Can spatial data substitute temporal data in phenological modelling? A survey using birch flowering.

In addition to the evaluation of long-term series, the analysis of spatial gradients, such as urbanization gradients, may be helpful in assessing phenological responses to global warming. But are phenological responses of birch (Betula pendula Roth) assessed by temperature variations comparable over time and space and can spatially calibrated models predict long-term phenological data adequately? We calibrated and tested linear regression models and the process-based DORMPHOT model on phenological and temperature data sampled along an urbanization gradient in 2010 and 2011 in the German cities Munich and Ingolstadt (spatial data). Additionally, we analysed data from the German Meteorological Service for the period 1991-2010 (long-term data). The model comparison showed that the DORMPHOT model performed better than the linear model. Therefore, the importance of forcing and chilling sums as well as photoperiod, factors which were all considered in the DORMPHOT model, was evident. Models calibrated on spatial data produced good predictions of spatial data, but they were less adequate for predicting long-term data. Therefore, a time-for-space substitution might not always be appropriate. This finding was also confirmed by a comparison of temperature response rates. The rate of change in the spatial data (-4.4 days °C(-1)) did not match the changes observed in the long-term data (-1.9 days °C(-1)). Consequently, it is important not to generalize results derived from one specific study method, but their inherent methodological, spatial and temporal peculiarities have to be considered.