Modelling post-emergent hemp phenology (Cannabis sativa L.):: Theory and evaluation

Temperature and photoperiod can be used to simulate post-emergent hemp (Cannabis sativa L.) phenology. With reference to hemp in Italy, our main objective was to model field crops grown under a range of temperature and day length regimes. Dates of emergence and 50% of flowering were collected at Cadriano (Bologna) from serially sown field experiments (1996-1999, 2003-2005) on five cultivars: Carmagnola (late maturity), Felina 34 (medium maturity), Fibranova (late maturity), Futura (medium-late maturity), and Tiborszallasi (medium-late maturity). The database of phenological records was segregated into calibration and validation subsets. A phenology model was developed which utilises the beta function for response to hourly air temperature, and a switch-off function for response to day length. The life cycle of hemp from emergence to 50% of flowering was defined in terms of physiological development days (chronological days at the optimum photoperiod and temperature) and considered in three phases: juvenile phase (BVP), photo-sensitive phase (PIP), flower development phase (FDP). Critical temperatures (T-b, base; To, optimum, T,, ceiling), which did not vary widely across phases and cultivars, were estimated as common values: T-b = 1.9 degrees C for BVP and 11.3 degrees C for the other phases, T-o = 26.4 degrees C and T, = 40.0 degrees C for all phases. Other parameters, i.e. day length of half-maximum development rate at PIP, and physiological development days for FDP, were also estimated as common values for all cultivars. Different genotypes were mainly characterised for the sensitivity to photoperiod (shape parameter n) and BVP length. With n similar to 50, Felina 34 and Futura are regarded as low sensitive cultivars. Tiborszallasi was estimated as the highest sensitive cultivar (n close to 70), whereas Carmagnola and Fibranova showed an intermediate sensitivity (n similar to 62). Felina 34 also differentiated for its relatively short BVP length at optimum conditions, i.e. similar to 13 days; duration of about 20 days was the estimate for the other cultivars. Model performance against calibration dataset was good (percent relative root mean square in the range similar to 6-20%), and comparison against independent data also confirmed the general applicability of this model. Owing to the importance of flowering date in hemp management techniques, these results can be used in decision support for hemp production though further evaluation of the model under a variety of latitudes is required. (c) 2007 Elsevier B.V. All rights reserved.