It is the nature of the scientific method to use a variety of models, procedures, techniques and data to quantify and analyse natural phenomena. Recently, in Reddy et al. (2006), we published a study in which υ 18 O values for precipitation, surface-and ground-water samples from the Shingobee River Headwaters Area (SRHA) were analysed using an amplitude-attenuation (convolution integral) approach for estimating mean residence times (MRTs). This approach has been used in many small watershed studies (e.g. Stewart and McDonnell, 1991; DeWalle et al.disagree with our method and present an alternative approach for interpreting υ 18 O values in the SRHA. We feel that this comparison with our methodology is unfair. Criss and Winston (2006) focus their attention on the weaknesses in our method and the strengths of theirs, while overlooking our detailed evaluation of model predictions with other independent watershed hydrologic properties.The comment by Criss and Winston (2006) makes several points, some of which we agree are valid criticisms of our paper. In this response, we will address their concerns regarding (1) our interpretation of the current state of the literature, (2) the data we used in the analysis, and (3) the relative merits of the two modelling approaches. In so doing, we will update the dataset with precipitation υ 18 O data from 1990 to 2004, and compare the two approaches to provide a fair basis for understanding the strengths and weaknesses of both methods. Additionally, we will discuss some technical issues that arise for scientists in choosing between simple and complex models.
CURRENT STATE OF THE LITERATUREThe goal of the Reddy et al. (2006) paper was to test the applicability of the υ 18 O amplitude-attenuation approach for estimating MRTs in a well-characterized catchment with significant surface-water-ground water interactions, and to evaluate the limitations of this approach by comparison of model predictions and other