3. Results
3.1. Streamflow response to restoration of native forests
Predicted values of seasonal streamflow were within +/- 80 mm of observed values, and were evenly distributed as positive and negative deviations (Figure S1 a). In relative terms (i.e., %), predicted seasonal values were within +/- 40% of observed values (Figure S1 b). Changes were considered to be practically significant when they were more than 50 mm or 40% different than predicted. This level of uncertainty is comparable to confidence intervals obtained from long-term studies of paired catchments (e.g., Jones & Post, 2004; Perry & Jones, 2017).
Catchments differed substantially in the relationship of runoff to precipitation (Figure 3). Before clear-cutting, streamflow was almost two times higher per unit of precipitation at RC6 than at the contiguous catchment, RC5, with which RC6 shares a long catchment divide. Streamflow also was almost two times higher per unit of precipitation at RC11 relative to its nearby but not contiguous catchment, RC10 (Figures 1, 2 and 3). After clear-cutting, the slope of the double mass curve increased in all catchments, indicating increased streamflow (Figure 3). The greatest increase occurred at RC10, and the least at RC6, which was not clear-cut. The sum of cumulative increases at RC6 and RC5 in the post-treatment period (2011 to 2019) is similar to the increase at RC10, and higher than RC11. These observations suggest that hydrologic processes differ among the catchments, and that some portion of the flow from RC5 is transferred into RC6.
Over the entire restoration period (2011-2019), streamflow increased by 24% at RC5, 73% at RC10 and 21% at RC11, relative to the pre-treatment period (2006-2010), but precipitation changed by only 3% (Table 2). In most post-treatment years, annual streamflow increased by >200 mm after clear-cutting in most catchments, relative to predicted values based on precipitation (Figure 4). As noted above, streamflow in RC6 increased immediately after the clear-cut, whereas streamflow did not increase at RC5 until the second year after the clear-cut. (Table 2, Table S2).
Seasonal streamflow response in absolute terms (mm) in the three catchments under restoration was greatest in winter, followed by fall, spring, and summer, based on the before-after method (Figure 5, Figure S2). Seasonal streamflow increased by more than 300 mm (16 to 31% of pre-treatment annual streamflow) in all three catchments only in winter of 2014 (Table 2, Table S2). Streamflow response based on predicted streamflow relative to 2009 and 2010 (%) was greatest in summer and fall, and smaller in winter and spring (Figure 6). Streamflow increased by more than 150% relative to pre-treatment in fall and summer of some years at RC5, RC10, and RC11. Streamflow increased by more than 50% in all four seasons at RC5 and RC10 in most post-treatment years. Some differences were apparent in streamflow responses relative to the 2006 to 2010 pre-treatment period compared to those relative to the 2009 and 2010 pre-treatment period. The 2006 to 2010 pre-treatment period included 2006 to 2008, when streamflow was estimated manually once a day, or from regression relationships with precipitation, and years with high and low precipitation. In contrast, the 2009 to 2010 pre-treatment period included years with consistent automated measurements of streamflow, and years with average precipitation relative to the study period.
3.2. Vegetation recovery in catchments under restoration
Vegetation cover of non-tree species from natural regeneration increased rapidly after clear-cutting and planting of Nothofagus seedlings in RC5, RC10, and RC11. Non-tree vegetation cover reached a maximum of 63 to 106% between 2012 and 2016, one to five years after restoration began, and then declined in 2020 as cover of planted and naturally regenerated tree species increased (Table S3, Figure S3). Cover of non-tree species and survival of planted trees was lowest in RC5, whereEucalyptus stumps sprouted vigorously, compared to RC10 and RC11, where Eucalyptus sprouting was prevented by application of herbicide (glyphosate) to cut stumps. Eucalyptus recruitment from seeds occurred in all three catchments (Figure S3, Table 3).
By 2020, vegetation cover of tree species ranged from 48% (RC5) to 78 % (RC10) (Table 3). The density of surviving N. dombeyi saplings planted in 2011 was lower in RC5 (43%) compared to RC10 (90%) or RC11 (73%). The density of saplings and seedlings of naturally regenerated native tree species was thirty times higher than that of the planted species, N. dombeyi (Table 3, Table S4). Over the period 2012 to 2020, densities of saplings of native tree species increased in all three catchments, and densities of Eucalyptus saplings declined in RC10 and RC11 (Figure S3). In 2020, total basal area of plantedN. dombeyi and naturally regenerated trees was much lower in RC5 (2.3 m2/ha) than in RC10 (9 m2/ha) or RC11 (9.8 m2/ha) (Table 3). In contrast, basal area of Eucalyptus trees that seeded in or sprouted from cut stumps was greater in RC5 than in RC10 or RC11 (Table 3). From 2010 to 2020, basal area of the Eucalyptus plantation in RC6 increased from 38 to 61 m2/ha (Table 1). By 2020, basal area of trees in the portion of the catchments under restoration was 16 to 20% of native forest basal area in the riparian zones and 12 to 18% of the basal area in the untreated Eucalyptus plantation in RC6 (Figure 1, Table 3).
3.3. Factors affecting streamflow response: precipitation variability, catchment hydrology, native vegetation recovery
Streamflow response varied with precipitation, vegetation development, catchment hydrology, and their interactions. Seasonal precipitation varied during the study period (CV 0.2 to 0.41 for 2006-2019, Table 4). Annual precipitation was lowest in water years 2007 and 2016 and highest in water years 2006 and 2012. Summer precipitation was lowest in 2006 and 2014 (Table 4). Runoff ratios were positively related to precipitation among years and among seasons (Figure S4). The largest increases in streamflow during the restoration period occurred in the winter of 2014 (Figure 5, Figure 6, Figure S2), after unusually high fall and winter precipitation (Table 4). The largest post-harvest streamflow deficits occurred in 2016 (Figure 5, Figure 6, Figure S2), when precipitation was very low (Table 4).
Base flow on average over the study period accounted for 42 to 45% of total flow in fall, 50% in winter, and 52 to 54% in spring in all catchments, with little variation among years (Figure S5). Average summer base flow was lower at RC5 and RC10 (base flow 37 to 41% of total) compared to RC6 and RC11 (50 to 53%) (Figure S5).
Summer base flow was consistently lower and more variable at RC5 and RC10 than at RC6 or RC11. During a series of years with comparatively low precipitation in spring and summer (2014 to 2016, Table 4), summer base flow declined steadily in RC5 and RC10, but it recovered by the end of the 2016 water year (i.e., January to March of 2017, Figure 7). In contrast, summer base flow remained constant in all years at RC11, except at the end of water year 2014, when it declined abruptly for one year. Base flow was consistently high in all years in RC6.
Base flow was significantly positively related to precipitation in the same season in fall, winter, and spring at all catchments. Summer base flow was significantly positively related to spring precipitation in RC5 and RC6; these were the only significant lagged responses of base flow to prior precipitation (Figure 8).
Base flow increased by 28 to 87% in the catchments under restoration for the period 2011 to 2019 compared to the pre-treatment period (Table 5). Annual base flow as a percentage of total flow was consistently higher than the long term mean in 2017 to 2019 at all catchments, despite below-average annual precipitation in 2018 and 2019 (Figure 9). Increasing base flow trends are most evident in winter, spring, and summer (Figure S6).
Streamflow increases were consistently high during the restoration period in RC10 (Figure 3, Figure 4, Figure 5, Figure 6), which had the highest survival of planted Nothofagus , highest density of other native saplings and seedlings, absence of Eucalyptus saplings or trees, and highest percentage of the catchment covered by the riparian buffer (Table 1, Table 3, Figure S3).