TIME STRUCTURE OF SELECTED INTEGRATED MONITORING OF ENVIRONMENT DATA SERIES

Regional Monitoring of Natural Environment 2002, No 3, 61-68

TIME STRUCTURE OF SELECTED INTEGRATED MONITORING OF ENVIRONMENT DATA SERIES

 

Alfred Stach

 

Summary

 

Semivariogam analysis was applied to detect hidden time structure of three types of data: daily mean soil temperature at 50 cm depth, daily cumulative average of gas/aerosol pollutants concentrations and monthly weighted mean precipitation pollutants concentrations. The first data set are of high time inertia, high measurement frequency and low level of processing of original measurements. The second one differs by high time variability, looks even chaotic and lacks of any processing of original measurements data. The third data set comes from much less frequent measurements and was highly processed. Soil temperature time pattern is dominated by seasonal component. Subtracting seasonal component exposes short time autocorrelation structure with 12 and 51 days range. It’s effect of soil thermal conductivity and reflects time delay in heating and cooling at 50 cm depth. Combining deterministic (seasonal sine function) and probabilistic (semivariogram) approach yields prediction of missing data at 0.1ºC precision level so it is comparable with measurement error. Monthly precipitation solute data shows three kinds of time pattern: seasonal (Ca, Mg, Na, S-SO4, N-NH4), long time tendency (trend, Cl, Zn) or lack of autocorrelation (CTY, Mn, P-PO4). Gas/aerosol pollutants time series shows similar, strong, short time autocorrelation structure, with 3.1–3.9 and 7.5–10 days ranges. It is the effect of typical frequency of weather changes in Poland. Each three for four days (100 times in the year) new air masses flowing in and weather changes. This data shows that complete air replacement (and lack of similarity in its physical and chemical properties) take place after two weather front shiftings. Pollutants variability in the time shorter than sampling interval (1 day) accounts 25 to 35% of total variability. Semivariogram analysis can also be an aid in choosing optimal measurement frequency, loosing some redundant information.