Climate is quickly changing with unprecedented consequences for ecosystem functions and services. Heat and erratic rainfalls are the main factors challenging plant production, with an increasing frequency and intensity of drought stress that adversely effects several processes of crop growth, physiology and metabolism. Even the temperate regions of Europa, including Austria, are affected by more frequent drought periods and heat waves, with decreasing trends in yield for several key agricultural crops. Thus, coping with the uncertainty in water supply is essential to better adapt agriculture to climate change.

The fundament of a climate change adapted plant management is an improved understanding of the physiological defense mechanisms used by plants to avoid damages during stress periods. This requires measurement methods to effectively monitor and early predict the specific plant stress responses, which provide the relevant information to translate plant signals into management decisions. 

Most of the currently available physiological techniques are invasive, time consuming, mostly limited to measure single plant and often hardly applicable under field conditions. Experience from remote sensing has shown that spectroscopy can provide a tool to infer on the plant physiological status via spectral features encoding plant stress signals. Particularly spectral imaging is a promising approach for plant stress monitoring: it combines the potential of spectroscopic measurement of biochemical properties with spatial imaging. Originating in remote sensing of plant canopies via satellite images, spectral imaging has now advanced towards high resolution proximal sensing in the framework of plant phenotyping.

The aim of our project is to advance imaging approaches towards monitoring and prediction of physiological functioning of plants. Capturing the fundamental stress response processes of plants undergoing water shortage via imaging would allow an innovative early detection method for mitigating crop damage. In this project, we focus on grapevine and develop a novel setup that combines different imaging sensors (hyperspectral, thermal, RGB-D) with continuous direct physiological point measurements. Based on the multimodal dataset of directly measured grapevine physiological time series combined with spatial imaging, we develop learning algorithms that allow identification and early prediction of grapevine physiological stress response cascade during drying.

It is expected that this physiological imaging approach can provide improved solutions for no-invasive plant stress monitoring and support management decision via early prediction of the onset of drought stress in grapevine. 


Figure 1. Image time series of Zweigelt during a drying experiment. The colored dots show a pixel that is tracked over time for extraction of the hyperspectral information. 

Figure 2 Spectral lines of a grapevine leaf (colors correspond to the pixels marked in Figure 5) during a drying process. Red arrows indicate spectral regions that visually qualify for characterizing the drying process. Pink area marks the NDVI-region.