Geoscience Reference
In-Depth Information
Table 25.2
Examples of key biotic and abiotic variables in arable agricultural ecosystems
which would need to be quantifi ed in terms of variability, in order to allow prediction of
crop yields and to inform ecosystem management
Biotic
Abiotic
Crop genotype/variety
Soil texture
Crop density
Soil type
Crop yield (biomass)
Available soil moisture
Crop health
Soil organic content
Pathogen presence
Soil nutrients
Herbivore presence/abundance
Drainage (surface and subsurface water movement)
Weed presence/abundance
Topography
Precipitation
Temperature
of ecosystem variability (e.g., Kerr and Ostrovsky, 2003; Bocchi and Castignanò,
2007).
Regardless of the data collection methodology, common statistical metrics are
used to defi ne variation over set periods of time and space, such as mean, median,
range, standard deviation, frequency, and size and shape of distributions (e.g.,
Landres et al., 1999). However, one of the main problems with measuring and
quantifying variability in complex ecosystems is that different processes or compo-
nents will require different descriptors of variability. For example, quantifi cation of
the variability of fi re disturbance within a landscape will require measurement of
fi re frequency, intensity, size, and spatial arrangement. Measurements of fl ood dis-
turbance will focus on fl ood frequency, discharge, stage variations, and spatial
extent of riparian and fl oodplain inundation. Measurements of species population
variability might include repeated counts or estimates of individuals, quantifi cation
of life history cycles, and metapopulation dynamics (Landres et al., 1999). All of
these can be expected to vary notably over time, and so either long-term monitoring
(which is often unfeasible) or reconstruction of past conditions is necessary, though
this too has limitations. Time lags are a common feature of non-linear processes
and also cause problems with obtaining and interpreting measurements.
For all of these reasons, quantifying the variability of ecosystem states and
process is diffi cult (e.g., Adachi et al., 2005). Nevertheless, such measurements are
essential if predictive models are to be utilised for ecosystem management.
Effectiveness of Prediction in Ecosystem Management
All of the issues discussed above create problems for ecosystem and resource predic-
tion. But how unreliable are our current efforts at prediction? Modelling has tradi-
tionally been based on specifi c data and is closely related to the measured parameters
of the ecosystem as it was during the measurement period, which limits its wider
prediction potential (Tan et al., 2006). Empirical models use a broader range of
data and parameters to make general assumptions about a system, and to predict
system response where data are unknown. Often these assumptions greatly limit the
applications of empirical models, and there is now more focus on non-linear models
to predict ecosystem response and resource management (Tan et al., 2006;
