Lettuce (Lactuca sativa L.) yield prediction under water stress using artificial neural network (ANN) model and vegetation indices

Abstract

Water stress is one of the most important growth limiting factors in crop production around the world. Water in plants is required to permit vital processes such as nutrient uptake, photosynthesis, and respiration. There are several methods to evaluate the effect of water stress on plants. A promising and commonly practiced method over the years for stress detection is to use information provided by remote sensing. The adaptation of remote sensing and other non-destructive techniques could allow for early and spatial stress detection in vegetables. Early stress detection is essential to apply management practices and to maximize optimal yield for precision farming. Therefore, this study was conducted to 1) determine the effect of water stress on lettuce (Lactuca sativa L.) grown under different watering regime and 2) explore the performance of the artificial neural network (ANN) technique to estimate the lettuce yield using spectral vegetation indices. Normalized difference vegetation index (NDVI), green NDVI, red NDVI, simple ratio (SR), chlorophyll green (CLg), and chlorophyll red edge (CLr) indices were used. The study was carried out in vitro conditions at three irrigation levels with four replicates and repeated tree times. The different irrigation levels applied to the pots were 33, 66 and 100 % (control) of pot water capacity. Spectral measurements were made by a hand-held spectroradiometer after the irrigation. Decrease in irrigation water resulted in reduction in plant height, plant diameter, number of leaves per plant, and yield. Using all indices in a feed-forward, back-propagated ANNs model provided the best prediction with R2 values of 0.86, 0.75, and 0.92 for 100, 66, and 33 % water treatments, respectively. The overall results indicated that spectral data and ANNs have high potential to predict the lettuce yield exposed to water deficiency.