Identifying Water Stress in Potatoes Using Leaf Reflectance as an Indicator of Soil Water Content

Abstract

Soil water content must be monitored and maintained at adequate level for optimal productivity. Accuracy of traditional sensors used to monitor soil water content depends on the installation technique and proper contact between soil and sensor, which is difficult to achieve in light textured sandy soils. Non-contact sensing technique does not have the limitation of contact with soil and can monitor plant status continuously. In this study, hyperspectral imaging was used as a non-contact technique for detecting changes in spectral reflectance of Umatilla Russet potato plants grown under varying soil water content. An experiment was carried out in a greenhouse to subject potato plants at different levels of soil water content from extreme stress to surplus. Yield data was also collected, which showed that maximum yield for Umatilla Russet potato can be achieved at 18% to 21% soil moisture content. Various spectral indices were calculated using spectral reflectance data at different water stress levels. Principal component analysis was used to identify indices that represented maximum variability in the data. Simple Ratio Index and Modified Red Edge Simple Ratio Index were identified as the two most relevant indices for differentiating soil water content. K-Means clustering with these two indices resulted in an accuracy of 75% in identifying highly stressed plants and 92% accuracy in identifying stressed plants (that included both high and low stress levels). These results showed a promise for development of a non-contact sensor for detecting plant water stress in potatoes, which may lead to an automated irrigation system for maintaining optimal soil water content during potato growing season.