10.2.5 Discussion and Outlook
At the plot level, the magnitude of soil eroded from maize plots (Table 10.2) was in
the range of that found in the reference experiments carried out on similar slopes
and soils in Chieng Khoi (Tuan, personal communication). Simulated soil conservation measures on maize plots were effective at reducing soil erosion on these
plots and also on other plots downstream, although even erosion under soil conservation was at times considerable. Still, reduced erosion rates had a positive effect on
maize yields in the first years after implementation of the measures.
After 8 years, yields under the legume scenarios (C and D) dropped below those
under no tillage (B) and even those of the baseline (A). In this case, the initially higher
nutrient export of C and D through maize harvest could have led to soil mining, but
after a further number of lower maize yields under C and D, this tendency should have
been reversed again (which was not the case). Two potential explanations can be
given at this stage: (a) the higher water demand under crop plus legume as compared
to a single crop (effects of weather in years 7 and 17 point to water stress), and (b) the
effects of not burning on the availability of nutrients to plants.
While yields under scenarios A and B were above those of the legume treatments
C and D, higher nutrient export for the maize harvest could have been the cause of
the yield decline. However, this tendency would have reversed after several years of
lower maize yields under A and B. The fact that the high fertilizer scenarios did not
show the same trend supports this assumption.
At the landscape level, the effects of soil conservation measures on maize were
limited when looking at sediment loads leaving the entire catchment. Although
absolute quantities of eroded soil at the catchment outflow differed clearly between
scenarios, these differences remained small in relative terms (data not shown), due
to the fact that the large areas under forest and tree plantations, those contributing
little to erosion, remained unchanged between scenarios. Seemingly larger erosion
reduction effects in paddies, as compared to maize plots, stemmed from the fact that
the model simulated sediment loads and thus did not distinguish between eroded
soil originating from a pixel and such passing through a pixel (except for pixels
without an inflow, e.g., next to a ridge). As sediment from the entire catchment
passed the lowland and outflow cells, total amounts were always higher than in the
upland source cells.
The LUCIA standalone model captured the spatial variability in erosion and crop
yields observed in the field (Lippe et al. 2011). The high temporal and spatial
resolution of the model allowed us to identify erosion hotspots (in terms of reduced
topsoil thickness), distribution of sediment loads and patterns of soil fertility (e.g.,
high fertility along previously forested footslopes, outputs not shown) and their
development over time. The unchanged land cover and management practices over
25 years, even though not a necessarily realistic scenario, facilitated the tracing
back of causal relationships between variables.
In a coupled model with dynamic land use (Sect. 10.8), the effects observed here
could not be expected to appear to the same degree, because agents facing waning
yields would resort to different land uses or fertilizer levels. Given that soil
10 Integrated Modeling of Agricultural Systems in Mountainous Areas
383
At the plot level, the magnitude of soil eroded from maize plots (Table 10.2) was in
the range of that found in the reference experiments carried out on similar slopes
and soils in Chieng Khoi (Tuan, personal communication). Simulated soil conservation measures on maize plots were effective at reducing soil erosion on these
plots and also on other plots downstream, although even erosion under soil conservation was at times considerable. Still, reduced erosion rates had a positive effect on
maize yields in the first years after implementation of the measures.
After 8 years, yields under the legume scenarios (C and D) dropped below those
under no tillage (B) and even those of the baseline (A). In this case, the initially higher
nutrient export of C and D through maize harvest could have led to soil mining, but
after a further number of lower maize yields under C and D, this tendency should have
been reversed again (which was not the case). Two potential explanations can be
given at this stage: (a) the higher water demand under crop plus legume as compared
to a single crop (effects of weather in years 7 and 17 point to water stress), and (b) the
effects of not burning on the availability of nutrients to plants.
While yields under scenarios A and B were above those of the legume treatments
C and D, higher nutrient export for the maize harvest could have been the cause of
the yield decline. However, this tendency would have reversed after several years of
lower maize yields under A and B. The fact that the high fertilizer scenarios did not
show the same trend supports this assumption.
At the landscape level, the effects of soil conservation measures on maize were
limited when looking at sediment loads leaving the entire catchment. Although
absolute quantities of eroded soil at the catchment outflow differed clearly between
scenarios, these differences remained small in relative terms (data not shown), due
to the fact that the large areas under forest and tree plantations, those contributing
little to erosion, remained unchanged between scenarios. Seemingly larger erosion
reduction effects in paddies, as compared to maize plots, stemmed from the fact that
the model simulated sediment loads and thus did not distinguish between eroded
soil originating from a pixel and such passing through a pixel (except for pixels
without an inflow, e.g., next to a ridge). As sediment from the entire catchment
passed the lowland and outflow cells, total amounts were always higher than in the
upland source cells.
The LUCIA standalone model captured the spatial variability in erosion and crop
yields observed in the field (Lippe et al. 2011). The high temporal and spatial
resolution of the model allowed us to identify erosion hotspots (in terms of reduced
topsoil thickness), distribution of sediment loads and patterns of soil fertility (e.g.,
high fertility along previously forested footslopes, outputs not shown) and their
development over time. The unchanged land cover and management practices over
25 years, even though not a necessarily realistic scenario, facilitated the tracing
back of causal relationships between variables.
In a coupled model with dynamic land use (Sect. 10.8), the effects observed here
could not be expected to appear to the same degree, because agents facing waning
yields would resort to different land uses or fertilizer levels. Given that soil
10 Integrated Modeling of Agricultural Systems in Mountainous Areas
383