Using RadialPheno to Make Sense of Phenocam-Collected Phenology Data

Remote sensor technologies such as cameras, weather stations, and GPS trackers have transformed biological data collection in the field.

Scientists can currently capture continuous datasets in various terrains, at a scale inconceivable before these technologies became available. But as this torrent of data has poured into laboratory computers globally, scientists have found themselves without well-built analytical tools to interpret and understand it all.

In a study presented in the latest issue of Applications in Plant Sciences, Dr. Greice Mariano and colleagues launch a tool referred to as RadialPheno to examine leafing patterns of plants based on remote camera data.

Before the extensive use of remote sensing equipment for field data collection, manual observation performed by researchers was a hard, monotonous, and error-filled task, producing much smaller and more uncertain datasets. This is perhaps mainly true of phenology—the study of the timing of developmental events such as leafing, flowering, and fruiting—because solid phenological observation necessitates being in the correct place at the correct time.

On-the-ground phenological observations are accomplished periodically, require much more labor, and rely on the people that perform the observations. On top of that, results are stored in spreadsheets, which do not allow interoperability between data, making it difficult to analyze the data.

Dr. Greice Mariano, Study Corresponding Author and Post-Doctoral Research Fellow, OCAD University

The ability to get detailed sets of observations by arranging relatively economical remote cameras has altered all that. But these observations will not matter much unless they can be interpreted into integrated, tractable datasets that can produce telling insights.

Since remote monitoring is something new to phenology studies, there is a need to develop and improve methods for detecting changes in phenological series and data images, setting up standards.

Dr. Greice Mariano, Study Corresponding Author and Post-Doctoral Research Fellow, OCAD University

She carried out this research during her PhD at the Institute of Computing, University of Campinas, Brazil. She added, "Thus, in this research we introduced RadialPheno as a tool to support phenology experts in their analyses."

To identify the requirements of these phenology professionals, Dr. Mariano and her team partnered with scientists at the Phenology Laboratory at São Paulo State University (UNESP). Dr. Mariano examined what they wanted to see from data visualization tools and the problems they encounter with present-day software.

Based on this study, they "developed a tool focused on the visualization of temporal data, where the identification of recurrent events is the most important task. We also thought about the integration of the visualization with the common statistical methods used by the experts," explains Dr. Mariano. "We choose radial representation because those structures are useful for understanding cyclical events, such as phenological ones."

The team tried out their new tool by measuring leafing patterns in the Brazilian cerrado, a massive savanna ecosystem with a predominantly wet and dry season.

Many plant phenology studies in Brazil have been conducted on a cerrado savanna area, and a digital camera for near-surface monitoring was installed in that area. Thus, we also have phenological data based on on-the-ground observations, which allowed us to validate our tool with both on-the-ground data and [camera-derived] data.

Dr. Greice Mariano, Study Corresponding Author and Post-Doctoral Research Fellow, OCAD University

The camera network they arranged is part of a set of cameras fitted in various areas called the Phenocam Network, which can be accessed online.

Solid data about phenological patterns are more immediately wanted than ever, as climate change changes the timing of events in ecosystems worldwide with probably dismal consequences for ecological interactions. Fortunately, remote sensing technologies have made collecting these datasets feasible; tools such as RadialPheno can currently render them more meaningful.