Next Generation Agronomic Weather for 'Soft' Commodities
Contributed by John Corbett and Michael Ferrari, aWhere, Inc.
A Look into the Current Soybean Production (South America) and the Market
In a market where the only constant in recent years has been volatility, understanding risk factors can make all the difference. This article focuses specifically around how the Risk aWhere™ platform can be used to understand factors affecting the oilseed market.
aWhere™ provides data, analytics, and risk management tools for all agricultural and soft commodities as we track and monitor all production areas across the globe. The foundation of aWhere’s platform is a next generation, high resolution agronomic weather database. We have integrated more than 10,000 ground stations with data from multiple satellite constellations (including exclusive access to global, hourly precipitation) to deliver daily weather observations every ~9km across the agricultural earth. Leveraging tremendous skill in 3 dimensional curvilinear functions for temperature and humidity, this data foundation brings unparalleled fidelity in spatial coherency and temporal acuteness to monitoring agricultural commodities.
aWhere’s global, daily, agronomic weather database has a representative agricultural meteorological (ag-met) dataset at a regular spacing of 5 arc-min or about 9km x 9km. On a daily basis at each location we add to each location 8 variables: minimum and maximum temperature, minimum and maximum relative humidity, precipitation, solar, wind speed, and wind direction. This database is kept current (up to yesterday]. For each location we also provide current conditions and hourly forecasts out 8 to 15 days.
Many weather sources rely almost exclusively on forecasts and forecast ensembles, we couple actual observations with agronomic models (specific to each crop) to fully understand crop health and the impact on specific growth stages. We include 8-15 day forecast data and such data are fully integrated into the modeling but it is our tracking of the actual agricultural weather that differentiates of our next generation platform from existing systems.
Perhaps the most significant improvement in soft commodity information for those with risk exposure is the Platform’s number of observations. For example, the soybean production area of Argentina (as defined by IFPRI’s Spatial Allocation Model – see maps below and Harvest Choice citation at end) contains 13 WMO weather stations. For the same area, the aWhere risk management platform has 6,151 independent observations of rainfall and for each of these locations, a complete agricultural meteorological station’s worth of daily data (i.e. the accurately interpolated temperature and humidity data). Understanding production and therefore market movement takes on an entirely new opportunity when your assessment includes this agronomically optimized weather asset.
Our analytics and crop stress measures identify risks to crop development and subsequent price reactions ahead of conventional sources. The Risk aWhere data and tools enable us to track South American growing conditions at the field level, specific to the crop/origin of interest without being subject to (a) developing analogs that are tied to a specific weather station which may be 10s or even 100s of kilometers away, and (b) errors associated with traditional satellite and survey based crop monitoring (i.e. cloud cover, infrequent revisit times, etc.). Further, where crop indices from other services may have a representative sample that amounts to 10-20 locations per origin, we have thousands.
After the United States, the #2 and #3 soybean producers are Brazil and Argentina, respectively. The first set of maps describe some information important to the current and forthcoming soybean crop in these two origins. The first map depicts the agronomic weather for the primary soybean origins in Argentina between 19 Nov and 18 Dec vs. the 2006-2014 ‘normal’ (average). The map shows the outline of the soybean production area of Argentina, and the 2015 rainfall during the early vegetative stage as compared to the long-term normal. The histogram shows the distribution of 2015 rainfall across the whole of the Argentina soybean region compared to the long-term normal. About 25% of the region had lower than normal rains though only 4% of that was significantly lower during this 30 day period. Some 6,151 independent rainfall observations describe the conditions and for all observation locations, the complete daily data are available for more detailed assessment.
For the same period in Brazil, the same map and histogram present slightly more area with lower than normal rains for the 30 day period November 19-December 18. Here the growth stage is at the start of the flowering stage in soybean development (planting in Brazil is typically a month earlier than in Argentina). Zones A, B, C, D contain 4,210 complete daily ag-meteorological data sets. Zones E and F total another 9,429 independent rainfall observations (thus the total Brazil soybean area as depicted below totals 13,639 ag-met data sets).
Risk aWhere’s daily foundation enables examination of conditions in any geography (a production area, sales territory, political administrative unit) for any time period and any ag-met index. For example, the precipitation difference from LTN for the last 4 weeks (period Dec 24, 2015 – January 22, 2016) shows a reversal in the wet/dry by geography.
Examination of rainfall during various growth stages and since planting provides insight but with all ag-met variables available for each location, other indices from full-on crop simulation models to calculating daily PET (Potential Evapotranspiration) are possible. PET is a useful variable as it describes the drying conditions of the atmosphere. Just as a cloudy humid day would be less impactful to already water limited situation, a hot, clear, windy day, typical of drought conditions, impacts field crops far more. The ratio of P/PET is a useful index for quickly describing the plant-available water conditions.
With El Niño hitting its maximum in January 2016, the situation in Brazil is striking. This next pair (Brazil map and histogram) shows the P/PET percent difference between this year’s P/PET ratio since planting through January 13, 2016 and the 2006-2014 average P/PET for the same period. The northern part of Brazil’s soybean zone is showing signs of much drier than normal conditions.
The histogram of the same data shows that 31% of the soybean area in Brazil exhibits conditions 15% drier than normal. At the other extreme, heavy rains indicate that 17% of the area is >90% wetter than normal.
The key to agricultural production and the Risk aWhere quantitative differentiation is the ability to monitor large areas - and individual fields - day by day. Examine the map and histogram below: the recent rains (the 8 days between 01/13 and 01/22) in northern part of the Brazil soybean area does start to change the signal as to overall seasonal crop water conditions. These data show the period from October 15, 2015 through to January 22, 2016 (adding the last 8 days as compared to the previous map/histogram). Now only 25% of the soybean production area in Brazil is more than 15% drier than normal and only 10% are significantly wetter.
For Argentina, the period from November 15, 2015 through to January 22, 2016 – a period that started out wetter than normal, continues to look quite good in terms of crop water availability with only 9% of the area drier for P/PET than the 2006 – 2014 normal.
Putting the data to work
As noted, we believe that the interpretable quantification of these detailed, spatial extensive daily weather data are just beginning. At this stage, there is already more useful, and more granular, quantified information here than what is being utilized by most market sources in the commodity risk management space (i.e., thousands of independent rainfall observations vs. 10s). As we dig deeper and assess crop physiological activity at the field level, we add more precise planting dates by location. We model growth stage (i.e., growing degree days) by temperature and we asses stress specific to each growth stage. Soil data can be included as well as pest and disease information specific to each location, can be modeled, observed (ground truth) and added to the overall assessment. Further, we add an approximation of the number of hectares planted so that each ag-met data set gets further weighted by this area planted.
Putting what is described above in the context of current market conditions for soybeans, there appears to be a generally favorable outlook regarding South American production. With March 16 soybean futures currently just coming off of a 12 month low, and a broad bearish sentiment across the commodity sector and favorable crop prospects, the prevailing theme is for more pressure on the oilseed sector. However, this view should be balanced with information contained in the USDA’s recent WASDE report. The 12 Jan 2016 report noted that the global oilseed 2015/16 production estimate is now at just under 527 mm tons, down 2 mm tons from the previous month’s estimate; of this the global soybean estimate is at 319 mm tons, which is down 1.1 mm tons from December, largely from smaller crops in North America. Further, global oilseed stocks are estimated in the WASDE at 90.9 mm tons, which is a 4.2 mm ton reduction. So while broader fundamentals paint a bearish picture, supply side constraints may prevent prices from sliding further from their current levels.
When examining the distributions generated for one period, it is essentially a snapshot. The real value in utilizing aWhere data is extracted by monitoring how this changes each period over the course of a growing season, from weeks to months. By doing this, risk managers and traders can track crop progress without relying only on surveys. Survey resources can be optimized with field assessments being requested from specific areas because known conditions drive the requirement. Further, risks as well as opportunities emerge from both a physical and financial perspective, ahead of the market. This allows forward hedges to be constructed with more information than the market is currently using, from a crop health and potential perspective. Again, what is described in this brief article is just a sample of the parameters where these types of analyses can be generated. Risk aWhere can do this and more for every global commercial crop in every origin.
HarvestChoice, 2014. "Crop Production: SPAM." International Food Policy Research Institute, Washington, DC., and University of Minnesota, St. Paul, MN. Available online at http://harvestchoice.org/node/9716.
John Corbett is a member of the speaking faculty for Oilseed Congress Europe MENA, February 9-10 in Barcelona Spain.
The opinions expressed in this editorial are the authors' own and do not reflect the views of Oilseed & Grain News.