Detecting climate-induced patterns using wavelet analysis.

Abstract

One of the difficulties encountered in the detection of ecosystem responses to climate change is distinguishing climate-induced patterns from those created by other sources. For example, changes in the trend of stream discharge records over time may reflect a composite response of changes in the climate (i.e. precipitation and temperature), land-use (e.g. timber harvesting and grazing), and local basin characteristics. Methods which quantify and relate information of temporal and spatial patterns across scales are critical to assess climatically induced changes in the forest and stream ecosystems. A methodology utilizing wavelet analysis is introduced for the purpose of identifying and isolating inferred climatic components of the hydrologic record. Trends observed in stream discharge records from eastern Oregon, USA are identified and used to illustrate the utility of a new time series technique, wavelet analysis, as a complementary approach for discerning pattern. This methodology affords an informed procedure for choosing filter dimensions for the purpose of signal decomposition. The wavelet cross-covariance is applied to precipitation and discharge records to identify the climatic component in the discharge record. Reconstruction of these dominant frequencies is effected to isolate the climatic components. The discharge pattern shows two dominant scales of pattern coincident with the precipitation record. A 3-year half-period pattern is found to be correlated with the Southern Oscillation Index at the same frequency.