The pattern of soil settling over time was distinct for the two locations (soil types); median soil height declined fast over the first 11 days and slowly thereafter in Kirton (see Figure 3-30). In Wellesbourne soil height declined in a more constant way and only half as much as in Kirton. Interestingly, the amount of rainfall over the first 27 days in Wellesbourne was much higher than over the first 25 days in Kirton (see Table 3-8), suggesting the pattern of soil decline is more inherent to soil characteristics than amount of rainfall per se. In both locations, the pattern of decline appeared similar regardless of plough depth. In fact, the difference in median height between soils ploughed at 18 and 24 cm that existed just after ploughing remained intact until the end of the sampling period (i.e 81 and 99 days after ploughing (DAP) for Kirton and Wellesbourne respectively).
From the frequency distributions of changes in soil height over different time intervals it is clear that the average decline in soil height at Kirton is not constant over time but most pronounced in the first 11 days, lower in the next 28 days and even less in the last 42 days (see Figure 3-31).
Figure 3-30 Change of median soil height over a three month period after ploughing; data
from replicates ploughed at 24 cm in Wellesbourne was not pooled because date of ploughing was different. Standard error is calculated as SEmedian =1.253σ/ N
Furthermore, the sample variance (s2) of the data over the first 11 days was larger than the sample variance over the next two intervals, in other words, absolute changes in soil height are larger over the first 11 days than thereafter.
Comparison with model predictions
Due to different decline rates in median soil height at Kirton and Wellesbourne, it was not justified to pool the data of both locations for each plough depth at the end of the sampling period. Whereas immediately after ploughing the observed median soil height for ploughing at 18 cm seemed to agree most with the predicted median soil height given 2 slivers, this had increased to 6 slivers (Wellesbourne) or to more than 12 slivers (Kirton) at the end of the sampling period (see Figure 3-32). For soils ploughed at 24 cm the median soil height immediately after ploughing was reflected best by a value in between that predicted by models with 1 and 2 slivers and median soil height at Wellesbourne was a little higher and at Kirton a little lower than the median soil height for 3 slivers at the end of the sampling period. Because at each location the absolute change in soil height was the same for both plough depths and the decrease in absolute soil height from one sliver to the next is much greater for a lower number of slivers (see Figure 3-24), more slivers were needed to explain median soil height after 81 / 87 DAP for 18 cm than for 24 cm. This highlights the
Figure 3-31 Frequency distribution of changes in soil height at Kirton, compared for
consecutive time intervals after ploughing (DAP = days after ploughing). Data from two replicates and both plough depths were pooled (N=484) since no differences were detected between their individual frequency distributions.
fact that the number of slivers in the model should not only be a function of the degree of soil compaction but also of plough depth.
3.4.4
Discussion and model implementation
Had soil height readings been taken over a larger area, it is possible that lower variation along compared to across the direction of ploughing would have been detected as was observed by Zhixiong et al. (2005). The relief aspect is less crucial however than the systematic underestimation of soil expansion after ploughing and the way soil settling over time is addressed. From this experiment it is clear that soil expansion occurs not only because of the voids created by the diagonal panning of the slivers but because soil volume itself expands, i.e. macro pores are created over the entire depth of the plough profile. Since germination probability is strongly influenced by a seed’s vertical position in the soil, in a model that underestimates the thickness of the soil layer, seeds will be incorrectly assigned to the zone from where emergence can be successful, thus overestimating the numbers of emerging seedlings. Obviously, this affects only the scenario in which soil is ploughed in autumn and left over winter.
Figure 3-32 Comparison between observed and predicted distributions of soil height, 81 (Kirton) or 87 days (Wellesbourne) after ploughing, for soils ploughed at 18 and 24 cm. Outliers are represented by red asterixes and are defined as in Figure 3-28.
In addition, this experiment showed that soil settling is more simply modelled as a natural compaction of the soil profile independent of plough depth, rather than as an increase in soil slivers over time which is dependent on plough depth. The Colbach and Roger-Estrade models require an estimate of the number of slivers, both for modelling seed movement and as soil settling over time. This would add a further component to ECOSEDYN and given the more straightforward use and interpretation of transition matrices and the problems identified with the Colbach and Roger-Estrade models in this study it was decided this added complexity was not warranted.
The scenario of plough cultivation in autumn and then leaving the field over winter will not be implemented because of the failed experiment (Section 3.2). Although the information on slumping has contributed to deciding which model to choose, it bears no further use in the modelling framework.