5.1. Evaluation of Possible Factors Influencing Declines in Dissolved Silica
 Bedrock lithology is known to influence the rate of physical and chemical weathering [Bluth and Kump, 1994], and therefore it is not surprising that high silica concentrations are associated with watersheds underlain by the more reactive basaltic and granitic bedrock, while lower concentrations are found in siliciclastic watersheds. What deserves further scrutiny is the overall decline in dissolved silica concentrations in western Virginia streams and the fact that the magnitudes of the trends are related to bedrock geology and mean silica concentration. The geographical analysis provides an empirical context for the dissolved silica trends, however a process-based explanation is sought.
 Trends in atmospheric deposition were considered as a possible cause for changes in the chemical weathering rates of the study watersheds. Decreases in acidic deposition have been reported in the eastern U.S. over the past 12–20 years [Stoddard et al., 2001; Skjelkvåle et al., 2005] due to state and federal emission control programs. These decreases, however, do not necessarily yield commensurate improvements in stream water quality. In the Blue Ridge region of Virginia, reduced sulfate deposition has not lead to significant changes in stream water sulfate concentrations or acid-neutralizing capacity, making this region distinct from areas in the northern U.S. that have more recently formed soils [Webb et al., 2004]. The availability of sulfate in watershed soils, which forms sulfuric acid and weathers silica ions from the regolith and underlying bedrock [Rice and Bricker, 1995], has not changed considerably in the western Virginia region due to the large sulfate absorption capacity of the watershed soils.
 Even if watershed acidity within the study region had become less acidic, the effects on dissolved silica concentrations would have been minor. Stream water concentrations exhibit little correlation with pH for the six core sites (Figure 3), a finding that was supported by the equilibrium dissolved silica concentrations estimated by the PHREEQC model applied to the dominant bedrock mineralogy. These observational and model results are consistent with Brady and Walther , who reported that dissolution of most silicate forms is independent of pH within naturally-occurring ranges. Trends in atmospheric deposition and watershed acidity are thus not likely to be responsible for the declines in dissolved silica concentrations.
 Precipitation has an influence on stream water dissolved silica concentrations through its effect on hydrological flow paths [Scanlon et al., 2001]. In general, water associated with overland or shallow subsurface flow is depleted in dissolved silica relative to groundwater due to reduced contact time and less availability of weatherable material. If precipitation had increased over the study period, groundwater would account for a smaller fraction of stream flow, leading to a decrease in dissolved silica. This was not the case, however, as both precipitation and stream discharge showed no significant trends. Furthermore, dissolved silica concentrations differed between the early and late portions of the time series at like levels of discharge (Figure 4), providing clear evidence that a factor other than hydrology caused the declining trends in dissolved silica.
 Analyses of stream chemistry data for the two core watersheds with records extending back to 1979 reveal that dissolved silica concentrations actually increased until a time coincident with gypsy moth disturbance, after which the trend reversed (WOR1 data are shown in Figure 5). The declining portion of the WOR1 record overlaps the majority of the timeframe for the more widespread trend analysis (1988–2003). Rather than geochemical or hydrological factors being responsible for the declining trends in dissolved silica concentration, the extended analysis shown in Figure 5 points to the disturbance event and its legacy as being central to the observed declines.
 The gypsy moth, an invasive species native to Europe and Asia, has had a devastating impact on hardwood forests throughout eastern North America [Elkinton and Liebhold, 1990; Davidson et al., 1999]. It infested the northern region of the Virginia survey area beginning in 1986 and reached the southern end by 1991. The gypsy moth attacked several key tree species, including many oaks and hardwoods, causing intense defoliation as well as some tree deaths. Gaps in the forest canopy led to increased growth of understory vegetation, in some cases leading to species replacement [Jedlicka et al., 2004]. Species-specific uptake rates of silica, a key nutrient for plant health, are unknown for the majority of trees present in SNP, as the literature has focused mainly on agricultural plants. In a review of leaf carbon to silica ratios in tree species, Fulweiler and Nixon  report that this ratio for oak is relatively low compared to other sampled species, suggesting that if oak had been replaced on a widespread basis following gypsy moth disturbance, the species-specific uptake rate would not necessarily be enhanced.
 Gypsy moth disturbance reduces the overall net primary productivity of the forests during the period of defoliation, followed by a rebound and thus greater silica uptake by the terrestrial vegetation during forest recovery. This does not appear to be responsible for the long-term declines in dissolved silica concentrations following gypsy moth defoliation, however, since tree productivity is impaired for several years at most [e.g., Muzika and Liebhold, 1999] before returning to pre–gypsy moth conditions. An alternative model is needed to describe the trends that are observed over the more long-term, decadal timescales.
5.2. A Conceptual Model: Benthic Diatoms
 In keeping with the findings from the above analysis, an adequate conceptual model to account for the observed trends in dissolved silica concentrations must: (1) result in long-term decreased concentration through time, (2) be associated with the gypsy moth defoliation, (3) primarily impact concentrations during low-flow conditions, and (4) explain why streams with highest mean dissolved silica concentrations have experienced the greatest declines through time. Benthic diatoms, which remove silica from the water column in the formation of their shell during growth and are fundamental to the food webs of stream ecosystems [e.g., Mayer and Likens, 1987; Lamberti, 1996], were present in the streambeds at each of the core sites. Here we develop a conceptual model to examine the role of benthic diatoms in affecting long-term trends in stream water dissolved silica.
 Gypsy moth defoliation resulted in increased sunlight penetration through the forest canopy, and this increased exposure in streams could have caused photosynthetic diatom populations to expand [Round et al., 1990]. In addition to increased radiation, nitrate levels in streams dramatically increased following gypsy moth defoliation and remained high over a period of many years (Figure 5b). Since freshwater diatoms can be nitrate- or silica-limited [Kilham, 1971; Wall et al.,1998; House et al., 2001], the enhanced nitrate concentration in the streams could have exceeded a threshold such that this limitation on population growth was removed. A steady increase in diatom population would cause a reduction in dissolved silica concentrations in the years following the defoliation event.
 For this conceptual model to hold, the statistically significant relationship between mean dissolved silica concentrations and slopes of the dissolved silica trends should be explained. If nitrate was no longer limiting the diatom populations during the post-gypsy moth time period, stream water silica instead could have been limiting. Both silica uptake rate and diatom cell division rate are commonly modeled according to a Michaelis-Menton equation with respect to dissolved silica concentration [Martin-Jezequel et al., 2000]. This implies that streams with higher mean dissolved silica concentrations should also have the largest diatom populations and uptake rates. In fact we see that PINE and NFDR, the two core sites with the highest mean dissolved silica concentrations, also have the highest diatom densities. Beyond this, the relationship between diatom density and mean stream water dissolved silica concentration is not well-defined. A more spatially extensive diatom sampling program would be needed to more rigorously test this hypothesis.
 Another possible explanation for the statistically significant mean versus trend relationship is the fact that the gypsy moth defoliation progressed from the north (dominated by basaltic and granitic bedrock) to the south (dominated by siliciclastic bedrock) of the study area during the early portion of the 1988–2003 timeframe. The later arrival of the gypsy moth in the southern watersheds could have reduced the slope of the linear best fits to the dissolved silica time series.
 Uptake of dissolved silica by diatoms would more prominently influence low-flow concentrations. This is due to the fact that during these periods there is less overall mass of silica in the water column, and therefore uptake would have a greater affect on the relative amount of mass. For high-flow conditions, although approximately the same amount of mass would be removed, the change in mass would be small compared with the large initial mass of dissolved silica within the water column, and therefore the concentrations would not be greatly impacted. This can be illustrated by a simple conservation of mass model, assuming steady state conditions (and therefore assuming constant residence time with the stream network):
where C is dissolved silica concentration (μmol L−1), Q is stream discharge (m3 s−1), r is diatom dissolved silica uptake rate (μmol s−1), and the 1000 is a conversion factor (L m−3). This assumes a power-law relationship between concentration and discharge for flow into the stream network, described by the coefficient a and the exponent b. Results of this simple model are shown for Staunton River in Figure 6, depicting the influence of in-stream uptake rate on the concentration-discharge relationships. Values of uptake rate, r, were chosen on the basis of fitting the model to the data for the early and late portions of the time series. Literature values, specific to the diatom genera and environmental conditions of the stream, are not available as an independent confirmation of these fitted estimates. This simple model suggests that increases in uptake rate by the diatom community could lead to decreases in stream concentrations, most notably during low-flow conditions.
Figure 6. Concentration-discharge plot for Staunton River, with the time series divided into early (1992–1997) and late (1998–2003) periods. Also shown are the modeled effects of in-stream uptake of dissolved silica by diatoms. Note that diatom uptake more strongly influences concentrations during low-flow conditions.
Download figure to PowerPoint