Can Nepal’s food security really be improved through modelling?
Scientists have developed a computer model for Nepal that shows income from different agricultural scenarios or market policies. But modelling is still not properly recognised
By Rachmat Mulia and Betha Lusiana
In January 2015 in Kathmandu, Nepal, in a small workshop with team members of a research-for-development project as audience, we introduced a beta-version model for estimating household income and food security in the country.
The model was developed as part of a 5-year project (2013–2018) funded by the Australian Centre for International Agricultural Research and the CGIAR Research Program on Forests, Trees and Agroforestry: Enhancing Livelihoods and Food Security from Agroforestry and Community Forestry in Nepal (EnLiFT). The aim of EnLiFT is to ‘determine ways for enhancing livelihoods and food security from improved implementation of agroforestry and community forestry systems and for better utilisation of under-utilised or abandoned agricultural land in the mid-hills of Nepal’.
About 80% of Nepal’s 27.8 million people are dependent on subsistence farming and more than 30% have a monthly income of less than USD 14. More and more rural young people find jobs in the city or go abroad to earn money. The lack of interest in managing land because of low economic returns and low soil fertility is responsible for the increasing amount of underused land in the country. Improving incomes for farmers is an urgent priority to slow the rate of land abandonment and increase the level of food security.
The current version of the model we created to help address these problems includes annual crop, agroforestry, livestock, and off-farm income as the main sectors. Applying it to the mid-hills of Nepal showed that it could explain income levels at baseline conditions and give insights into the impact of different scenarios representing plot-level agricultural innovations or national-level changes in market policy.
To start with, we translated our knowledge about household income in Nepal into the Stella programming language. One reason for using Stella as the model platform was to stop the model’s code becoming a ‘black box’ which only the modellers could understand. The software is not free to download, however, which to a certain extent prevents a wider application of the model.
Responses from the audience at the Kathmandu workshop to the model and its application were as expected: no one argued about the model’s concept but rather most were curious about the use of a model in general and within the project in particular. Questions prevailed of whether the modelling would have any direct impact on local people and the degree of confidence in the model’s results.
For us, these were relevant questions because most of the audience had had no involvement with modelling in previous projects and modelling was not a ‘stand-alone’ activity. A model needs inputs, so modellers need to collaborate with other team members to carry out surveys and field measurements. To be fair, at first glance it looks like nonsense that a set of mathematical equations in a laptop can actually produce anything meaningful for most people. Moreover, these equations were supposed to help improve the level of food security in Nepal. Surely, that is total nonsense! However, between those equations that make up the model and the people whose livelihoods are to be enhanced there is a set of innovations, interventions and policies that link them. The model isn’t the set of actions but it can be used to determine the quality of the actions.
Computer modelling has been a feature of agricultural research for decades. A well-designed model is a way of knowing more about a system and can be used as a base for designing hypotheses. As our colleague Prof Meine van Noordwijk has often said, ‘There is a great deal of statistical method to test hypotheses but we lack tools for designing them’. In the world of business, a model is a tool at your disposal to project loss or gain from your business strategies before you finally decide what kind of shop you will open. The same principle applies in the agricultural world.
At the very least, a model can give us more confidence that the set of actions we chose will not fall apart because the strengths and weaknesses have been tested in the model. Modelling saves a lot of time, energy and money rather than trying out each possible set of actions through trial and error.
Yet sceptics continue to claim models aren’t relevant to the real world while others paradoxically want to have a ‘one size fits all’ model that can be used to produce everything in a few mouse clicks. But if we had such a ‘magic-box’ model, we might view it as more useful or sensitive than others simply because of the ease of ‘playing’ with it. As modellers, while we acknowledge that ease of use and the comfort of users are critical ‘selling points’, we cannot forget the need for a sound concept behind the nice look.
We trust that gradually more and more people will understand, though not have to necessarily use, modelling and why it is important for research. Indeed, after the workshop in Kathmandu, one of the audience members expressed his sincere hope that the model could be used to convince young people to stay in their hometown and pursue agriculture. For us, that was an incentive to continue to develop the model to make it as effective as possible and also a reminder about the ultimate purpose of developing a model for the project: improving people’s livelihoods and securing their food supply.
Edited by Robert Finlayson
This work is supported by the CGIAR Research Program on Forests, Trees and Agroforestry