Agricultural and Forest Meteorology 101 2000 203–216
A method of land evaluation including year to year weather variability
Gordon Hudson
∗
, Richard V. Birnie
Land Use Science Group, Macaulay Land Use Research Institute, Craigiebuckler, Aberdeen AB15 8HQ, Scotland, UK Received 14 May 1999; received in revised form 5 November 1999; accepted 16 November 1999
Abstract
Land evaluation is sensitive to the effects of annual variability in weather. A method to incorporate this variability into land evaluation systems is proposed, using the land capability system for Scotland as a case study. Land capability classes
were found to be sensitive to the climate reference period from which data are taken. Individual stations rarely occupy their long-term land capability class. In addition, the relative position of stations in the land classification alters from year to year,
indicating variations with time in spatial correlation structures. Markov chain analysis was used in a risk assessment approach to estimate the mean return time to a land capability category for individual stations and for areas of land. The main conclusions
were: that land evaluation systems should not be applied using data from a different period to the baseline weather period used to establish the classification; there is a need to establish whether groups of stations tend to behave in similar ways over
space and through time; mapping zones of risk could provide a means of formally incorporating weather variability into land evaluation. ©2000 Elsevier Science B.V. All rights reserved.
Keywords: Land evaluation; Climate variability; Markov-chain; Risk assessment; Agriculture
1. Introduction
Land evaluation is internationally recognised for providing qualitative information about land, such
as its cropping potential or land degradation risk. It comprises a range of methods developed to enable
the assessment of land in terms of either capability for general land uses e.g. agriculture, forestry or
suitability for specific crop types e.g. wheat, bar- ley. Pioneering work on developing land evaluation
systems was carried out in America Klingebiel and Montgomery, 1961 and developed subsequently by
the Food and Agricultural Organisation FAO, 1984 for application mainly in Africa. In general, the sys-
∗
Corresponding author. Tel.: +44-01224-318611; fax: +44-01224-311556.
E-mail address: g.hudsonmluri.sari.ac.uk G. Hudson.
tems use a range of land qualities, derived from measurable land characteristics, to classify land. The
land qualities are modelled values that include crop growth requirements e.g. temperature regime, mois-
ture availability, impinge on management practices e.g. land workability and affect land conservation
e.g. erosion hazard FAO, 1984. The land quali- ties are derived in a variety of ways. The simplest
method is by direct conversion; e.g. the potential for mechanisation based on slope calculated from
elevation differences. Indirect interpretation is made using relationships between land characteristics, e.g.
soil hydraulic conductivity can be modelled from more easily measured soil properties such as texture
or structure. Crop simulation models can be used to integrate several land qualities to predict yields.
There are several generic issues associated with the application of land evaluation. In particular, land qual-
0168-192300 – see front matter ©2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 1 9 2 3 9 9 0 0 1 5 8 - 6
204 G. Hudson, R.V. Birnie Agricultural and Forest Meteorology 101 2000 203–216
ities derived from measurements of dynamic variables e.g. temperature are converted to static variables for
the purposes of land evaluation. For example, tem- perature measurements are converted to land qualities
such as length of growing period or accumulated temperature summed over a growing period. These
land qualities, derived for a single seasonal cycle, are then summarised over a sequence of consecutive
years, generally using robust statistics e.g. median. Long-term summaries are used to construct empirical
land evaluation systems, exemplified by Bibby et al. 1982 and the FAO 1996. A key weakness in using
summarised land qualities is that by treating dynamic variables in a static way much of the variability that is
an essential property of the land is removed. Farmers do not farm average landscapes under average cli-
matic conditions. So, whilst land evaluation methods based on this approach are of value in land use plan-
ning, for land management decision making it may be more useful to have information on variability from
which risk may be assessed.
There are few examples of land evaluation where dynamic variables have been explicitly included. van
Lanen et al. 1992 showed that it is possible to de- velop a mixed approach for incorporating weather
variability. In their method, land is first classified qual- itatively using biophysical or socio-economic data to
form land evaluation units. Then dynamic variables are re-incorporated uniformly within each land evalu-
ation unit using a representative climate station. They have applied this approach using simulation modelling
with data from single climate stations. This attempt to incorporate dynamic elements is rare and a recent
review by Rossiter 1996, in which he explored the theoretical basis that could underpin land evaluation,
highlighted the lack of internationally accepted meth- ods for incorporating dynamic variables.
This paper explores the effects of long-term decadal and short-term annual weather variability
on the classifications derived from land evaluation systems. The aim is to develop a robust and repro-
ducible method for incorporating weather variability into land evaluation to make it more relevant to land
management problems. The method is developed us- ing the Land Capability for Agriculture LCA clas-
sification system as applied in Scotland Bibby et al., 1982 as a case study. The effects of using weather
data from two different periods on the LCA classifica- tion are described. In addition, a method is developed
to enable inter-annual variability in weather to be quantified in terms of the mean return time to a land
category based on the LCA classes. The mean return time is derived from analysis of transition sequences
between categories, and uses concepts developed from formal risk assessment procedures. Finally, regional
estimates of the mean return times are mapped using interpolated weather data to show how the risk can
be expressed geographically.
2. Methods