Coffee weather report November 10, 2017. awhere, Inc., an agricultural intelligence company, is pleased to provide this map-and-chart heavy report focused on the current coffee crop in Brazil. Global stocks of coffee have declined over the last few years, though there is conflicting news on this status. Nearly all stock and inventory information reports, however, add two caveats: 1) global supply is expected to stabilize should the current Brazil crop achieve its expected record production of about 60 million 60kg bags and 2) the only way this current crop will achieve this record is if there are no significant weather impacts. In this report, we will leverage awhere s ag-intelligence to provide insight on Brazil s current coffee crop: * biological ag-information / ag-weather * current conditions compared to the last 10 year average * comparison of forecast to what actually happened (Very Important) awhere monitors the agricultural earth in great detail. This report highlights some of our insight. For more information please contact beawhere@awhere.com. All information, data, maps, charts etc. are available from awhere. For additional background on awhere s tracking of coffee production - and the supply situation planet wide: http://blog.awhere.com/coffees-precarious-position-aug-2017 For information on awhere s global, localized, real-time, quantified ag-weather asset: http://blog.awhere.com/foundation-big-data-asset Our foundation ag-information database contains a weather s stations worth of daily data every 5 arc-min or ~9k x 9km across the planet. This platform is incredibly agile. Not only can we compare current conditions to past averages or individual years, but we can also assemble aggregations by any spatial area: administrative (a country, state, district); natural (watershed, crop production area); business (sales territory or area of interest) -- or by any time period: a day, week, month or since a key biological signal (planting, flowering, or specific growth stage). Fundamental to our platform is the ability to model a wide spectrum of agronomic insights from crop specific and targeting yield and production to environmental trends, crop suitability (where could a particular crop or even variety grow?) or even site suitability (what crops could grow in a specific locations). 1
Where coffee is grown Map 2 Coffee production areas across South America Map 1 Where coffee is grown weighted by hectares of production Map 3 These maps highlight where coffee is grown in South America (Map 1) and then in main coffee producing regions of Brazil (Maps 2 & 3) Coffee acres have expanded in the past decade and these data represent a mashing of multiple sources. The remainder of this report presents coffee production ag-weather information focused on the main production zone as show in Maps 2 and 3. 2
Rainfall with coffee areas October 26 to November 1, 2017 Rainfall mm total Note the very high variability - spatial and weekly total (time) - of rainfall. Moreover, notice the large areas of yellows or less than 60 mm. This will start to become more important as soil moisture recharge is dependent upon surplus moisture (more than PET). Map 4 Here we present observed rainfall between October 26 and November 1, 2017 with the general coffee area overlayed to train the eye on the key areas to examine. Yes, we could present the maps showing only the ag-weather for the areas where coffee (the crop under review here awhere monitors all major crops across the planet Brazil sugarcane area example at the end of this report) but we have found that most people prefer to see the continuous surface of the ag-weather information. 3
Rainfall - October 26 to November 1, 2017 Rainfall mm total Note the very high variability - spatial and weekly total (time) - of rainfall. Map 5 For this view, we have removed the coffee area for a more clear view of the highly variable rainfall accumulations. The next map (map 6) shows the observed rainfall for the period 2-8 November 4
Rainfall - November 2 to November 8, 2017 Rainfall mm total Minas Gerais did not appear to catch much rainfall this past week. Note the large areas of yellows or less than 60 mm. Map 6 Examine closely the 14 day period as shown on maps 5 and 6 rainfall across much of Minas Gerais (the Brazilian state that produces the most coffee) was pretty light in some key production areas. 5
P/PET P/PET - November 2 to November 8, 2017 Minas Gerais and Espirito Santo pretty dry this past week Map 7 P/PET an expression that relates rainfall to the evaporative demand of the environment (see http://blog.awhere.com/how-thirsty-is-the-atmosphere for more information on PET)., The atmosphere is warmer today and it is simply not enough to track only rainfall. PET is a calculated measure, in mm, of the evaporative demand of the environment. The ratio P/PET provides a useful index to express whether or not sufficient rainfall fell to add to the soil moisture. In Map 7, we see that much of Minas Gerais and Espirito Santo were quite dry this past week there is irrigation in Espirito Santo (though reservoirs are already low) and much less irrigation available in Minas Gerais. Soil water holding capacity is a physical expression of the buffer that keeps plants healthy between episodic rainfall. For now we will leverage P/PET fairly extensively but know that crop simulation models, which include ways to more realistically manage the ag-weather (for example, very large rainfall events have run-off which does not increase soil moisture and then a simple P/PET for that period can be a little misleading) are part of awhere s advanced modeling portfolio. 6
30 day P/PET (9 Oct 8 Nov) 30 day rainfall total (9 Oct 8 Nov) Rainfall mm total P/PET Map 8 Map 9 One week of rainfall or low P/PET may not impact plant health / crop growth but over 30 days the signal starts to have yield impacting interpretation. One reason for this report of 10 November was that there was a good rainfall in late September and early October and this rainfall triggered a spectacular flowering of coffee trees over much of this area of Brazil. There were some exceptions: part of northern Parana state (the southern most coffee production state in Brazil) had an unusual flowering rainfall event between 15-20 August 2017 and there is some suspicion that the high temps and very dry conditions between that August rainfall event and late September may have led to coffee berry abortion this still to be quantitatively confirmed. Not much rain fell in early to mid-october either (Maps above) though temperatures were warm but more seasonal. The 30 day period here then examines that period from just after the flowering event through 8 November. As can be seen, conditions across much of Minas Gerais were actually pretty dry and this is not what the many rainfall forecasts of late October and early November were predicting. 7
Rainfall Forecast on 2 November for 2-8 November Rainfall Observed for 2-8 November Rainfall mm total Map 10a Map 10b As an example of this optimistic forecast, compare maps 10a and 10b. Clearly the rainfall forecast (GFS) of 2 November for the next 7 days significantly overestimated the amount of rain that would fall. The extreme localized larger rainfall amounts continue to impact people and agriculture but the dryness is also starting to impact healthy plant growth. The next set of maps leverage and more simplified visual and they provide a quantitative assessment of the forecast over the coffee production hectares. 8
In mid-october rainfall entered the forecast (Map 11a) and 8 days later, looking back over that specific week, we see that the forecast actually underestimated the rainfall (Map 11b) Map 11a The next slides will present map pairs of forecast and observed for rainfall (Precipitation) and P/PET. The map shows south and eastern Brazil. The black dashed lines schematically highlight the coffee production zone. Map 11b Note the histograms: these graphics provide a quantitative breakout of the variable (rainfall or P/PET) over the coffee production hectares only as shown in Maps 2 & 3 for Brazil. An explanation of the more generalized maps and charts awhere generates daily. The next few slides leverage this information style. 9
Forecast +7 days Forecast vs Actuals Map set 11 10/18/2017 10/24/2017 Map 11a Map 11c Observed last -7 days Map 11b Map 11d For this mid-october forecast, the rainfall was more than expected over the coffee areas temperatures were also higher than normal and PET was high. The resultant P/PET observed (Map 11d) showed fewer areas of the lowest ratio (less than 0.40 forecast was 74% of the area (Map 11c), observed was only 54%) but the next category, P/PET 0.40 to 0.80 captured the bulk of the difference between the forecast (only 17% of the coffee area) and, 8 days later, the observed (33% of the area). 10
Forecast +7 days Forecast vs Actuals Map set 12 10/25/2017 10/31/2017 Map 12a Map 12c Observed last -7 days Map 12b Map 12d For this period, the forecast again underestimated the amount of rainfall but for this week, there was sufficient rain to start to recharge soils in some areas. The P/PET forecast was pretty close forecasting 50% of the area to have a ratio > 1.4 (Map 12c) and, 8 days later, we observed that 59% of the area exceeded 1.4 (Map 12d). 11
Forecast +7 days Forecast vs Actuals Map set 13 11/01/2017 11/07/2017 Map 13a Map 13c Observed last -7 days Map 13b Map 13d Map set 13 shows where things moved into a new space. The forecast of 1 November (Maps 13a and 13c) greatly overestimated the rainfall over the coffee areas. The observed data (Maps 13b and 13d) paint a very different picture: what actually emerged was the water stress conditions also shown in Map 6 and 7. Why the forecast missed no idea. But the reality is that we need to monitor closely this coffee crop as the ag-weather conditions are starting to wobble away from normal and should this persist, then the record production (60+ million bags) may get further compromised. If this crop cannot confidently re-fill dwindling global supply, there may be some very acute price increase response. 12
In conclusion, closely monitoring Brazil s coffee areas is needed. Observed information will explain the actual yields and production. The forecast is clearly valuable and we will continue to review the accuracy of various forecasts over the coffee production areas. For the record production to occur, there can be no negative bumps in the conditions. Already there is great variability in the ag-weather and that record production is likely already compromised. 13
P/PET September 29 October 5, 2017 PPET September 29 October 5, 2017 Coffee production area Map A Map B Which view of the ag-weather do you prefer? Map A where the information is provided in a comprehensive view, or Map B where the presentation is for the selected crop area. Let us know beawhere@awhere.com 14
Precip total September 29 October 5, 2017 Sugar Cane production area Precip total September 29 October 5, 2017 Coffee production area awhere s agricultural intelligence covers all crops 15