The rain had just stopped, leaving a gentle, wet breeze in its wake. I was standing on a farm 150 kilometers from Bangalore with a despondent farmer by my side. I had just resigned from NVIDIA with the intention of starting a company that provided high-tech farm management solutions. We were looking at his partially submerged saplings, when he remarked,
“I am not going to make any money this season”.
It felt like deja vu: almost exactly one year before, I was standing on the same farm when the same farmer uttered that very same sentence. However, the circumstances were starkly different then. It hadn’t rained at all, and his saplings were dying. He pointed out the sacks of fertilizers with “NPK” written on their sides in a small shed behind us:
“This is what I didn’t use last year. I probably won’t use them this year either.”
Commodities traders may find the information in this post relatable. Agricultural commodity prices often see ups and downs due to weather impacts on supply. In this post I will describe just how well supply and prices are tightly correlated by taking the examples of coffee, the returns traders can realize by predicting the supply months or even a year in advance, and how gap environmental forecasts can give them the edge and generate alpha.
Climate Risk and Reward
Arabica, the coffee variety consumed by the western world, was selling at $1.00/lb just a few years ago. Now it is selling at $2.33/lb. In 2011, it even hit $3.00/lb. The chart below shows just how much volatility coffee prices have seen in the last 25 years. Every peak and valley in the coffee price chart can be attributed to weather-related impacts on production in Brazil and Colombia, the two largest producers of Arabica coffee. Let’s examine the biggest peak, i.e., the one in 2011.
Before I go into the detail of what caused the peak, it is important to know that there is a predictable zig-zag pattern in coffee production: Coffee plants have a biennial cycle where they produce high quantities of berries during on-years and low quantities in off-years. So we do expect to see coffee prices fall during on-years and rise during off-years. We are only interested in explanations for what caused deviations from this pattern.
This chart shows just the deviations from what is predicted by the zig-zag pattern for Colombia. You can see how, in 2011, Colombia’s production was 20% below what is normal.
According to the USDA’s Foreign Agricultural Service (FAS), the causes of the production slump were "heavy rains that increased humidity and the occurrence of coffee rust, while cloudy weather created an ideal environment for the coffee berry borer". If you read any annual FAS report for coffee, you’ll see several recurring themes: whether production in Brazil and Colombia (for Arabica) was in line with their on/off-year normals or higher/lower than expectations, and how weather conditions inhibited or facilitated the spread of coffee rust and the berry borer.
Coffee rust is a fungus that covers coffee leaves with its spores creating brown patching giving rise to its name. It reduces photosynthesis in the leaves impacting the plant’s ability to produce beans. Temperature and moisture play the largest role in infection rate. The disease peaks at 21°C, but the spores can survive only if there are droplets on the leaves and it spreads only when there is sufficient humidity and wind. There are pesticides that can control the fungus, but it impacts the quality of the coffee and is prohibitively costly – up to 50% of total production costs according to NIH.
Coffee borer is a beetle that makes holes in coffee berries. The weaker the sunlight and the greater the moisture, the greater its spread. The USDA calls it the world’s most devastating coffee pest which can cut yields by up to 80 percent. And there is pretty much nothing growers can do about it. As this University of Florida article puts it, “The problem is that there are very few options for control of the coffee berry borer. How could farmers control an insect that spends almost its entire life inside a berry, protected from pesticides and natural enemies? Currently the only practical management option is to be very vigilant about the cleanliness of a coffee plantation, and to remove as many of the infested berries as possible during harvest”.
As you’ve seen, slumps in coffee production are almost entirely due to the weather. Pesticides and even biological pest control methods are largely ineffective against the coffee berry borer, and controlling coffee rust is prohibitively expensive for many growers. If you can predict precipitation, sunlight, humidity and cloud coverage, you can forecast exactly what coffee production will be. But there is a crucial condition: you need to predict these environmental variables simultaneously and accurately on a region-by-region basis.
Economic Value of Environmental Forecasting
Commodity futures contracts for Arabica are popularly traded on ICE. They have five series running in March, May, July, September, December. To illustrate the potential financial rewards of accurate environmental forecasting, let's examine the example of the May 2011 coffee futures contract for Colombia.
In Dec 2010 KCK11 was trading at $2.00/lb. The initial margin was $6,600, the maintenance margin was $6,000, and the contract size was 37,500 lbs. Suppose one accurately predicted the spot price at expiry, one could have closed the contract at $2.87 in mid-March when liquidity was still high and made $0.87 * 37500 = $32,625. For an input of $6,600 this represents a 494% return. Not bad for 3.5 months, is it? Of course, the leverage of 11.36x works both ways - if you are wrong, then you could easily lose your entire investment. This is why forecasting accuracy is so important.
Unlocking the Power of Precision Environmental Forecasting for Commodity Traders
Planette can forecast precipitation, sunshine (solar irradiance), temperature, and other environmental variables with high accuracy. Our predictions combine the latest physics-based Earth System Models (ESMs) with deep neural networks to accurately forecast 1 month to 1 year in advance, at spatial resolutions of 25 km across the globe. The following figure shows how well our models predicted precipitation several months in advance in Colombia.
Planette forecasts can be accessed using API calls, in an interactive dashboard, or alternatively, we can supply the forecasts on a regular basis as charts, reports and bulk data for the exact areas in the world for which they are needed. By leveraging Planette's advanced environmental forecasting capabilities, commodity traders can gain a crucial edge in predicting supply disruptions and capitalizing on the resulting price movements in agricultural futures markets.