Impact of Environmental Conditions on Long-Memory of Surface Air Temperature

The memory of dynamic processes refers to the characteristic of dynamic systems where historical states influence their current and future behaviour. In modelling dynamic systems, it is crucial to ensure that the model design allows simulations that consider the system’s history and its impact on future dynamics beyond the integration interval. An important step toward this goal is identifying the characteristic time scale of the memory, i.e. memory longevity.

 In this research, we use the atmosphere as a dynamic system and air temperature at 2 m height as a variable that can carry memory information. To assess the impact of environmental conditions on memory characteristics, we selected meteorological data measured at climate stations and data from rural areas near Novi Sad and Vrsac (Serbia) measured at automatic weather stations (AWS) as part of the Forecasting and Warning Service for Plant Protection of Serbia (PIS) meteorological monitoring system.

Following Caballero et al. (2002), we analyzed the power spectral density (PSD) and autocorrelation functions (AC) of three air temperature time series. According to results of previous studies, processes represented with PSD following power law function, S(w) at low frequencies (w), [S(w) = w^-2d] are collectively referred to as “long memory” processes if 0<d<0.5. Parameter d is commonly described as the “intensity of long memory”.

Initial results obtained for climate station data series show clear differences between high- and low-frequency trends in PSD estimates for air temperature at the selected locations. Differences in AC are particularly noticeable for longer lags. 

Caballero, R., Jewson S., Brix, A., 2002: Long memory in surface air temperature: Detection, modelling, and application to weather derivative valuation. Clim. Res., 21, 127-140. 10.3354/cr021127.