Automated leaf temperature monitoring of glasshouse tomato plants by using a leaf energy balance model Kristof Vermeulen a , Jean-Marie Aerts b , Jan Dekock b , Peter Bleyaert c , Daniël Berckmans b , Kathy Steppe a,⇑ a Laboratory of Plant Ecology, Ghent University, Coupure links 653, B-9000 Ghent, Belgium b M3-BIORES, Katholieke Universiteit Leuven, Kasteelpark Arenberg 30, B-3001 Leuven, Belgium c Department Protected Cultivation, Inagro, Ieperseweg 87, B-8800 Beitem, Belgium article info Article history: Received 23 November 2011 Received in revised form 12 March 2012 Accepted 3 May 2012 Keywords: Crop monitoring Speaking plant Early-warning Mechanistic modelling Plant-based stress detection abstract In order to detect biotic and abiotic stress at leaf level thermal indices based on leaf temperature mea- surements have been commonly used. The application of these indices within glasshouse crops is, how- ever, restricted due to the specific humid conditions and the large spatial variability of irradiance and air temperature inside a glasshouse. In this study, a novel diagnostic algorithm is proposed as an alternative method to automatically monitor the leaf temperature of a glasshouse tomato crop based on the eco- physiological interactions between a leaf and its surrounding microclimate. Given that this algorithm is intended to be implemented as a software tool in glasshouse climate control systems, a critical over- view of all relevant equations found in literature was first given. Next, the most appropriate equations were selected by using two objective criteria, i.e. the commonly used R 2 and the less conventional Young Information Criterion, which also takes into account the complexity of an algorithm, so that the most fea- sible algorithm for automated monitoring purposes was built. Our results also showed that an in situ cal- ibration of the selected algorithm was needed, for which a novel procedure was proposed. Once calibrated, this algorithm successfully simulated the leaf temperature of a well-watered tomato plant during several days given that the environmental conditions in its microclimate were accurately mea- sured. Finally, the 95% confidence limits on the leaf temperature simulations provided the requested dynamic thresholds necessary for an effective automated monitoring tool. It was demonstrated that by using this novel diagnostic algorithm unexpected and likely harmful stomatal closure can be detected before visual signs of turgidity loss are observed. Ó 2012 Elsevier B.V. All rights reserved. 1. Introduction For automated monitoring of glasshouse crops, leaf tempera- ture is considered by many researchers as one of the most promis- ing and valuable plant responses (Baille, 1992; Hashimoto et al., 1981; Ehret et al., 2001; Kaukoranta et al., 2005). It is generally known that leaf temperature affects the rate of key ecophysiolog- ical processes, such as photosynthesis and transpiration and that extreme temperatures can result in irreversible damage (Taiz and Zeiger, 2006). Nevertheless, Kaukoranta et al. (2005) pointed out that leaf temperature measurements of glasshouse crops are mainly interesting because of the early-warning signal they can provide resulting from an unexpected stomatal closure due to plant diseases or due to unfavourable growth conditions, such as a too low moisture content in the root zone environment. Handheld (temperature guns) or stationary infrared thermome- ters are already in use in agricultural research to remotely measure surface temperatures for crop status monitoring (O’Shaughnessy et al., 2011). However, applications for monitoring in glasshouses are scarce. For field crops under clear sky conditions, automated early-warning systems based on leaf temperature have already been developed using thermal indices (Jackson et al., 1981; Takala, 1996; Jones, 1999). These indices are often based on the tempera- ture differences between a real leaf on the one hand and artificial ‘wet’ or ‘dry’ reference surfaces on the other hand, which might be regarded as leaves with minimal and maximal stomatal open- ing, respectively. Although these thermal indices have been widely used, De Lorenzi et al. (1993), Jones (1999) and Blonquist et al. (2009) emphasised that this approach has severe limitations under humid or overcast conditions resulting in unreliable indices, since the temperature differences between a leaf and the reference surfaces are then quite small. These climatic conditions, however, prevail 0168-1699/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.compag.2012.05.003 ⇑ Corresponding author. Tel.: +32 9 264 61 12; fax: +32 9 224 44 10. E-mail address: kathy.steppe@UGent.be (K. Steppe). Computers and Electronics in Agriculture 87 (2012) 19–31 Contents lists available at SciVerse ScienceDirect Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag