Results

2.3.1 Physical environment

The depth contour of water temperature (15.9~32.1℃; Fig. 2.2A) revealed that the water column was well-mixed during winter (Dec~Feb of the next year). Stratification occurred during the period of early Apr to late Oct. Surface water temperature (17.5~32.1℃; Fig. 2.3A) varied seasonally with the coldest and warmest temperature recorded in Feb and Aug, respectively. Values of the mixing layer depth (MLD) ranged 3~90 m, with the shallowest and the deepest MLD in summer and winter, respectively (Fig. 2.3B). Weekly light intensity (0~102 mE m^-2 d^-1; Fig. 2.3A) showed apparent seasonality but seemed varied more than that of surface temperature. Daily precipitation ranged from 0~406 mm (Fig. 2.3C) with lower values recorded in the dry seasons, which covered the period from Nov to Feb of the next year. Higher precipitations came from two sources, the summer evening thunder-showers and the rainfalls induced by typhoon (Fig. 2.3C).

2.3.2 Chemical variables

In the epiliminion, most of the individual SRP (<0.02~0.25 μM P; Fig. 2.2B) and depth-integrated averaged SRP (SRPDIA; <0.02~0.15 μM P; Fig. 2.4B) concentrations were under detection limit (<0.02 μM P) during the stratified seasons except several spikes recorded during post-typhoon periods (Fig. 2.3C). Higher individual SRP and SRPDIA concentrations occurred in the mixing seasons, especially in the winter. The temporal changes of the SRPDIA concentrations in the hypolimnion (20~90 m) were higher than those recorded in the epilimnion (0~20 m) (Fig. 2.4B), but with similar trends (r = +0.60, p<0.01, n=126). Strong inter-annual variation of SRPDIA in both stratified and mixing seasons was noted (Fig. 2.4B). For the stratified seasons, SRPDIA

concentrations in 2007 were very high, even higher than the winter-spring (Dec~May of

the next year) values recorded in 2008. SRPDIA concentrations in the mixing season of 2007 were most undetectable.

During the investigation period, individual dissolved inorganic nitrogen (DIN, = nitrate + nitrite; 14~87 μM N; data not shown) concentrations were always detectable, even in the epilimnion during summer (Fig. 2.4A). In term of stoichiometry, DIN seemed to be much surplus to SRP. The molar ratios of N/P ranged from 230 to 2,995 molN molP^-1 with an average of 1,432±685 mol N mol P^-1 (Fig. 2.4C), which was about 100-fold greater than that of the Redfield ratio (N/P = 16; Redfield 1958). Total phosphorous (TP, = SRP + DOP) in the surface water ranged from 0.16 to 2.06 μM P with an average of 0.47±0.38 μM P (Fig. 2.5). SRP constituted ca. 8% of the TP.

2.3.3 Biological measurements

Vertical contours of Chl a (data not shown) indicated that phytoplankton biomass was restricted in the upper 20 m. Epilimnic depth-integrated averaged Chl a concentrations (ChlDIA; Fig. 2.6A; range, 0.5~9.7 μg L^-1; mean, 2.4±1.2 μg L^-1) varied seasonally, and basically followed the trend of temperature (Table 2.1). In this system, the scale of spring bloom was less significant when compared with that of autumn, as evident by the very high ChlDIA (>9 μg L^-1) recorded in Oct 2006. Vertical contour of picocyanobacteria (data not shown) indicated that they distributed mostly in the upper 20 m, and the abundance was higher in the surface water and then decreased with depth.

Epilimnic depth-integrated average of picocyanobacteria abundance (CYADIA) ranged from 0.3 to 3.7×10¹¹ cells m^-3, with an average of 1.5±0.6×10¹¹ cells m^-3 (Fig. 2.6B). The values of CYADIA were positively correlated with ChlDIA (Table 2.1), and higher CYADIA

values were generally recorded in late autumn. CYA were the most abundant species of the phytoplankton community in Feitsui Reservoir. During 2006 and 2007, CYA on

average constituted 87±12% of the total algal cell counts (Fig. 2.7). Heterotrophic bacteria abundance (BA) was high in the upper 20 m, and then dwindled with depth (data not shown). In term of seasonal variation, depth-integrated average of BA (BADIA; Fig.

2.6B; range, 0.8~4.6×10¹² cells m^-3; mean, 2.3±0.8×10¹² cells m^-3) changed positively with temperature and CYADIA (Table 2.1). A negative correlation was observed for BADIA vs. MLD.

2.3.4 Seasonal and inter-annual analyses of bulk APA

Signals of bulk alkaline phosphatase activity (APA) were only observed in the upper 20 m, and then decreased significantly with depth (Fig. 2.2C). Epilimnic depth-averaged bulk APA (APADIA; Fig. 2.6C) varied ~100X with a range of 1.6~95.2 nM h^-1, and a mean of 40.4±21.5 nM h^-1. In the epilimnion, values of the biomass normalized bulk APA (i.e. specific APADIA; Fig. 2.6C) varied ~10X with a range of 124~1,253 nmol mgC^-1 h^-1, and a mean of 391±207 nmol mgC^-1 h^-1.

In term of seasonal variation, values of bulk APADIA were positively correlated with of the changes of temperature, light intensity, ChlDIA, CYADIA, and BADIA; values of bulk APADIA were also negatively correlated with MLD, DINDIA, and SRPDIA

concentrations (Table 2.1). A closer examination indicated that the temperature response of bulk APADIA of each year were different (Fig. 2.8). The slope of bulk APADIA vs.

temperature of 2009 was insignificant, while the slopes of the other 3 years were significant with values ranged 0.08~0.14. The slope of 2008 was significantly different from those of 2006 and 2007 (ANCOVA, p<0.05), while the latter two were not different from each other (ANCOVA, p>0.05). The relationships of specific APADIA to other variables were the same as those of bulk APADIA (Table 2.1).

Results of the multiple linear regression analysis indicated that 65% of the bulk

APADIA variability could be explained by the combination of light intensity, MLD, DINDIA, SRPDIA, and CYADIA (Table 2.2). The relative importance (standardized regression coefficient, i.e. Beta weight) of these independent variables on bulk APADIA

in order was -0.40 for MLD, 0.30 for CYADIA, -0.22 for DINDIA, 0.21 for light intensity, and -0.15 for SRPDIA. The same procedure was performed on the data of each single year. The results showed apparent inter-annual variability of the independent variables in explaining the variation of bulk APADIA. In 2006, ChlDIA and BADIA were the best factors for the changes of bulk APADIA. In 2007, the best combination switched to MLD only. In 2008, MLD and ChlDIA were responsible for 72% of the variation. In 2009, light intensity and CYADIA explained 72% of the variation. Basically the analysis results of specific APADIA were the same as those of bulk APADIA. For the pooled data set, light (Beta = 0.45) and MLD (Beta = -0.41) were the two factors affecting the variability of specific APADIA most.

2.3.5 Typhoon impact

During the investigation period, a total of 18 typhoons had swept through the study site during summer periods (Fig. 2.3C). To test typhoon effects on the behaviors of bulk APADIA and specific APADIA, the data collected during the period of Jul~Sep were analyzed. Table 2.3 indicated that 2007 and 2008 could be categorized as the strong-typhoon years with typhoon impact indices of 12.2±9.2 and 13.1±6 m² S^-1, respectively. On the other hand, typhoon impact indices of 2006 (4.3±2.4 m² S^-1) and 2009 (2.8±3.3 m² S^-1) were ~one-third of those in 2007 and 2008. Accordingly, 2006 and 2009 were considered as the weak-typhoon years. Physical parameters (i.e.

temperature, light intensity, and MLD) showed no difference between strong- and weak-typhoon periods. SRPDIA concentration (0.06±0.03 μM P) in 2007 was ~2-fold higher than those of the other years (p <0.01). CYADIA value (1.3±0.4 ×10¹¹ cells m^-3)

recorded in 2008 (one of the strong-typhoon years) was significantly lower than the weakest typhoon year (2009; 1.9±0.3×10¹¹ cells m^-3). BADIA values of the strong-typhoon years (3.2~3.4×10¹² cells m^-3) were higher than those of the weak-typhoon years (2.2~2.7×10¹² cells m^-3). Bulk APADIA and specific APADIA showed no difference between the strong- and the weak-typhoon years. Correlation analysis of the pooled data set showed that bulk APADIA was positively correlated with ChlDIA and CYADIA, and specific APADIA correlated positively with bulk APADIA (Table 2.4).