Precision Agriculture, 2, 39᎐54 2000
ᮊ 2000 Kluwer Academic Publishers. Manufactured in The Netherlands.
Managing Uncertainty in Site-Specific
Management: What is the Best Model?
MATTHEW L. ADAMS,
SIMON COOK, AND ROBERT CORNER
CSIRO Land and Water, Pri
ate Bag, P.O., Wembley 6014, WA, Australia
Abstract. Models which enable the representation of spatially variable crop performance are central to
site-specific management. Rarely have these models been considered in relation to the different sources
of uncertainty facing the decision maker. This paper describes various sources of uncertainty temporal,
metrical, structural, and translational in the context of the site-specific management problem and the
model types proposed to solve the problem. An example involving the site specific application of
nitrogen fertilizer in dryland wheat production is presented to show how these model types might be
evaluated for suitability to a given situation. We conclude that in data poor situations, knowledge-driven
models may be less accurate but preferred by the farmer, while in data rich situations data-driven
models may be more appropriate.
Keywords: model selection, uncertainty, decision-making
Introduction: Decisions and uncertainty
Decisions in broadacre agriculture are inherently risky because they are character-
ized by large uncertainties about their consequences. Anderson et al. 1977 specify
the decision problem more explicitly according to the following components: states,
acts, outcomes or consequences with their likelihoods, and choices or strategies in
response to expected outcomes. For example, consider the decision whether or not
to apply a given rate of fertilizer at a location, i, at a given time, t. The decision can
be influenced by the following unexhaustive list of uncertainties:
1. Uncertain states caused by spatial or temporal variation:
The demand for fertilizer will vary over space and time, depending on the rate
of crop development and other limiting factors, such as water availability.
The ability of the soil to provide nutrient varies widely, depending on
antecedent conditions, current status and conditions through the growing
The ability of the crop to acquire nutrient on demand, if available, may be
influenced by competition from weeds or constrained by disease.
The value of additional grain or grain of higher protein may vary over time as
a result of fluctuations in commodity markets.
The risk of nutrient leaching may vary as a result of heterogeneous climate
and soil conditions.