4.3 Tree- and watershed-level estimates of transpiration
Sapwood area explained nearly 85% of the variability in daily tree-level transpiration (Ts ) (in the monitored trees) across all zones, with tulip poplar generating higher than predicted transpiration for its diameter and sapwood area in the riparian buffer and mid-hillslope (Figure 3). In contrast, loblolly pine had lower than predicted transpiration for its sapwood area in the mid-slope and upland-slope. Unlike species J sthat was related to zone, an effect test revealed thatTs rates were significantly correlated to sapwood area and species but not to zone (sapwood – F ratio 122.8, P = <0.0001; species – F ratio 9.2, P = 0.0076; zone – F ratio 2.1, P = 0.20). This is not surprising as Ts is closely linked to and driven largely by sapwood area (Wullschleger and Norby, 2001).
Annualized zone weighted watershed-level transpiration (T w) was very similar from a dry year to a wet year (361 mm vs. 370 mm, Table 4) in part becauseJ s was not significantly different for loblolly pine, white oak, and tulip poplar between years. These three species occupied 50% of the sapwood area and were responsible for almost 75% of the water loss in both years (i.e., 73% in 2015) and (i.e., 76% in 2016). Oishi et al. (2010) also found similarities or small variations in transpiration across wet and dry years, 329 mm to 349 mm over four years. Hawthorne and Miniat (2018) also found that species composition may explain the similarity between T w between wet and dry years.
There were differences in scaled T w, depending on which scaled T z was used to represent tree water use (Table 4). We compared Tz based on scaledJs data from the buffer, mid-slope, and upland-slope against Tw based on weightedTz from all three stations to determine their relative differences. Riparian buffer zone estimates of annual water loss were 24% higher in 2015 and 21% higher in 2016 thanT w based on weighted Tzfrom all three stations. Annual water loss from the mid-hillslope was only 2% higher in 2015 and 2016 than T w based on weighted Tz from all three stations while water loss from upland-hillslope was 7% lower in both years thanT w. The riparian buffer and upland-hillslope produced larger percent differences than mid-hillslope likely because they did not fully capture the spatial changes in speciesJ s response to VPD as well as the range of watershed soil moisture conditions and stand structure across the watershed (Mitchell et al., 2012). In addition, the hydrologic controls on tree J s in these two zones likely shifted from climatic factors to soil moisture as VPD increased, and soil moisture rapidly decreased (Emanuel et al., 2010). The riparian buffer and upland-hillslope zones also had larger amplitudes for drainage and soil moisture compared to the mid-hillslope. The coefficient of variation in soil moisture was 24% in the mid-hillslope, 32% in the riparian buffer, and 48% in the upland-hillslope. If we assume weighted Tz across the three zones offered the best estimate ofT w, then it appears Tz from the mid-hillslope was within a reasonable range to determine annualT w. J s-scaledT w from the buffer and upland-hillslope resulted in broader estimates of T w than mid-hillslope and should probably not be used as standalone zones to determineT w.