A country scene of a maize or corn field with eucalypts and tall palm trees
In the spirit of leaving the hardest problem ’til last – how will I model the complexity of farmland:
- how changes in soil fertility effect the growth of crops (and weeds)
- the effects of weeds on crops
- effect of water (droughts and deluges) on crops
- what happens to seeds if they are planted onto soil in different states (germination likelihood)
This is a similar issue that we faced modelling fishing – how ‘realistic’ does the model need to be for the experience (of playing with the godel) to be interesting/useful/educational…or in our case, allow us to decide whether ABMs with game-like elements are a useful research tool.
It is easy to imagine a simple model rules:
- crops die (are strangled) with a frequency proportional to the weed concentration around them
- seeds can only germinate when the soil has a certain level of ‘preparedness’ (cleared, tilled etc.)
- plants grow at a rate that is hindered by too much water, and enhanced by fertiliser (as might the weeds)
- wind levels above a certain strength will push over plants of a certain height
- and so on…
Maize lifecycle
The basic heuristic we need to stick to is: create a model that presents interesting challenges in a semi-realistic and rule-based way. We want to ensure farmers are ‘asked’ what they would do with increased climate variability e.g. less/more water, stronger winds etc.
- soil patches (a kind of agent)
- lose fertility
- sprout weeds (with frequency)
- sprout crops (when farmer plants them)
- fertility (0-10 with a threshold at 7 above which crops and weeds grow faster)
- moisture (0-10)
- weed killer (Y/N, if yes, weeds die)
- cows
- wonder around trampling crops underneath (if any)
- speed/direction
- weeds
- grow at a certain rate
- die if too many crops nearby above a certain height
- crop
- grow at a certain rate
- die if too many weeds nearby
- sway proportional to wind strength (and fall over at threshold)
- height
- farmer
- sow seeds
- kill weeds by weeding
- spread weed killer (and so change weed killer chemical concentration of soil patches)
- spread fertiliser (and so change fertility of soil patches)
- weather (or rain and wind separately)
- add water to soil
- wind strength
Regarding the last comment, I googled ‘agent-based models of crops’ and there were plenty of hits including Google Scholar. E.g.
http://onlinelibrary.wiley.com/doi/10.1111/j.1574-0862.2001.tb00205.x/abstract
http://www.springerlink.com/index/3266272078436kv7.pdf
thanks – i got the first one, but shiboleth doesn’t work for the second (Springer publisher site). Will try again tomorrow and when back at work.
There is a fairly simple crop model coupled to ABM in this paper by Sukaina
doi:10.1098/rstb.2005.1742
Not sure how easily to adapt it for Cameroon context.
Perfect – thanks. This is like Alice in Wonderland – the further I get into thinking, the deeper the warren. Fascinating.
I can see the chain of thought that went into prioritising what gets into the model i.e. “Three crop choices are included in the Mangondi model (table 1). The possible number of choices observed during fieldwork was reduced to represent
salient drivers of cropping choices established during interviews and a participatory knowledge elicitation process”
We’re doing this project a little upside-down it seems – we hardly have any data about how farming is currently done in our study area. Maybe this won’t be a problem.