4.2 What are the effects of TMCF conversion to shaded coffee on streamflow?
At daily scale, shaded coffee had a lower mean streamflow and lower rainfall to runoff ratio, despite having similar rainfall inputs compared to PF. In the dry season, SC also showed a slightly lower mean annual low flow (MALF), and a lower recession constant (k ) (Table 3). A significantly lower RR and MALF was expected in SC, because this catchment is located at a lower elevation (472 m difference), which promotes a higher evapotranspiration (Ramirez et al., 2017), while ET in the PF is likely to be energy-limited, as described in previous paragraphs. Additionally, the soil evaporation in a shaded coffee farm may account for around 13 % of the total ET (Holwerda & Meesters, 2019), while the dense canopy in PF can minimize this output. However, the lower aboveground biomass of SC is also associated with a lower interception capacity, and an approximately 10% to 15% lower ET compared to PF (Holwerda et al., 2010; Ponette‐González et al., 2010). This interception difference may compensate for the increased of soil evaporation in SC. In this work, SC had a significantly lower RR than PF, as expected, but surprisingly, SC also exhibited a relatively higher MALF and k , when compared to the other studied land covers (Table 3) and with previous studies (Muñoz-Villers & McDonnell 2013); regardless of the expected higher ET described earlier. This suggests that shade coffee may preserve a large part of its original capacity to sustain baseflow during the dry season, associated with its higher soil infiltration capacity when compared to IP (Table 1).
At the event scale, the Dunn’s test indicated that the hydrologic response of SC in terms of total runoff, quickflow, baseflow and peak discharge were not statistically different (alphabet grouping Figure 4a-d) compared to the PF. The metric normalized by streamflow (Qbf/Qt ) also suggested that SC has larger dominance of baseflow compared to YF and IP. The ANCOVA analysis largely confirmed these results (Table 5). These findings indicated that the SC micro-catchment shows a similar capacity to modulate peak discharge compared to that of PF. However, the average time to peak and lag time responses of the SC were the lowest (Table 4), indicating a faster response, even when this micro-catchment exhibit the largest area of the studied sites (Table 1). We attributed this result to the fact that SC had an intermediate soil infiltration capacity (48 mm h-1) and the highest soil bulk density (Table 1). Our results agreed with previous research conducted in the study region which suggested that shade coffee soils had higher bulk density and lower soil porosity than forest soils, in addition to intermediate saturated hydraulic conductivities (Geissert & Ibáñez, 2008; Marin-Castro et al., 2016). Our study demonstrates that shade coffee preserved large part of the pre-disturbance capacity to sustain baseflow and modulate peak discharge, despite the differences in elevation and slopes that may have affected our results.
At the daily scale, the primary and intermediate forest presented similar responses characterized by lower variability in their daily flows, higher water storage capacity and lower flow velocities (high recession constants), and higher MALF , compared to that of the young forest (Table 3). Figure 3a demonstrates that, during the dry season, the intermediate forest had a streamflow closer to that of primary forest. These results correlated well with the soil infiltration capacity, where, primary forest and intermediate forest showed a similar and higher Kfs, compared to the young forest. Thus, our results suggested that daily scale hydrologic regime of the intermediate forest more closely resembles that of primary forest than that of a young forest. However, the YF catchment also had the lowest annual rainfall inputs and the highest rainfall variability (PVAR and DAYP0), in addition to a greater land cover heterogeneity (Figure 1), which may have obscured the response of this land cover. While PF and IF have dominant land covers in their respective micro-catchments (100% and 77% respectively), the young forest covered only 68% of the area and the rest of the micro-catchment was covered by more developed land. It has been shown in other catchments that 20% or less forest removal is sufficient to cause substantial effects on peak flow and water yield (Bosch & Hewlett, 1982, Schueler et al., 2009; Evaristo & McDonnell, 2019). Yet, the mean annual low flow and mean annual high flow estimates (Table 3) for YF were of the same order of magnitude to the results reported by Muñoz-Villers and McDonnell (2013) for a 20 yr-old regenerating forest in a neighboring catchment (0.21 and 32 mm day-1, respectively); this suggests that all the studied catchments showed high mean annual low flows compared to previous studies in the region.
In terms of baseflow normalized by total dischargeQbf/Qt , all three catchments presented a high baseflow percentage (Table 4): primary forest (81%), intermediate forest (78 %), and young forest (67 %). This metric also appears to be positively correlated with the percent of forest cover described in the previous paragraph. Similar results of high baseflow dominance during storms were previously reported in this region for an old-growth TMCF by stable isotope-based experiments (Table 3 from Muñoz-Villers & McDonnell, 2012), where groundwater dominated up to 90% of the storm runoff during the wet season. In terms of quick flow and total runoff responses to rainfall, YF and IF were less responsive to storm magnitude compared to PF (Figure 4e and Figure 4g). The high baseflow contribution to total stormflow in PF appeared to be associated with increases in the rainfall-runoff response, as observed in Figures 4g. Similar effects were previously reported for an old-growth TMCF, a 20-yr old TMCF, and a pasture catchment (Muñoz-Villers & McDonnell, 2013). Surprisingly, the peak discharge response to rainfall was very similar for the three forested micro-catchments (Table 5, Figure 4d). These results indicated that the three forest covers presented a seemingly similar capacity to attenuate peak flows. Our work suggests that the secondary forests contributed to restored hydrologic responses to pre-disturbance conditions. Nonetheless, uncertainty remains to confirm this hypothesis, since the micro-catchments were not entirely, nor equally, dominated by the studied forests, as indicated earlier.
We were not able to completely answer the question of whether forest age impacts hydrology, in terms of baseflow maintenance and peak discharge, because differences in slope, elevation and percent of forest cover among the three catchments influenced our results. For instance, the PF micro-catchment has a steeper mean slope than that of IF and YF, being 6° and 12° greater, respectively. Steep slopes have been associated with higher quickflow, and lower lag time and time to peak responses (Mu et al., 2015; Nainar 2018), as observed in our study (Figure 4e-f and Table 4). On the other hand, steeper slopes have also been linked with higher water storage capacity (Karlsen, 2010; Gabrielli & McDonnell, 2011; Sayama et al., 2011; Uchida et al., 2008). Thus, the steeper slope of PF may explain its higher contribution from baseflow and its high rainfall to runoff ratio (0.67). Additionally, the higher mean elevation of PF (1756 m a.s.l.), compared to that of IF (1604 m a.s.l.) and YF (1453 m a.s.l.) seems to correlate with its higher RR.
Our findings showed that 20 and 40 years of natural forest regeneration were sufficient to restore large part of the hydrology on micro catchments previously occupied by intensive pastureland and annual crops (local inhabitants communication). These results clarify that, despite the commonly accepted notion that these forests have high ET rates, both regenerating and primary forests provide important hydrological services, such as dry-season baseflow sustenance and modulation of peak discharge, due to the higher infiltration rates and water storage capacities.