greater extent, the effects of changes in single parameters cannot be easily
identified. Thus, to experimentally attribute effects to specific causes, standalone
versions or more specialized individual models may be the best choice (e.g., Pansak
et al. (2010) for interactions between maize yields and erosion at the plot scale),
whereas the approach taken here may be more appropriate for assessing the effects
of integrated scenarios on certain outcomes at the landscape scale, and not analytically dissecting a process into its component parts. In our case study, it was not
possible to unambiguously attribute plot-specific soil erosion rates or topsoil depth
to a specific soil conservation treatment, for soil conservation was applied only on
maize plots, and over time the locations of the maize plots – or the crops grown on
each specific plot – changed in the coupled model system.
Although the inclusion of agent decisions at the farm level led to more plausible
results, further improvements may be achieved by taking into account plot-specific
decisions. In practice, farmers monitor plot soil fertility and optimize cropping
decisions accordingly. Implementing these mechanisms may lead to a reduction in
spatial variability of soil fertility over the long run, as farmers would fallow
degraded plots or to increasing spatial variability as distance to the farmstead
may become a decision criterion to preferentially allocate labor and fertilizer
resources. However, to incorporate this into the model would significantly increase
the computing power required. A major challenge in future versions will thus be to
bridge the different aggregation strategies that exist between the plot and farm
levels, while using a reasonable amount of effort.
10.9 General Conclusions and Outlook
The fact that the case study locations presented in this chapter were all located in
the Uplands Program’s study area, plus the exchange among colleagues regarding
modeling approaches allows to compare trends modeled and observed in the field
and aspects of the different models. This chapter showed the different perspectives
of the models presented and the resulting level of detail used according to the
research question posed by each approach: Detailed spatially distributed biophysical processes need to be represented in the LUCIA model, which looks into matter
translocation in small watersheds. As land use change dynamics are of secondary
importance in this approach, they are exogenous to the model. In comparison, the
FALLOW model operates at a semi-quantitative biophysical level, but integrates
socio-economic factors. MP-MAS focuses on household economics and thus puts
less emphasis on biophysical processes, which are simplified in the TSPC as
nutrient balances or empirical equations on yield potential. CropDSS serves as a
decision support tool for policy makers that predicts yields and thus combines a
process based crop model with a large-scale GIS database, but clustering spatial
units and not taking matter flows in the landscape into account.
Each of these approaches has its validity and no single model can address all
questions that turn up within given settings. In general, the models applied to the
426
C. Marohn et al.
identified. Thus, to experimentally attribute effects to specific causes, standalone
versions or more specialized individual models may be the best choice (e.g., Pansak
et al. (2010) for interactions between maize yields and erosion at the plot scale),
whereas the approach taken here may be more appropriate for assessing the effects
of integrated scenarios on certain outcomes at the landscape scale, and not analytically dissecting a process into its component parts. In our case study, it was not
possible to unambiguously attribute plot-specific soil erosion rates or topsoil depth
to a specific soil conservation treatment, for soil conservation was applied only on
maize plots, and over time the locations of the maize plots – or the crops grown on
each specific plot – changed in the coupled model system.
Although the inclusion of agent decisions at the farm level led to more plausible
results, further improvements may be achieved by taking into account plot-specific
decisions. In practice, farmers monitor plot soil fertility and optimize cropping
decisions accordingly. Implementing these mechanisms may lead to a reduction in
spatial variability of soil fertility over the long run, as farmers would fallow
degraded plots or to increasing spatial variability as distance to the farmstead
may become a decision criterion to preferentially allocate labor and fertilizer
resources. However, to incorporate this into the model would significantly increase
the computing power required. A major challenge in future versions will thus be to
bridge the different aggregation strategies that exist between the plot and farm
levels, while using a reasonable amount of effort.
10.9 General Conclusions and Outlook
The fact that the case study locations presented in this chapter were all located in
the Uplands Program’s study area, plus the exchange among colleagues regarding
modeling approaches allows to compare trends modeled and observed in the field
and aspects of the different models. This chapter showed the different perspectives
of the models presented and the resulting level of detail used according to the
research question posed by each approach: Detailed spatially distributed biophysical processes need to be represented in the LUCIA model, which looks into matter
translocation in small watersheds. As land use change dynamics are of secondary
importance in this approach, they are exogenous to the model. In comparison, the
FALLOW model operates at a semi-quantitative biophysical level, but integrates
socio-economic factors. MP-MAS focuses on household economics and thus puts
less emphasis on biophysical processes, which are simplified in the TSPC as
nutrient balances or empirical equations on yield potential. CropDSS serves as a
decision support tool for policy makers that predicts yields and thus combines a
process based crop model with a large-scale GIS database, but clustering spatial
units and not taking matter flows in the landscape into account.
Each of these approaches has its validity and no single model can address all
questions that turn up within given settings. In general, the models applied to the
426
C. Marohn et al.