Dear All
 
Thanks very much for all the responses so far to my questions on model impacts. I don't think that a bit of 'soul-searching' (as one respondent termed it)
from time to time does any harm. Its helped me at least to clarify a few ideas. There were a lot of interesting ideas and I will try and summarise these,
particularly for those of you who only subscribe to one of the two listservers that this discussion has been occurring on, and add a few thoughts of my
own. I haven't digested everything yet, but there are a few points that I would like to pick up on:
 
There seems to be a general feeling that models have a useful part to play in development (but subscribers to these discussion groups are likely to say
that anyway, I guess!), but there seems to be a difference of opinions about the way that modelling should go to have a greater impact. One is that the
main limitation to greater use is that the models we have are not accurate enough and hence that we need to improve on them (e.g. Matthias
Langensiepen ([log in to unmask]) and others). The responses on the possible combination of simulation models with neural
network techniques probably also come under this heading. While I agree whole-heartedly with the need to have better models, I am not so sure that
this is the limitation to their use in development so far. I would have thought that what we have so far are quite capable of making a first stab at solving
questions to a reasonable approximation (after all it is more the accuracy of the relative response to a change in the inputs of the model that we are
interested in rather than an absolute value of yield or whatever). But with one or two exceptions (e.g. the RAPS model of Peter Harris) it doesn't seem
that they have been used for this. One of the reasons must be related to the second strand of thought that most developmental problems tend to be
socio-economic in nature and that crop/soil models "attempt to specify a small (and often not particularly important) component of a system with
excessive (and largely empty) precision." (from Peter Cox" ([log in to unmask])). So if biophysical models are to have any impact they must
link to socio-economic models (which according to Rodney Beard ([log in to unmask]) are in the ascendency!). I think that I would
agree with this, but again I am surprised that this hasn't already occurred. Other innovative potential ways of using models (e.g. to take out interactions
we already know about to cheapen and sharpen the sensitivity of field experimentation - Peter Cox) are also worth thinking about.
 
I think that Bob Caldwell ([log in to unmask]) made a valid point when he said that the impact effect of agricultural simulation models should be put
into context with other agronomic 'tools' and approaches - a lot of these also would probably be hard put to come up with solid examples of significant
impact too, at least if used in isolation. So probably modelling is no better or worse than a lot of areas. One that springs to mind (mine at least) is the
'farmer-participatory' approach that is being stressed at the moment (at least in DFID, so much so that very few projects are now approved if they don't
have a strong socio-economic component in them). And yet, in the area of tropical agroforestry, a senior scientist at ICRAF recently admitted that after
10 or so years of research (with a strong emphasis on socio-economic methodologies), they were still no closer to being able to say where and when in
Africa and elsewhere agroforestry techniques would or would not work. Hence the funding by DFID of a project on modelling of agroforestry systems with
the aim of addressing this question. My feeling therefore is that we have to see modelling just as another tool in the lineup of tools that we have at our
disposal (along with socio-economic methodologies and all the rest), and to use it where and when it is appropriate. No one tool by itself will solve the
problems. But it would be nice to have some concrete example to point to to say there's a case where modelling played an important role, even if it
wasn't the main role!
 
I guess all this leads to the question of how you measure impact on development. Much of the impact of modelling may be quite long term - for example,
 as Brian Baer mentioned, training programmes such as the SARP project at IRRI, although probably not contributing directly to poverty alleviation in the
short-term give the participants another way of looking at their agricultural research, as part of a system rather than an isolated segment of it, and the
benefits of this may not be realised to well into the future. Investment in development needs both short-term and long-term aspects, hence I find myself
disagreeing with Peter Cox's statement that "The opportunity costs, in terms of other development opportunities foregone, have also been high." Of
course we all want our own areas to be funded more, but the challenge is to get the balance right.
 
Just out of interest, our group is planning to use a tea simulation model (http://www.silsoe.cranfield.ac.uk/CUPPA/index.htm) to help prioritise
researchable constraints in the tea sector of Tanzania where we have the remit to develop the newly formed Tea Research Institute of Tanzania, using it
for yield gap analysis, identifying potential tea growing areas, evaluating optimum management practices (e.g. harvesting regimes, irrigation, fertiliser
policies) in different regions, etc. We would be glad to hear from anyone who has done a similar thing of any pitfalls that we should be aware of, or
suggestions of how we could be more successful (e.g. linking with socio-economic models, PRAs, etc.)
 
This is turning into quite an essay, so I think I had better shut up now.
 
 
=======================================================================
Dr Robin Matthews
Department of Natural Resources Management
Cranfield University, Silsoe, Bedfordshire MK45 4DT, UK.
Tel: +44 (0)1525 863008; Fax: +44 (0)1525 863384
Email: [log in to unmask]
WWW: www.silsoe.cranfield.ac.uk/agricsystems/people/RobinMatthews.htm
     ourworld.compuserve.com/homepages/rbmatthews/
=======================================================================