It’s is pretty ancient advice. But what if you knew that the sun was going to shine all week and you could safely do another urgent task and still not miss out on the haymaking weather? Welcome to the world of climate & weather forecasting for agriculture.
I recently had the privilege of chatting with meteorologist Jaclyn Brown from CSIRO which got me thinking about this in more detail. She pointed out that historically, forecasting was based on previous years’ data, but climate change is rendering this less useful so other tools are needed. Climate modelling allows a narrower range of outcomes. Such forecasts are still probabilistic, but more reliable.
“What if you knew you could safely do another urgent task and not miss out on haymaking weather?”
The long range weather forecast
Three big improvements are making climate forecasting ever more useful for agriculture. It remains the case that the longer the timeframe, the more the uncertainty increases, but improvements are continual. Firstly, better prediction. The reliability of forecasts is greater than ever before, whether you want to know what will happen this afternoon or next week. Secondly, meteorologists are able to forecast – with useful levels of reliability – over longer and longer periods of time. I see your 7 day forecast and raise you a 6 month forecast. Thirdly, the granularity is steadily improving. This means both geographic (the storm passes through your county but the next one is sunny) and temporal resolution.
What sprang to mind for me initially when thinking about forecasting in agriculture, is the farmer using a long range forecast to decide, for example, when to harvest or muster livestock. Or, over a longer timeframe, decisions like whether to choose risk a drought resistant variety vs a higher yielding but water dependent variety become possible. But actually, the farmers themselves are, generally speaking, not the initial beneficiaries of this pushing these technological frontiers.
For the individual farmer, an extremely high level of granularity is required. The farmer needs to know whether the rain will fall on her/his land, not the average for the region. The timing of rainfall or heatwave is also critical – for crops such as wheat, timing can have a much bigger impact than the fact of its occurrence per se.
“farmers themselves are generally not the initial beneficiaries of pushing these technological frontiers”
Ambitious projects using ground based sensors and/or novel algorithms to derive increasingly localised forecasts are making this a realisable dream. Projects like CSIRO’s Digiscape project, companies like Ignita and aWhere, to name just few. In general, however, meaningful forecasting is currently at the scale of “100’s of km”.
Another reason the individual farmer isn’t an immediate beneficiary, is that handling probability isn’t their core competence. Translating the forecast probabilities into actionable insights for the farm is a task in itself. As Jaci Brown puts it, the outputs for climate forecasts are “not which way to jump, but which way to lean”. So if there’s a 60% chance of wet harvest, how should I use that to decide on when to start, or how to trade off cost of drying grain vs risk of losing a harvest? There are legions of entrepreneurs, user experience professionals and social scientists working on different ways of making such insights useable.
So actually, industry players who work at a regional or national level are the first beneficiaries. Armed with insights into the next month or 6 months weather, a whole range of organisations who use probability as their lingua franca can make more informed decisions about how to allocate resources over the coming season. Commodity traders, bulk handlers who need to know how the country or the region’s harvest will benefit. Numerous start-ups – such as Understory and Skymet or got going by focusing on providing forecast data to the insurance and financial services markets.
No data set is an island
But before we get too carried away, it’s worth reflecting that climate isn’t everything.In fact, forecasting models can be made much more powerful for many purposes when linked with other data sets.Knowing soil type, for example, is a crucial component of yield forecasting, as is knowing at aggregate level how many hectares of each crop is planted.This is one of the reasons a lot of work is ongoing on to use satellite data to differentiate between crop types.
As with the grand arc of software development in other industries, there comes a point where companies realise they cannot, and don’t have to do it all. Farm management software companies are steadily realising that they need to link with others; becoming a component that links well with other systems, or a platform upon which niche services can be built is probably the future for many of these services. This is exactly the path that Climate took, after its acquisition by Monsanto. It created a network of sensors and allowed other providers to access its data platform, aiming to create a platform for agriculture specific companies focused on data and sensors.
“companies realise they cannot, and don’t have to do it all”
Of course, here we run into data ownership and data blowback issues which must be navigated in all big-data environments. What level of transparency is appropriate if, by using services on the Climate’s FieldView platform, or any other, your data in some form is being shared with others? Will the abuses of personal data be repeated in Agriculture with companies’ data? What incentive does a farmer have to share data about their farm and practices if it might result in the cost of the services and products they buy increasing? Anyone who’s paid the price of out by making a decision to be totally transparent their insurance company will recognise this feeling.
All of this will be debated and resolved over time. Integration of disparate data sets and data rights are themselves topics for a whole separate post. For now, the utility of being able to better anticipate what the increasingly variable climate will throw at us remains enormous. The trend towards finer resolution, longer range forecasts and greater accuracy can only be beneficial.