2 Modelling – Part 1: Conceptualisation
2.5.3 Germination and emergence
2.5.3.1 Background
As part of the objectives for DEFRA, extensive work was conducted by Paul Neve to characterise weed seed populations of different origins of S. media and T. inodorum in terms of base-water potential and base temperature. In addition, studies were conducted in which the parameters for pre-emergence growth were derived. The derived information was implemented and added to an existing model for carrot germination so that the effect of relative crop and weed emergence could be studied in detail.
It was impossible from a time point of view to implement that data in ECOSEDYN as developed within this Phd. There were two alternatives: either implement a much simpler germination model component or use the more accurate model to produce output that could be used in some way as input data in ECOSEDYN as developed within this Phd.
In terms of answering the research questions, the germination model component that delivers the most accurate data should be preferred but considering the construction of a modelling framework of different components the inclusion of an autonomous germination component is preferable. Since designing and implementing a simpler conceptual component model for germination and pre-emergence was deemed to be more time-consuming than using the output produced by the more complex germination and pre-emergence model as developed by Finch-Savage et al. (1998) and Rowse and Finch-Savage (2003) at the Seed Science Group at Warwick HRI the latter option was chosen.
The germination and emergence model as produced by the Seed Science Group at Warwick HRI (2008) predicts the daily number of germinated seeds based on hydrothermal time (HTT) (Gummerson, 1986) or on the principles as implemented in the ‘Virtual Osmotic Potential’ model (Rowse et al., 1999) and the number of emerged seedlings based on post-germination seedling growth (Whalley et al., 1999). For the crop and weed seed germination scenarios only the HTT option was used. The model runs on the assumption that weed seed germination occurs predominantly as a consequence of seedbed preparation. The day of seedbed preparation therefore is set as the ‘trigger’ from which hydrothermal time is accumulated.
2.5.3.2 Parameterisation and implementation
For each of the 85 weather year - sowing time combinations the number of emerged carrot seedlings / day was recorded and saved in tables. Each simulation comprised 500 carrot seeds sown at a dept of 12 mm. ± 2.0 (standard deviation). The simulation lasted from the day of sowing (seedbed preparation) until 60 days later. Regarding the other settings in the model: the parameterized carrot germination data originated from Paul Neve, ‘Option 1’ was checked and was fitted using the HTT method as implemented by ‘Carole’. In fact, after all the simulations had been completed it became clear that the final percentage carrot germination did not vary much and the only characteristic that was assumed to impact on weed population dynamics was the day at which 50% of the carrot crop had emerged, d
Carrot
Cs. This was calculated from the
data in the tables using MatLab.
The model has not been parameterized for winter wheat and therefore a simplified set of decision rules was implemented. In principal the timing of 50% emergence was timed to take place after 150 day degrees (above a base temperature, Tb, of 0) had been accumulated (Hodges and Ritchie, 1991). If within the interval of one week prior to crop sowing to one week after crop sowing the cumulative amount of rainfall was less than 10 mm or more than 50 mm, a delay of 7 days was imposed.
Winter wheat
Although parameterized hydrothermal time models were available for T. inodorum
the soil depth structure (single point, e.g. 1.5 cm deep) of the germination and emergence model did not match with the soil depth structure (layer, e.g. 0.5-1.0 cm deep) in ECOSEDYN. The germination model requires specifying a depth and ‘spread’ and then allocates a seed distribution according to a normal distribution. The maximum depth from which T. inodorum can germinate is roughly 15 mm (Grundy et al., 2003a). By overlapping normal distributions with the same standard deviation but different means, a uniform distribution can be generated over most of the relevant interval. Using an Excel spreadsheet, seed depth distributions were generated with different combinations of mean (seed depth) and standard deviation (‘spread’) and
the soil surface are in fact allocated to the surface. An unrealistic number of seeds at the surface is likely to affect the germination and emergence results. Therefore the ‘shallowest’ normal distribution should contain a ‘mean’ and ‘standard deviation’ such that results in as low a number of seeds at the surface as possible. The best compromise between the uniformity of the distribution and the number of seeds at the surface was found when the first normal distribution had a mean of 1 mm and a standard deviation of 0.4, the mean soil depth of the remaining normal distributions was every 1 mm down to 15 mm. For each soil depth there were 136 simulations (17 weather years, 8 sowing times (carrot + winter wheat). In total there were 2040 simulations with 500 seeds each of the non-dormant T. inodorum population as characterized and implemented in the model by Paul Neve.
The number of germinated and emerged seedlings / day was initially converted to a proportion / day relative to the total number of germinated seeds / emerged seedlings at the end of the 60 days. This is however re-calculated in MatLab to a daily proportion relative to the size of yesterday’s seedbank. The daily germination and emergence proportions for the 1 to 5 mm, 6 to 10 mm and 11 to 15 mm depths were then averaged to get an estimate for the 0-5, 6-10 and 11-15 mm soil layers in ECOSEDYN.
In raised bed systems the seeds experience severe soil compaction in the tramlines but no soil compaction within the beds. Weed species emergence due to soil compaction is variable with both positive (Jurik and Zhang, 1999; Boyd and van Acker, 2004) as negative effects (San Roman and Fernandez, 1991) reported. Compared to other weeds, T. inodorum seeds have one of the narrowest depth ranges over which they can emerge. This implies that the seedling does not have enough vigour to be able to emerge from other depths. It is therefore likely that soil compaction also reduces the number of seeds that can emerge. Two seedbanks are distinguished, the between-bed area (BB) and the within-bed area (WB). From the BB area fewer seeds are likely to germinate due to severe compaction. On the other hand, compaction would bring some seeds at a distance from where they could emerge whereas they could not before. Without data to indicate which effect would be more important, no reduction of germination in the BB area was applied. No distinction was made either between the within-row and between-row areas in the WB area.
In ECOSEDYN decision rules were introduced to reduce the predicted proportion of germination. Firstly, it was imperative to apply a depth-dependent germination reduction scaler. If the temperature and soil moisture are sufficient, the model
estimates that if T. inodorum seeds are placed at 3 cm depth or deeper, around 80% of the seeds will still germinate, which would result in 100% fatal germination and thus a massive depletion of the seedbank. Seedbanks of T. inodorum are relatively persistent (Thompson et al., 1997) which suggests that the seeds possess a depth- mediated germination response.
A germination reduction factor (GRF) based on seed depth was calculated based on a Beta distribution function:
Equation 2-5
(
)
− − − − − − + = e m s e depth depth depth depth s e s m e e max depth depth depth depth depth depth depth depth 1 GRF depth GRFwhere the maximum reduction of germination (GRFmax) was 0.9 and was reached at
16 mm depth (depthe). The starting depth at which germination was assumed to
become reduced (depths) was at 6 mm. and the point at which germination reduction
increases fastest (depthm
Fatal germination levels vary between 5 and 40% of the total proportion germination, depending on weed species, soil depth and presence of pathogens (Benvenuti et al., 2001b; Benvenuti et al., 2001a; Davis and Renner, 2007). In ECOSEDYN the value for fatal germination of T. inodorum below the layers for which germination and emergence was calculated by the germination and emergence simulation model (Seed Science Group, 2008) was set at 15% and 5%, in carrot and winter wheat respectively, over the 60 day interval over which germination and emergence was simulated to occur.
) was assumed to be at 12 mm.
Secondly, the degree of crop development determines to what extent weed germination is suppressed. A germination suppression factor increased linearly from 0.0 to 1.0 over the interval of critical period of crop competition (0.20 – 0.52 of growing period). The shorter the time from sowing to harvest, the earlier the critical period of weed competition is initiated. Hence, when comparing equal sowing times for varieties with different times of sowing to maturity, the shorter the time from sowing to maturity, the more germination is suppressed and therefore the lower the weed density in the crop.
The germination and emergence model gives one value for the proportion germination whereas in ECOSEDYN two seed states, sg−light and sg−dark, and four seed ages are
considered. Since seeds of the first age have not been produced yet at the time of seed germination there are in fact six separate seed categories that all contribute to the
overall daily germination. To ensure that each of the six seed categories is reduced by the appropriate proportion of germination the following calculation is carried out per soil layer:
1. The number of available seeds is calculated for each of the six groups (three seed ages, two seed states, sg−lightand sg−dark). Due to the more intensive
seedbed preparation for a carrot crop as compared to a winter wheat crop, it was assumed that 75% of the seeds that can germinate only after receiving a light trigger,sg−light, are ‘excited’, i.e. available, during carrot seedbed preparation but only 25% are excited during the winterwheat seedbed preparation. In contrast, all seeds in the sg−darkstate are available.
2. For each of the two seed states the number of seeds per seed age is expressed as a proportion of the total number of seeds of that seed state. For example, 65% of ‘sg−light’ seeds are 1 year old and 35% are 2 years old (since all seeds lose the light requirement at the end of the 2nd
light g
s −
year (see Section 2.5.1.3 in thesis), there are no ‘ ’ seeds that are 3 years old).
3. The daily proportion germination as calculated by the ‘Germination and Emergence’ model is multiplied with the proportion that each seed age of a particular seed state represents of the total number of seeds of that seed state (as calculated in 2) to obtain the total number of seeds that germinated.
4. This number is subtracted from the relevant category (seed age, seed state) of the seedbank.
The number of emerged seedlings per soil layer was calculated by multiplying the predicted number of emerged seedlings per soil layer by the ‘Germination and Emergence’ model with one minus the proportion pre-emergence mortality due to linuron application (see Section 2.5.4). This was repeated for the other two soil layers and the number of emerged seedlings was then summed. According to the ‘Germination and Emergence’ model the maximum period over which germination was predicted to continue after seedbed preparation was 60 days. Rather than account for the emerged seedlings on each day individually, a maximum of 12 weed cohorts were created by grouping the weeds of each 5-day period together. The median date of each interval was then assigned to be the day of emergence, dWs.