Field-Based High-Throughput Phenotyping Using Newly Developed Proximal Sensor Device

To cope with changing climate and increasing human population, agricultural production requires more data on plant status. High-throughput phenotyping represents new avenue of data collection, increasing throughput in both time and space. We developed a low-cost proximal sensing node retrieving reads at six wavelengths in red (610 and 680 nm), near infra-red (730 and 760 nm) and infrared (810 and 860 nm), intended to capture important plant reflectance indices and fluorescence signals. The reads were collected in fields of maize, barley and wheat and coupled with reads from a commercial multispectral sensing device and a handheld fluorimeter, measuring plant photosynthetic efficiency in a JIP-test framework. Reads of our sensing node were standardized by calculating vegetation indexes (VI). Strong correlations were observed between VIs retrieved from our sensor and NDVI measured by a commercial device, reaching
$\mathrm{R}^{2} > 0.91$
. Correlation analysis also showed strong link between VI810610 and primary electron transport, vanishing between VIs and higher order photosynthetic reactions. Collection of meaningful ecophysiological data facilitates plant state retrieval in scales from seconds to months (temporal throughput) with increasing spatial density (spatial throughput). Proximal sensing nodes combined with novel communication devices (Internet of Things), emerging phenotyping technologies and methods (high-throughput phenotyping) and data science frameworks (machine learning) show promise in transition to Agriculture 4.0/5.0.