Visualizing and Interpreting Large Sensor Datasets for Daily Specialty Crop Management Decisions

With increasing use of wireless networks for gathering and reporting sensor-derived information, many needs arise for end-users to accurately interpret and use the data. As part of an SCRI-funded national project, we are implementing networks with specialty crop ornamental growers to provide them with daily management information for irrigation and nutrient management from soil moisture data, together with many other decisions that can be aided by a suite of microclimatic sensors and derived data (such as degree-days and vapor pressure deficit). This translation of information into knowledge is not trivial, since growers are time-limited and wish to make better decisions, but within a relatively short (10-15 min.) time frame. This requires that we develop software decision support tools that can handle complex computational tasks, but deliver information in intuitive ways—for example using graphical user interfaces. We are using a commercially-available software package that uses a mySQL database to download and organize large volumes of data. We are also developing a more advanced graphic user interface in collaboration with the Carnegie Mellon Robotics Institute that uses sqlite3 for the database (since this tool is serverless), making transfer of data files very easy between users. This presentation will focus on how we handle and visualize large datasets to provide growers with precision information from specific networks. This will be illustrated by using specific case-study examples of how we use simple spatial and temporal data trend analysis to give growers an insight into what is normal biological variability, and what is an anomalous reading from a bad sensor, or from incorrect calibration.