We are pleased that Mattson and Likens have throughly reviewed our recent paper on the acid-base trends in surface waters in Maine. This type of scrutiny is what fosters discussion and consensus in science. Mattson and Likens re-iterate several points we made in Kahl et al. (1993), but also take issue with some points and citations in our paper. We show here that our original statements were accurate, although we did incorrectly cite Kramer (1990) and Metcalf and Gerlach (1990) as being published in 1989.
We agree completely with Mattson and Likens that ANC, rather than pH, offers the superior approach to trends detection in lake acidity. That is why Kahl et al. (1993) focussed on the ANC data for lakes in Maine. However, the historical ANC values are dependent on colorimetric pH data, so it is not clear what relevance their statement has for improving the determination of trends based on historical colorimetric data.
We cited Stauffer's (1990) conclusion that the ANC of some large Maine lakes has increased by as much as 40% since the 1940s, based on this re-examination of the historical data. Mattson and Likens state that errors of as little as 0.2 pH units in colorimetric pH can change these estimates by as much as 58%. They imply that this difference might have been biased in only one direction, and thus would change the magnitude or direction of the ANC change suggested by Stauffer. It is not appropriate to debate the validity of Stauffer's method here, but we are glad that Mattson and Likens have inadvertently emphasized our point for us again: It is not possible to correct the historical data to unequivocally discern trends within the limits of natural variation. It is probably true that for Maine data 'historic errors cannot be reconciled today' (paragraph five of their comment). However, it is not clear how they concluded that the historical data were useful in the Adirondacks, but not in Maine.
Mattson and Likens re-evaluated our data on changes in base cation concentrations in lakes at Acadia National Park. They conclude that the decline is significant at the 93% confidence level using a sign test. We also noted the decline, which was not significant at the 95% confidence level using a t-test (although we did not report it, the test result showed it to be significant at the 86% CL). We are glad that the two tests yield similar results.