Big Data & Precision Agriculture

Big Data & Precision Agriculture

GreenSeeker Sensor System


There is new term creating a lot of excitement in technological and scientific circles. It is “big data.” Big data refers to the generation of enormous amounts of data due to new technologies for measurement, collection and storage. Data are accumulated in such vast quantities that they defy conventional analysis techniques. As we will see, big data offers great opportunities but also major challenges. Before discussing how big data will impact precision agriculture, it may be instructive to learn how it has impacted another field which literally redefines itself through continuous observations.

The field of astronomy during the past two decades has undergone a rapid improvement in the ability of telescopes to collect data. If we assume that all data can be defined in terms binary digits or “bits,” then we can calculate how much data are in an image, book or table. This is because we can construct bytes from bits which numerically define colors, letters and numbers. The bigger a telescope is for making observations, the more data are collected in the same interval of time. The greater the density of picture elements or pixels in the same viewing area of a telescope, the more data are collected in the same interval of time. The faster the recording of images in a telescope of the same size and pixel density, the more data are collected in the same interval of time. With each new telescope launched into space or constructed on the ground, there has been acceleration in data realization.

The improvement in telescopes along with their “big data” capabilities is best illustrated through examples. The popular, exoplanet-hunting Kepler telescope, launched by the U.S. National Aeronautics and Space Administration (NASA), has a density of 95 million pixels with the capability to sum imaged data over 30-minute intervals. The Kepler telescope can precisely measure the light emitted from over 100,000 stars. This periodic tracking of stars has increased the number of known exoplanets from in the tens to in the thousands. The Atacama Large Millimeter Array (ALMA) is an international collaboration of ground-based, radio telescopes to observe galaxies, stars and planet formation. It began generating 40 terabytes of data per day when it became fully operational this year. The Large Synoptic Survey Telescope (LSST), a public-private partnership, has a 3-billion pixel density. It will be able to image an area of the sky that is 49 times the size of the Moon when it becomes operational in 2015. It has been estimated that there is now a doubling every year in the amount of data being collected in the field of astronomy.

The exponential growth of data requires new thinking on how to look at data. Today, it is humanly impossible to look at all but a tiny portion of astronomical data using conventional analysis. Consequently, in place of humans, computer programs conduct comparative analysis on large stores of data. Through a step-by-step procedure of calculations and/or comparisons, programs in the form of algorithms reveal patterns in large data sets. For example, in the case of the Kepler telescope, the tell-tale variation in light from a star reveals the presence of an orbiting exoplanet. A computer program can identify this variation with an algorithm and conduct an analysis of a star’s extended time series in the blink of an eye. While new telescopes are generating “astronomical” amounts of data, new algorithms are being written to analyze these data.

New Ways Of Thinking Needed

While agriculture is a latecomer to the big data phenomenon, it will soon follow a similar path as astronomy. Precision agriculture, with its time-dependent, geospatial data at the field scale, will be front and center in the realization of big data in agriculture. As sensors are placed on machinery, arrayed in soils and canopies and flown remotely overhead, the amount of data collected in the field will increase geometrically. As was the case in astronomy, it will be become humanly impossible to analyze data collected in an agricultural setting. And like in the field of astronomy, computer programs will be relied upon to conduct nearly all the analyses of large data sets.

Big data by its sheer volume of information offers a number of opportunities for precision agriculture. At the most rudimentary level, ingested data can be analyzed in real-time to flag critical values important for production decision-making. At a more sophisticated level, high resolution spatial maps of soil moisture can direct the efficient use of irrigation. Similarly, detailed maps of pest damage can allow for the precise targeting of controls in a field. At the most advanced level, remote-sensed data coupled with measurements made with sensors on machines or arrayed on the ground can be processed to create a dynamic, three-dimensional picture of soil, plant and environmental properties in a field. This picture would be composed of many layers of data, which singly or together can support specific management decisions.

The big data opportunities can be overshadowed by their challenges, especially in the early going. First, there are very few “data” scientists or persons who know how to create and execute the algorithms necessary for analyzing large of amounts of data. Second, there is commonly a mismatch in the scale, precision and accuracy of data coming from different sources. This mismatch can create an erroneous picture of what is actually happening in a field.

Third, big data, like all data, needs to be quality controlled before it is used in algorithms. The necessary quality control procedures can become pretty elaborate and time-consuming. The fourth and most important challenge is the interpretation of products created by algorithms processing large data sets. Interpretation of data patterns is very subjective. Every individual has their own way of looking at the world according to their beliefs, prejudices and preconceived notions of acceptable outcomes. Consequently, no two individuals will reach exactly the same decisions after interpreting big data patterns.

We can conclude that big data will increasingly become part of precision agriculture and will heavily influence our production decision-making in the not-too-distant future. We can also conclude that there will be a learning curve on the part of agricultural stakeholders making decisions based on big data.

The incorporation of big data in decisions, while challenging, may be one of the things precision agriculture needs to do to get back on a growth curve.