Award details

Model-based statistical methods for doing experiments at landscape scale

Reference	BBS/E/C/00004726
Principal Investigator / Supervisor	Professor Richard Lark
Co-Investigators / Co-Supervisors
Institution	Rothamsted Research
Department	Rothamsted Research Department
Funding type	Research
Value (£)	27,705
Status	Completed
Type	Institute Project
Start date	01/11/2004
End date	31/10/2006
Duration	24 months

Abstract

In conventional experimental design and analysis the effects of environmental variation are consigned to additive block, covariate or residual effects. This assumption may be plausible for classical trials conducted on plots within a single field, but is dubious if we wish to conduct experiments to compare the effects of treatments at the scale of a landscape. It is increasingly important to be able to do this to meet the requirements of policy makers. For this reason a model-based statistical framework, based on the geostatistical model, is proposed for the analysis of landscape scale experiments. Here treatment responses and contrasts are modelled as functions of location in space, and are estimated by kriging. This framework will be developed and then applied to a case study with a relatively simple treatment structure, but responses likely to vary across the landscape. The objectives of this proposed work are as follows. 1). To develop methods for the design and analysis of experiments at landscape scale using model based statistics to treat the response to a particular treatment as a realization of a random function with a more or less complex structure. This will allow spatial changes in the response to a treatment to be handled, whether arising from of additive OR interactive effects of inherent spatial variation. In particular, we will be able to test whether the joint effects of the treatment and the inherent environmental variability appear to be additive or interactive, and then to estimate the local treatment response at sites or blocks (e.g. management units such as fields) across the landscape. 2). To test this approach in a simple case study where we can measure a proxy variable which should account for much of the landscape scale variation in responses in order to test the plausibility of the conclusions from the analysis.

Summary

unavailable

Committee	Closed Committee - Engineering & Biological Systems (EBS)
Research Topics	X – not assigned to a current Research Topic
Research Priority	X – Research Priority information not available
Research Initiative	X - not in an Initiative
Funding Scheme	X – not Funded via a specific Funding Scheme

I accept the terms and conditions of use (opens in new window)

export PDF file

back to list new search