For example, consider the “precision agriculture” use case where farmers are
leveraging sensor data to maximize crop yields — and by extension, their prof-
its. To do so, they need to better understand what and when to plant, when to
water, when to fertilize and when to harvest. To make this possible, they are
deploying sensors to measure temperature, moisture and nitrogen levels in the
soil. These sensors may collect data continuously, or they may sleep and wake
up hourly. They may communicate data as it is collected, or, for purposes of
managing data costs and/or battery life, the data may be batched and sent,
store-and-forward style, on a daily or weekly basis.
These sensors need to be strategically deployed around fields in order to pro-
vide statistically useful data. While the quantity of data collected is quite small
per sensor, when analyzed at the aggregate level, it can become quite large, and
very powerful. The greater the density of sensors, the more granular the pre-
dictions and recommendations of the analytical models. Each reading in this
case is simply a set of data points that needs to be fed into machine learning
and analytical models to make determinations. Those models are frequently
FALL 2018 | THE DOPPLER | 45