Detection of faint, forced signals in the climate system is of importance in several areas of climate change research. First, the detection and discrimination of one signal from another (eg, greenhouse gas from anthropogenic aerosol) is important in establishing links between various anthropogenic activities and climate change. Second, the detection of such signals constitutes a measure of their strengths and therefore it provides an estimate of climate sensitivity. Third, the detection methods can be used to test climate models by comparing outcomes to predicted signals.
One of the problems in detection theory is the assessment of the errors in the analysis itself. In this study we examine the effects of errors in the signal waveform shape on the detection process. Typically, one must generate a signal from a climate model in order to look for a pattern in the data stream. If the model makes a slight error in the space-time signature how will this bias our estimate of the signal strength.
Furthermore, how will errors in one hypothetical signal contaminate those of another signal? In this study we develop a framework for studying these questions and we provide numerical results which should be of general use.