Weather variability has increased and will only become more extreme as the atmosphere warms. A general overview of why this current variability will continue to increase in the coming years can be found in this blog: Food Security and Weather Variability.
Coffee's Precarious Position (August 2017)
- Coffee stores in many areas are at historic lows
- Global demand for coffee is steady and growing
- Understanding each of the coffee tree’s growth stages is
key to accurately forecasting yields
- aWhere monitors all of the world’s coffee growing regions
with complete, highly localized coverage on a daily basis
- This complete global coverage allows us to forecast crop
stress, and ultimately, crop yields
The coffee plant, particularly Arabica (the more rich and tasty type), is vulnerable to decreased production due to many factors from too warm nighttime temperatures to variable rains and drought. Taking between 30-35 weeks from flowering to harvest-ready, the risk of stressed periods is much greater now than ever. Stressed trees are more susceptible to disease and insect damage. The last section of this document is a brief primer on growing coffee and the relationship between the coffee tree and the weather.
Current global coffee stocks are at historic lows. Country inventories appear healthy, but this inventory typically supplies less than 1/3 of a year’s demand. The chain that connects the bean in a farmer’s field to your morning cup is starting to look a bit like ‘just-in-time inventory’ with a very large caveat: The factories (the farms!) that produce the coffee bean are now subject to more yield impacting weather variability than ever before. The risk of a shortage is higher than anything in the historical record.
Complacency in managing coffee supply chains is unhealthy. To quote Nassim Taleb author of The Black Swan (2007), “Globalization creates interlocking fragility, while reducing volatility and giving the appearance of stability. In other words, it creates devastating Black Swans”
aWhere’s innovation is to construct a weather station’s worth of daily data every 9km across the agricultural earth. With this resource, we do not have a ‘sample’ of the weather across a production region, compared to the typical situation of a few irregularly spaced weather observations. We have the ‘population’. We miss no dry areas between existing on-the-ground stations. We see nighttime temperature everywhere. We know where the coffee is grown, and we have representative soils data for those locations. With these data as input to coffee stress and yield models, coffee production monitoring is taken to a new level of specificity and accuracy (see blog: about aWhere's big data asset).
Coffee production monitoring starts with knowing where the coffee plantations are, the characteristics of the soil, and multiple agronomic practices - the crop calendar (when is flowering?), type/variety, tree spacing, and access to irrigation. But no matter how detailed – or lacking in detail - these data, the over-riding key to production prediction is accurate localized weather data. Big Data vision and analytics helps explain an underlying key to this innovation: the law of large numbers. We know that rainfall varies tremendously across short distances, but with aWhere’s blending of many data sources and our modeling expertise, we are able to monitor across all production areas. From that modeling, the biological signal of coffee tree health is derived and interpreted. aWhere processes billions of data points each day, optimizing the information for agricultural use. We leverage big data science and our agricultural domain expertise to understand coffee production everywhere.
Today, traders and supply chain managers buy information on weather and receive general narrative on the impact of the weather across whole continents typically with general reference to crop growing regions. aWhere leverages the same forecasts, but we report the state of the weather with quantitative ag-interpretation on the crop in question. This is a daily updated robust, statistically and biologically valid information stream. aWhere’s quantitative expression of the same forecast coupled with biologically assessed observed ag-weather creates a next generation information stream.
If you are interested in quantitatively monitoring coffee production, please read on!
Monitoring global coffee production: a quantitative signal
The top 7 coffee producing countries account for just over 80% of the earth’s production with Brazil (35.7%) and Vietnam (16.6%) the only two countries with more than 10%. Colombia (9.4%), Indonesia (7.5%), Ethiopia (4.3%), Honduras (3.9%) and India (3.5%) round out the top 7 producers. Total global production was 153,869,000 bags in 2016/17 equating to ~9.2MMT (@60kg / bag = ~9.2 million metric tons). Clearly Brazil’s production defines the global supply as coffee is concentrated in a relatively small area.
With aWhere’s foundation ag-weather information asset, we track the ag weather over Brazil’s coffee production area, leveraging more than 4,900 of our detailed daily weather stations to cover the entirety of the coffee production zone. With these data, we correctly saw the lack of rain over Espirito Santo throughout early 2017. This was recently reported in June’s USDA coffee report (insufficient water to irrigate here at the start of the new production season), and we predicted the plants would produce smaller beans (also just reported) as the weather over much of Brazil’s coffee area continually left shallow roots dry for weeks at a time.
Purpose built for global agricultural monitoring, aWhere maintains this same resolution of a meteorological station’s worth of data every 9 km for all coffee producing areas. With this robust ag-weather foundation, we monitor all the coffee areas across the planet: from Costa Rica to Kenya, Thailand, Venezuela and Uganda. Some 27,250 weather station’s worth of data are processed each and every day – covering the planet’s coffee areas - to bring our clients actionable insight weighted by production hectares. Quantitative metrics with a timeline of key stress indicators coupled with where the indices are today as compared to the historical range (see Graphic 2) are all available and updated daily.
All our interpreted ag-weather, from soil moisture status to evapotranspiration deficit (what the tree needed but did not get), – and inclusive of the forecast – are specific to the impact on the coffee crop. Our quantitative metrics roll up growth stage sensitive crop health indices delivering statistics by the number of hectares impacted. We track soil moisture status via models so predictive functions leveraging historical weather are initiated with actual current soil moisture status. Our forecasts are applied to the locations where there are coffee trees. This quantitative interpretation of the forecast adds more insight than does a simple narrative description of the general area (i.e., ‘rain in central Brazil’ vs. ‘26% of the coffee area will receive less than 15mm, 74% more than 30mm’). aWhere has a scientific emphasis on accurate observed data, as with these data our models track critical coffee-sensitive attributes from the state of the soil moisture to nighttime minimum temperatures and daytime maximum (see the Growing Coffee primer below) all to more fully understand, explain, and predict yield and ultimately, production. Potential evapotranspiration – the water demand of the environment (see blog: How Thirsty is the Atmosphere) is calculated by utilizing all the meteorological variables and in this way aWhere understands the impact of wind, sunny skies, low humidity, and heat – exactly the conditions that impact productive coffee trees.
Global consumption is growing – recent announcements by Starbucks and Dunkin significantly expanding into China drives home this point. Even without this China expansion, coffee consumption has been growing at 1-2% per year and is rapidly overtaking production. Note that current ICO reports on consumption provides data for most countries but does not include numbers for China, Korea, nor South Africa (as examples).
Expansion of coffee hectares – the investment – has been hampered in recent years by low coffee prices, and new hectares are typically in more marginal areas as prime coffee lands are actually fairly limited (especially for Arabica). Indeed, coffee supply – the market – is in a precarious position.
Growing coffee - a coffee tree primer: Arabica focus
The physiology of the coffee tree – and here we focus on the Arabica plant – has a series of stages each with varying dependencies on ‘the weather.’ Flowering is triggered by rain or irrigation typically immediately after a dry season (commercial producers favor coordinating flowering for ease of management – in some areas, trees flower naturally over an extended time). Once pollinated (and Arabica is self-pollinating), the pin-head period (6-8 weeks) is followed by rapid growth through 15-16 weeks. This is a critical period and water stress is impactful and will permanently limit bean size if there are drought conditions weeks 7-16. During endosperm growth (weeks 16-27) the bean gains dry matter and here if the plant lacks water, the bean can become shriveled.. The final stage, cherry ripening, sees the color change from green to red about 30 to 35 weeks after flowering.
Arabica coffee is produced in relatively cool climates (optimum temps annually between 15-24C) as photosynthesis is slowed above 24C and cooler temperatures stress the plant and near 0C, frost will damage the tree. Quality is increased if the day and nighttime temperatures are large (diurnal gap) while high nighttime ‘low temperatures’ above 22-24C are actually detrimental to the plant. Arabica likes to cool down at night! Excessive rainfall favors proliferation of diseases while dry spells lead to tree dehydration that not only impacts bean size and quality but leaves the tree more susceptible to red spider mite, leaf miner, and the coffee berry borer.
Arabica roots are not especially deep – a main root to 80cm with a network of lateral roots concentrated in the top 20cm of the soil. This means that the Arabica tree does great as long as there is regular rainfall. Total rainfall is not nearly as important as the distribution – the daily supply.
In summary, the Arabica coffee tree prefers – actually require - stable cool temperatures and regular rainfall. For commercial crop management purposes, triggering flowering has many benefits as plant nutrition and harvest can be better coordinated. Highly variable temperatures, particularly extreme events, negatively impact production as does variable rainfall as even short dry spells have a negative impact. Commercial production can achieve amazing yields but yield is strongly correlated to the weather.
Graphic 1: example of tracking the daily forecast – 7 days - over the coffee area of Brazil. P/PET is explained in an earlier blog: How thirsty is the atmosphere? - which is designed to interpret daily rainfall (here accumulated over 7 days of forecast) with the water demand of the environment. Here we see that what little rain was in the forecast is essentially consumed by environmental demands – and the trees remained stressed.
Graphic 2: last 7 day observed. Here we are showing that the forecast shown in Graphic 1 was actually optimistic for the amount of water! Understanding the weather forecast accuracy is crucial.
Graphic 3: map of water stress, coffee, last 30 days. Now showing the crop condition (water stress) over only the coffee crop, these maps, also updated every day, provide insight into the past observed record. Quantitative charts and tables further illuminate interpretation.
Graphic 4: the timeline of aggregate stress, here for Brazil coffee production area, from 1 January through July, 2017. The current index, the black line, is shown alongside the historical pattern – the boundary red-yellow is one standard deviation of normal towards water stress while the boundary yellow-green is one standard deviation more moist. 2017 saw conditions all through the cherry fill stage that were drier than normal as averaged across the production area with an odd out-of-season more moisture in June. The small bean size currently reported reflects this slightly more stressed than normal growing season.
Graphic 5: an example of why the general moisture stress in 2017 was higher than normal when monthly rainfall totals were a bit low but not dramatically so (Brazil): Temperatures were much warmer than normal!
Graphic 6: drilling into some locations in Espirito Santo we see that there were long stretches of January with no or little rainfall coincident with maximum temperatures above normal.
Graphic 7: eastern Minas Gerais did catch some rain but note the higher than normal temperatures. Coffee trees become stressed with high temperatures and the evaporative demand of the environment increases (PET goes up with higher temperatures).