rivers, for the arctic and subarctic Mackenzie river watershed (data from Reeder et al., 1972), for the Lower and Central Amazon basin (Stallard, 1980), for the Andean tributaries of the Amazon (Stallard, 1980), and for the rivers of Japan and Thailand (Kobayashi, 1959, 1960). A global distribution is also considered based on 50 major world rivers (distribution weighted by water discharge, data from Meybeck, 1979). Three distributions for small streams are given: (1) a set of 250 monolithologic pristine French Streams located 20 to 600 km from the coastline (Meybeck, 1986); (2) a set of 75 monolithologic streams (Miscellaneous Streams) from various literature sources (chosen to be representative of global rock types) and grossly corrected for oceanic aerosol influence; (3) a set of pristine streams, the Temperate Stream Model, derived from set 1 after systematic correction of atmospheric inputs and selected as representative of the global distribution of rock types (Meybeck, 1987). The distribution of some ions and silica is given for these small, medium, and large rivers on Figure 4.4 and Table 4.3.
The K+ distribution (Figure 4.4A) can be considered as unimodal and log-normal. Potassium originates mainly from a few aluminosilicate minerals that weather at comparable rates. For the Temperate Stream Modal, K+ is much lower, due to an overcorrection of atmospheric inputs. Other distributions are similar, median values are within a factor of two and distributions are parallel, except for Thailand rivers for unknown reasons. The sample of major rivers is similar to the French Streams and Miscellaneous Streams, which indicates that the data used are still largely unaffected by pollution.
The dissolved silica pattern (Figure 4.4B) is similar to that of K+, which confirms a common origin. A major discrepancy is noted for the Mackenzie tributaries for which the SiO2 level is about half that in other regions and has a significant drop in the lower decile value. This is most probably due to the SiO2 uptake in numerous lakes of this basin, whereas K+ is not used by freshwater plants. As a result, the K-/SiO2 ratio of medians is 0.36 in the Mackenzie, compared to 0.04 to 0.07 for other pristine waters (Meybeck, 1987).
Bicarbonate distribution (Figure 4.4C) is usually bimodal. The first mode corresponds to the soil and atmospheric CO2 used in the weathering of noncarbonate rock (Garrels and Mackenzie, 1971), the second to the weathering of carbonate rocks, in which only 50 percent originate from calcite or dolomite. Concentrations related to the second mode are generally 5 to 10 times greater than those of the first one. In large basins where carbonate rocks are absent, as in Central and Lower Amazonia, the distribution is unimodal. In sedimentary regions, the HCO3- distribution shows an upper limit near 6 meq/l that corresponds to the CaCO3 saturation.
Sodium distribution (Figure 4.4D) is also complex, due to a triple source: weathering of aluminosilicate mineral, weathering of rock salt, and inputs of oceanic aerosols. The Central and Lower Amazon tributaries have an unimodal Na+ distribution reflecting the first source only. The influence of oceanic aerosols is well illustrated by Japanese rivers; the minimum Na+ values are much higher than those of the Amazon tributaries. Rock salt dissolution leads to the very high values, from 0.5 to 10 meq/l, noted for the Mackenzie tributaries, Andean rivers, French Streams, and Miscellaneous Streams.
The percentiles values of distributions are presented on Table 4.3A for small streams (French streams and literature