The development of cost-effective methods for analysing soil information to define crop management zones

Summary

Sector:
Cereals & Oilseeds
Project code:
PR171
Date:
01 January 1998 - 30 June 1998
Funders:
AHDB Cereals & Oilseeds.
AHDB sector cost:
£14,996 From HGCA (Project No. 0084,1/97)
Project leader:
R M Lark & H C Bolam Silsoe Research Institute T Mayr, R I Bradley & R G Burton Cranfield University P M R Dampney ADAS Boxworth

Downloads

project_report_171

About this project

Abstract

Yield maps of winter-sown cereal crops were obtained from five sites in England for two or more seasons. These sites contrasted strongly with respect to the underlying geology and the resulting soil variability. All sites had been previously investigated by soil surveyors and/or agronomists. The yield maps were analysed in two ways:

i) Spatially (by computing multivariate variograms). These indicate how great the spatial variability of the yield actually is, and the spatial scales at which it occurs over all available seasons. If all of the yield variation occurs over very short intervals, then the possibility of mapping underlying causes and responding to them by precision farming is probably limited.

ii) By continuous classification (fuzzy c-means). This technique sub-divides the fields into regions within which the pattern of season to season variation was similar. It is postulated that similar factors are likely to be limiting to crop yield within a region of the field which has shown similar season-to-season variation in yield.

The information on the spatial variation of crop yield and the classifications were then considered, in the light of existing knowledge about soil variation in the fields, and some additional field sampling was conducted to aid interpretation.

It was found that the extent to which yield is spatially variable differed substantially between sites, and this was well described by the multivariate varlogram of the yield maps. Sites which appeared uniform by this descriptor were also those which the detailed soil investigations had shown to lack soil variability.

At the most variable sites there was a clear relationship between the variability indicated by the analysis of the yield maps and that found in field investigations of the soil. The soil series classification used in the soil surveys did not always account for the important yield variations. However, there was evidence that single soil properties, particularly physical properties, were related to the yield variation. At one site, for example, where the Available Water Capacity of the soil had been estimated at a number of sample points, the classes defined from the yield maps accounted for a substantial proportion of the variation of this soil property. Soil series will be useful in accounting for within-field variation of yield where they differ substantially with respect to soil physical properties.

It was concluded that yield map analysis can aid investigations of soil variability at within-field scale; firstly by indicating the fields where such investigation is justified by the magnitude and spatial scale of the variation and secondly by providing a framework for a cost effective sampling strategy, since the classes defined from the yield maps appear to be related to underlying soil variation. The classes might be used as strata for sampling individual soil properties, or to guide initial investigative sampling to identify the principal soil types in the field. It was proposed that information from yield maps, and other sources such as digital elevation data or remote sensor data might be integrated in order to make best use of a limited number of costly soil samples. A rational procedure for using yield map data to proceed step by step from initial, low- cost assessments of spatial variation to more costly but clearly justified and directed sampling of individual soil properties is proposed.

Given these conclusions, what are the implications for precision farming ?

i) While all fields are variable, some are clearly more variable than others. Farmers would be advised to take a step-wise approach to investigating variation within their fields, rather than paying immediately for sampling and mapping of soil properties. Analysis of yield maps is proposed as a starting point. Firstly, farmers and their advisors may ask whether the yield variation within the field is substantial. Secondly, whether it occurs at a manageable spatial scale or is too intricate to manage. If it is concluded that variability is significant, and can be resolved and managed at a workable scale, then investment in some soil investigations may be justified. Our work in this report provides some standards for comparison in the analyses of yield maps from fields with differing levels of soil variability, but further research could enhance this.

ii) Any soil investigations should be carefully planned to be as cost effective as possible, making maximum use of all existing information. Thus rather than immediately paying for a grid survey of a spectrum of soil variables, none or few of which may emerge to be important, the farmer with his advisor could use the partition of the field using yield maps as an initial stratification. Each sub-region can then be investigated in turn, and the properties investigated within each sub- region could be carefully targeted. If, for example, a particular sub-region generally does well in all but dry years, then there is no point in looking for acidity or compaction problems, but an assessment of the risk of drought may be relevant.

iii) Having decided which soil properties are worth investigating and where in the field, then sampling should be planned to map properties at appropriate scales. Thus an estimate of the mean value of the property within one sub-region might be satisfactory in some cases, while other properties need to be mapped in more detail. In future it is to be hoped that maximum use will be made of remotely sensed information, data on topography, and other variables which can enhance the value of soil measurements and observations. Further research is needed to establish robust methods for doing this.

×