particularly the efficiency with which nitrogen is used.”

Using foundational genomic modelling, which Scott describes as being like “ChatGPT for DNA data”, Elaniti can look at raw biological data and infer its metabolic function. 

“It means we’re less interested in the name of a particular microbial species compared with understanding what it does as accurately as possible,” he explains.

“That helps us figure out how efficiently the nitrogen in the soil is cycled and made available to the crop in different situations.”

It’s not as easy as just saying if this group of microbes is present, then nutrient supply will be improved, he stresses.

“There’s underlying interrelations between various components – you might need a limit on one [microbe or soil property] in the soil to enable another to act, which is what our models are aiming to uncover.”

The farm data is being used to refine and test the model’s performance by hiding some of the tested parameters to see whether the model can predict them accurately.

Heat maps

Currently, the model produces “heat maps” that compare nutrient supply efficiency performance in fields across approximately 4ha zones. 

Participating farmers are also given a detailed biological breakdown of each soil sample, including where it sits on Elaniti’s performance scale and a description of biological function.

But that’s just the first step. The project’s central goal is the development of a nutrient use efficiency module for Elaniti’s problem-solving AI model, nicknamed “Robin”, that will help farmers decide on the correct management interventions to improve nutrient supply efficiency.

With Robin trained on thousands of academic research papers that associate nutrient supply efficiency with management practices, the aim is to provide customised interventions based on a field’s specific biogeochemical composition, and other characteristics such as soil type and crop variety.

Suggested practices could include physical interventions, such as changes to cultivations, product applications including specific biological products or seed treatments, or incorporation of practices like intercropping or growing specific varieties. 

Within the current project, Scott isn’t expecting farmers to validate whether those management practices work – that will happen after the project finishes.

“The current phase is dedicated to rigorous model development and baseline testing, which will pave the way for targeted, on-farm field validation in subsequent phases,” he stresses.

“We are building the AI models and infrastructure so that from year three, we have a licensable product that potentially can provide independent, farmer-friendly decision support on how to improve nitrogen use.”

Academic studies have suggested that in some situations, reductions of at least 20% in nitrogen fertiliser are possible without compromising yields by adopting the right set of practices, Scott says. 

“We’ve set that as a target, but it’s impossible to say currently whether that’s achievable or even conservative.

“Our job for now is to focus on building a mechanistic understanding of how the soil is working and how different practices impact it.”

NUE-Profits project cuts fertiliser use by 15%

Another Innovate UK funded project has demonstrated cuts in nitrogen fertiliser of around 15% on average compared with farm standard practice in wheat crops, while improving nitrogen use efficiency by at least 10%.

The £2.8m NUE-Profits project is developing a “Framework for Improving Nitrogen Efficiency” (Fine) by integrating new and existing technologies, such as data from in-field sensors, soil sampling and weather stations in one platform developed by AgAnalyst.

Running since 2022-23, the Databaler platform provides farmers with specific recommendations for the first nitrogen dressing and subsequent applications, based on the end market, drilling date and nitrogen available in the soil.

© GNP

The recommendations have been tested against the farm standard on around 50 farms each year since the project began, says AgAnalyst’s Clive Blacker, with Fine recommending average cuts of about 30kg/ha of nitrogen.

“What we’ve seen is that the some of the farmers involved in the project are tending to drop their rates because they are getting more confidence in what we’ve done, albeit they don’t tend to cut quite as aggressively as we try to.”

AgAnalyst chief executive officer Jim Williams says this means the difference between farm practice and our trials is getting more marginal.

“But we’ve still got more work to do to raise the confidence to the level where every farmer will use it,” he adds.

Confidence is higher in feed wheat growers, where the algorithm behind the recommendations is proving effective, Clive says.

“We’ve not been as successful at reducing nitrogen doses in milling wheat to the same level. “It is easy to reduce greenhouse gas emissions while improving nitrogen use efficiency in milling wheat. The challenge is to do it profitably,” he admits.

Overall, as the algorithm in Databaler continues to be refined, farmers following the advice are making more money from every kilo of nitrogen applied.

“In 2023 we made an extra £0.40/ha versus the farm practice, but a lot of the sites were negative,” Clive adds.

“In 2024 we improved our margin over nitrogen to £6.70/ha and last year it was £14.60/ha benefit for using the Fine recommendation.

“At current fertiliser costs and wheat prices, our modelling shows that benefit has increased to £30/ha given the shifts in pricing.”

One key modification to the algorithm that’s helped drive improvements in its advice is moving away from just using growth stage towards thermal time to trigger applications.

“By combining soil temperature, rainfall and weather forecasts, the system identifies the optimum window for applying the next application,” Jim explains.

The project also uses satellite imagery and data company Assimila’s modelling to forecast yield potential in season, helping to adjust nitrogen recommendations in April or May to ensure the farmer isn’t over-investing in a crop with limited potential or under-investing in one that could perform better. 

Typically, the project is highlighting that farmers’ target yields are often significantly higher than what the model or even historical data predict, Jim says. “It’s a triumph of hope over experience.” 

But over time, Jim suggests the project could help farmers tailor their inputs to what the land can actually produce by providing more realistic maximum and minimum yield ranges.