Farmers are engulfed with several hardships and challenges to help make their living. While most of the challenges can be taken care of, one that remains truly out of control is the weather.?
And irregular weather conditions have the potential of destroying everything a farmer has worked on, for days.?
However, IBM¡¯s Weather.com tries to solve this with the help of accurate AI forecasts and analytical tools that help farmers better plan their farming habits and help them avoid losses, while also helping them better their yields with crops that could thrive in particular weather.?However, the approach is a lot more indirect. We got to interact with Himanshu Goyal, India Business Leader, The Weather Company to better understand how they make it happen.
We often see how accurate temperature or rain data can get today on our smartphone. But in the background, a lot is going on to source the data in order to make analysis, explained Himanshu, ¡°This is, this whole activity is about data crunching. So we get data from over 50,000 flights, which take off and land in 24 hours. This is reduced now due to COVID-19, but if you dissect this 50,000 or 41,000 commercial flights. We get data from the atmosphere at various heights and various times.¡±
He further added, ¡°We also get data from companies who have mobile apps, as well as people who opt-in for our mobile data, with their location. So we get pressure readings from these smartphones. We also get data from a lot of government agencies nationally or internationally who are doing prediction of weather, wherever it is publicly allowed for third parties like us to be picking that data or usage.¡±
After collecting a ton of data, it won¡¯t mean anything if it's not processed appropriately, Himanshu explained, ¡°We take it into over 150 odd proprietary models. We have over a hundred meteorologists who work with the company, synthesize the data and bring it out in a form of consumable and decision making data for a variety of businesses.¡±
Another concern is the accuracy of the data that¡¯s collected. For that, they take help of famous meteorologist Jeff Master¡¯s crowd-sourced weather platform called ¡®Weather Underground¡¯ ((first acquired by The Weather Company, which was later acquired by IBM).
Himanshu explains, ¡°This has a crowdsourced collection of weather station data. What it does is that it actually helps us verify digital data at the location where this weather station probably is forecasting some data as well. So over here, a lot of government sources, let¡¯s say the weather stations on the aeroplane airports or any vital installations, uh, private or public, that data is used for verifying what we produce.¡±
He added, ¡°Now, all this data is then updated every 15 minutes and it is given at a resolution of half a square kilometres. Looking at this level, there are around 2.2 billion points on Earth and these points are updated every 15 minutes with our data. So that's the level of frequency we¡¯re updating this data. And this data is then consumed by the aviation sector or the energy sector or retail sector on an active sector or agtech sector.¡±
Himanshu revealed how initially, the weather forecast was provided in 10-kilometre zones but over the years they¡¯ve taken it to a much deeper level that helps in not just accurate weather information but a whole lot more.
¡°Until last year, we used to make a grid of 10 kilometres in the atmosphere and then figure out what the weather in that respect of 10 kilometres would be. Then we would downsize that 500 by 500 unit. I could, for instance, say that in Andheri there would be rains or let¡¯s say in Lower Parel there would be rains but that wasn¡¯t good enough.¡±
Himanshu added, ¡°After we came to IBM, we further decreased this bridge from 10-kilometre grid to a three-kilometre grid (the small pixel inside the larger grid) which has now started getting into a forecast and also in our B2C (business to consumer) apps, um, everywhere is essentially now I can tell you whether it will rain in Andheri East and not in Andheri West.¡±
While hyper-local data is all great for companies, we were curious to know how it helps the farmers or ag-tech startups. Himanshu explained, ¡°In ag-tech, there are two kinds of variables -- controllable and uncontrollable variables. Controllable variables are the ones which a man controls. So I'll decide when to fumigate, when to irrigate, when to put seeds, any decision a farmer will make.¡±
He added, ¡°But there are a lot of things that aren¡¯t in a farmer¡¯s control, like from how many days the sun will come up (we call it the growing degree days), how much precipitation will the farmer witness, what would be the temperature like for some crops, what would be the soil moisture etc. So we are as a company after these uncontrollable variables. Based on this data, then they are able to figure out what volume of these uncontrollable variables they need to spend. This is pretty much what we do.¡±
IBM¡¯s Weather Operations Centre is essentially a B2B application -- meaning that it is meant for helping companies. However, each company has ties with farmers who provide them with either wheat, grain, pulses etc and the data is shared via these companies to the farmers to better their farming yields.?
Himanshu explained,¡° ITC for example is now producing a lot of food products, various clients would look at pulses etc. When they have a steady running business, what really happens is that they have to look at the condition where things have to stay stable rather than improving quality.¡±
He added, ¡°So based on the longevity of their business and their desire to be in that business, they will start investing time and energy with the farmers. This way they will then try to build more farmer connections and they will try to enable the farmer get? better output with consistent quality.¡±
Himanshu gave us a glimpse of the dashboard used to observe a small farm, where not only did it show weather data of the location in real-time, but it also had some really cool tricks up its sleeve.?
For example, the dashboard could have an alarm in case you don¡¯t want the temperature to go over a certain degree for better crop health or using an AI algorithm to recognise crop health data based on the level of chlorophyll in a plant with the help of satellite imagery. There are farms which also can get data dependent on local sensors.?
He revealed how data on such a microscopic level helps then not just predict the weather, but also predict people¡¯s consumption patterns, ¡°This helps us to do very interesting things like determining consumption of retail -- for example soft drinks or water -- in a hyper-local level every year. There could be companies we're dealing with beverage sales and with their past sales data and our weather data. We are able to bring the artificial intelligence engine to see in five years with a certain degree of change in a certain locality in Mumbai so much water was consumed.?
He added, ¡°So on the basis of each forecast the next day we can determine seeing the trend of the last five years, what would be the consumption in that location for that particular soft drink or water. So that's the value we're able to bring to our clients.¡±
Himanshu shared how India is right now at the correct track with respect to embracing technology in farming and helping the nation innovate on a much diverse level, ¡°A couple of things which are happening in this area, with the whole advent of data and AI capabilities in India, which is growing, the availability of satellite data is also decreasing its costs constantly.¡±
He added, ¡°The government is also trying to open up resources for people with their data. And, there's proven success also -- success stories that are also coming with the usage of modern technology in agriculture and various other fields. So I think the ecosystem is fairly vibrant at the moment. The more data democracy we have, a lot of these things will improve.¡±