Real-time in situ sensing of soil nitrogen status to promote enhanced nitrogen use efficiency in agricultural systems

Lead Research Organisation: University of Nottingham
Department Name: Sch of Biosciences

Abstract

Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.

Planned Impact

UK agriculture uses over 0.85 million tonnes of nitrogen (N) fertiliser each year which is spread over 8.2 million hectares of tilled and grassland soil. A major proportion of this added fertiliser, however, is not taken up by the crop and is lost to the wider environment. This results in a major economic loss to farmers and can lead to pollution of water courses, groundwater and the atmosphere. As the use of synthetic fertilisers will continue to be pivotal in food production for the foreseeable future, new ways are needed to effectively target the efficient use of this resource. One of the major outputs from our research programme will be the creation of new decision support tools that are based on on-farm, real-time soil data which continually update during the growing season. This represents a major advancement in current fertiliser guidance systems (e.g. RB209, Planet, Farmscoper). The outputs of our research on society can be grouped as follows:

INDUSTRY: This research proposal is directly underpinned by key industry partners. These include (i) Yara UK who are one of the leading suppliers of N fertilisers, crop nutrient sensors (e.g. Yara-N-Sensor) and fertiliser guidance; (ii) Agricultural Industries Confederation (AIC) who are the agrisupply industry's leading trade association. AIC's Fertiliser Sector represents over 95% of the UK's agricultural fertiliser supply industry, worth about £2bn; (iii) British Grassland Society is a communication forum which through events and publications promotes the profitable and sustainable use of grass and forage; (iv) Agriculture and Horticulture Development Board (AHDB) is a Levy Board which represents the cattle, sheep, pigs, milk, potatoes, cereals, oilseeds and horticultural industries. AHDB are also responsible for reviewing current UK fertiliser recommendations associated with RB209. Our project directly aligns with the strategic priorities for all these industry organisations. All the main partners will be members of our management board, and will provide invaluable guidance throughout the project and will facilitate the dissemination of the project findings.

POLICY COMMUNITY: The results from this project will directly inform policymakers (e.g. Defra, DECC) by providing clear advice on future developments in precision agriculture including the environmental and economic costs and benefits and barriers to technology adoption. We will also provide guidance on the timelines and likely impact that adopting these technologies will have at the UK level and its potential impact on the UK N inventory. Policymakers are also central to our proposal (see WP5) ensuring dialogue throughout the programme. We will also build on our established links with Defra and Welsh Government to ensure effective dialogue.

WIDER COMMUNITY: A web page and Twitter feed from the Bangor website will provide ongoing information on the project and its results. Different aspects of the project will be used for teaching, generating student projects, and will be presented at open days at (1) the Bangor University Agricultural Extension Farm, which is one of Defra's Sustainable Intensification Platform flagship sites, and (2) by our industrial partners. We will also feature the project in School Science Week, using visualisation of nitrogen pollution to stimulate wider discussion about agriculture and the environment.

SCIENTIFIC COMMUNITY: Our research will inform scientists working in several areas of research (e.g. crop production, grazing management, water quality, greenhouse gas emissions and modelling). We will generate fundamental information on the use of in situ N sensors, plant-soil-microbial N cycling as well as providing new 3D mathematical modelling tools and information of the spatial heterogeneity of nutrients at a range of scales. These technologies will be promoted through the project-dedicated website, at national and international conferences and in journal publications.

Publications

10 25 50
 
Description In a prior report (before transfer of this award to University of Nottingham) it was explained how an optimized sampling design for spatial deployment of sensor networks had been developed for use in this project. The design was used in an experiment by colleagues at Bangor University, and data were returned for analysis. This allowed us to quantify the variability of nitrate concentration in soil water at different scales.

The results showed that very short-range (10-cm) variation dominated over the growing season, increasing in magnitude over the first half of the season, and then declining in the latter part of the season when variation at longer scales (1-m and more) became more substantial and relatively important.

These findings will have direct relevance for the design of further experiments in the project. They also show the need to include short-range replication in working sensor arrays to limit contributions to uncertainty from variation at this scale.

The initial analysis of the first experiment in the project highlighted some problems with data from certain sensors in certain periods. An approach to data editing was agreed. Two data sets, an edited set from the first season, and a set from a second season, have now been analysed using a linear mixed model to examine the variation at the scales of the sample design, to provide confidence intervals on estimated mean mineral N concentrations, and to examine how the uncertainty in estimates of mineral N concentration would differ between sensor arrays with different levels of replication at different scales. The analyses have shown :

(i) how the data from the sensor array can be analysed to provide a valid estimate of mean soil mineral N content of the soil day by day, with a representation of the uncertainty in this estimate in the form of a confidence interval. This is the sort of information stream from the underlying sensors which could be used to improve the timing and rates of fertilizer N bettet to meet crop demand.

(ii) that the sensors pick up real differences in the soil mineral N content between plots that have been managed differently.

(iii) that we can examine how changing the number of sensors, and the number of clusters in which they are deployed (each served by a data logger) affects the uncertainty in the estimated mean mineral N of the soil in a field. As sensor and logger numbers affect the costs of using the technology, being able to quantify this relationship is key to the design of feasible systems so that the costs are acceptable and the precision of the information generated is adequate to the task.
Exploitation Route While these are only provisional findings, ultimately these and further results from the project will have direct implications for how sensor technology is best deployed to provide real-time information to improve the efficiency with which nitrogen is managed in crop systems. This has the potential to improve crop yields while maintaining or improving nitrogen use efficiency, reducing costs to farmers and emissions to the environment.
Sectors Agriculture, Food and Drink,Environment
 
Title R code for the analysis of data from spatially nested sensor arrays 
Description The analysis of nested spatial data sets with linear mixed models and REML estimation is not novel. However, two new developments were required in this project. The first was the marked non-homogeneity of the variance of the mineral N data from plots under contrasting treatments in the second experiment. Interestingly the within-treatment distributions appeared normal, such non-homogeneity is often associated with a log-normal distribution of observations which can be dealt with by a transformation. That was not the case here. For this reason the LMM was extended to include scaling parameters for two treatments to scale their variance components relative to those for the control. This showed consistently larger variance in plots with applied fertilizer, and also temporal variation in the scaling factor. The code for analysis of these data also allows post-hoc evaluation of the variance of the estimated mean from notional sensor arrays with different numbers of sensors associated with different loggers. This allows one to evaluate how using more sensors, or changing the configuration of the array (with cost implications) affects the precision of the resulting data. 
Type Of Technology Software 
Year Produced 2019 
Impact This code allowed the robust analysis of data from the experiment. It also allowed us to address the key objective of assessing the returns to increasing cost of a sensor array with respect to the precision of the data, and so their usefulness for tasks such as managing the rates and timing of N fertilizer.