Oak budburst model

 

The budburst model of Hanninen (1990) which is based on daylength (and thus latitude) and accumulated temperature has been calibrated using data from the IPG site at Headley which operated between 1967 and 1981. The model has subsequently been run for three other sites (Ashtead, Middlesex; Walsall, West Midlands; Ponteland, Northumberland) and compared with observed data provided by Tim Sparks of the Centre for Ecology and Hydrology. When the observed date of budburst at Ashtead (the most complete data-set) is compared with that predicted by the model, agreement is generally good, although for some years (eg 1972, 1973, 1979), significant discrepancies are apparent.

When the three additional data-sets are analysed, with model predictions based on the Headley calibration, it is apparent that at an individual site level, the agreement between modelled and observed budburst is poor in some cases. However, when data from all four sites are analysed as a single data-set, correlation is good and the data fit well with the 1:1 line; indeed, when the regression is forced through the origin (with minimal effect on correlation (r2= 0.646 vs 0.652) the slope is 1.07.

The fit is poorest at Ponteland, with a lower slope indicated. This may reflect differences in climate between the closest suitable meteorological station (Cockle Park) and the site of observations (20 km apart, with Cockle Park 10 km distant from the coast, and Ponteland, 20 km). Given the potential for discrepancies in ‘observer performance’, together with differences in location of weather stations and trees, the fit is excellent, and provides a robust basis for the use of the model to predict budburst in oak across the UK for evaluations of climate change, water use, carbon exchange and pollutant uptake/exposure. The model has also been run for a number of sites in the UK to provide an indication of the geographical variation in flushing dates for two contrasting years (1986 and 1990) of observed late and early budburst at Ashtead (day 127 and 86, respectively).  

Regional variation in the flushing date of oak for an early (1990) and late (1986) year, as predicted by the oak budburst model

Location	Latitude	Longitude	Altitude	Flushing date
				1986	1990
Penzance, Cornwall	50.1	-5.5	19	117	77
Nettlecombe, Somerset	51.1	-3.3	96	132	80
Goudhurst, Kent	51.1	0.5	90	131	83
Alice Holt, Hampshire	51.2	-0.8	115	136	95
Westonbirt, Gloucestershire	51.6	-2.2	135	140	102
Velindre, Powys	52.0	-3.2	152	135	90
Santon Downham, Norfolk	52.5	0.7	24	131	81
Sutton Bonington, Nottinghamshire	52.8	-1.2	48	133	88
Aber, Gwynedd	53.2	-4.0	18	129	88
Harrogate, North Yorkshire	54.0	-1.5	66	140	98
Grizedale, Cumbria	54.3	-3.0	91	139	111
Clatteringshaws, Kirkcudbrightshire	55.1	-4.3	178	154	122
Kielder Castle, Northumberland	55.2	-2.6	201	155	122
Lochawe, Argyll	56.0	-5.4	5	138	122
Braemar, Aberdeenshire	57.0	-3.4	339	165	133

To aid analysis of phenology, and also make the budburst model more widely available, a user-friendly ‘front-end' has now been added to the model, and will enable accessible re-parameterisation for individual species.