Regression and decision trees in agriculture

When we normally think of use of science and technology in the rural sector, we think of mobile phones.  This week, I came across two very intriguing examples of use of scientific approach in agriculture.

My first example is from Ved Arya who founded Srijan.  He was working with soybeen farmers in Rajasthan, trying to help them improve their yields.  Farmers were using 14 different practices to grow soybeans.  He collected data on production inputs, climate etc. from 600 farms, and created a regression model with yield as the dependent variable.  Using regression analysis he was able to state that seven of the practices were useful and helped promote those, resulting in overall increase in yields for farmers.

The second example is from KC Mishra, who founded eKutir in Orissa.  He is developing science-based information services for farmers.  While it is straightforward to provide weather forecasts or information on today’s prices at the mandi, what he is trying to do is provide more actionable information for farmers.

Sunflower crop I had earlier written about Vijay Aditya’s Ekgaon Technologies which has developed SMS systems which send out alerts for soil preparation, weeding, etc. based on crop calendars.  Vijay’s organization depends on a team of agri-science researchers who provide science-based advice.

KC Mishra has taken this a step further.  He is developing algorithms for seed selection so that farmers can choose the best seeds for their farm conditions.  He wants to remove the need to consult experts for fairly routine decisions that farmers need to make. So he has developed a seed selection engine that can recommend the best seed varieties once you specify the soil type, local climate, irrigation, etc.

He is also working on pest advisory services on the same principle.  Imagine a farmer shows up and says there the leaves of his crop are drooping or there is rust on the leaves.  You already know the basic info about his farm (crop variety, soil type, irrigation type, etc.).  Could you input all these parameters into a system and get back the top three most likely causes of this?  This is the kind of algorithm and decision trees he is developing.

In a way KC Mishra is solving the same problem that healthcare initiatives are trying to solve in rural India.  Due to the paucity of doctors (India has one of the lowest per capita density of doctors in the world), they are either relying on para-doctors, para-nurses for early diagnostics or “de-skilling” it further by developing extremely low-cost diagnostic tools and algorithms.

The IT services industry in India prides itself on “de-skilling” software development through the use of robust processes.

What people like Ved Arya, KC Mishra and Vijay Aditya are doing is similar to the IT and medical professions: in the absence of easily available agronomists, and to reduce the cost of agri-advisory, they are relying on para-agronomist and algorithmic engines to provide diagnostic and advisory services.  Probably the algorithm based method can be used only for the more common diseases and pests, but even that is a big step forward from decisions based on gut-feel or advise from pesticide shops.

I hope that more and more NGOs take up such scientific methods in agriculture to provide valuable advice.

If you know of other such examples, do write in. I would love to feature more of them here on Stirring the Pyramid.

Advertisements
This entry was posted in Agriculture and farming, Science and technology and tagged , , , . Bookmark the permalink.

2 Responses to Regression and decision trees in agriculture

  1. Pingback: Soil nutrient management | Stirring the Pyramid

  2. Hi. How do you write to you?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s