DECISOR SOFT

Average corrected 0-1 m soil C and N stock in pastures ≥8 years were 10 ± 5% (15 ± 7 Mg C/ha) and 12 ± 4% (1.7 ± 0.5 Mg N/ha) higher after forest-to-pasture conversion. However, estimates were highly variable, ranging from -36% (-39 Mg C/ha) to 62% (78 Mg C/ha) for ∆ 0-1 m soil C stocks, and -41% (-3.7 Mg N/ha) to 44% (6.0 Mg N/ha) for ∆ 0-1 m soil N stocks (Table 2.5). The ∆ 0-1 m soil C and N stocks in the two pastures ≤2 years, which were excluded from the comparison of regression lines, were high (C: 49 and 96%; N: 52% for both). Corrected ∆ 0-1 m soil C stocks were correlated with ∆ 0-1 m soil N stocks in our pastures (r = 0.92, P < 0.01) (Figure 2.1c).

On average ∆ 0-1 m soil C and N stock estimates adjusted for potential tool bias in soil bulk density estimates were 21 Mg C/ha and 1.4 Mg N/ha lower,

respectively, than original ∆ 0-1 m soil C and N stock estimates (Tables 2.5 and 2.6). Differences between original and adjusted ∆ 0-1 m soil C and N stocks varied from - 93 to 36 Mg C/ha and -6.5 to 3.0 Mg N/ha compared to original estimates,

respectively. On average, adjusted ∆ 0-1 m soil C stock were -4 Mg C/ha while adjusted ∆ 0-1 m soil N stock indicted no change (Table 2.6 and Figure 2.5).

Nevertheless, adjusted ∆ 0-1 m soil C and N stocks due to pasture-to-forest conversion varied by pasture from -104 to 145 Mg C/ha and -6.8 to 10.5 Mg N/ha, respectively (Table 2.6).

The direction and relative magnitude of original ∆ soil C stocks differed by soil depth for some of our pastures (Table 2.5). While average ∆ soil C stocks in the Tropical moist forest and Tropical Premontane rain forest life zone increased in 0-1 m, they decreased in the 0-30 cm soil layer. For two pastures in the Tropical Lower Montane rain forest life zone ∆ soil C stocks decreased in the 0-1 m soil layer, while they increased in the 0-30 cm layer. In the Tropical Premontane wet forest-warm life zone, ∆ soil C stocks in 0-1 m soils were relatively small compared to 0-30 cm soils.

25 In Tropical dry forest life zone, negative ∆ soil C stocks in the 0-1 m soil layer were relatively large compared to the 0-30 cm soil layer.

The effect of age was not uniform for all life zones and the analysis was sensitive to the exclusion of certain pastures. Generally, ∆ 0-1 m soil C stocks in pastures ≥8 years in the Tropical wet forest life zone decreased with pasture age, but changes with pasture age were less profound or not present in other life zones (Figure 2.4ac). Absolute (original and adjusted) ∆ 0-1 m soil C stock estimates in pastures ≥8 years were unrelated to age (Table 2.2; Figure 2.4a). There was an interaction effect between age and life zone for absolute original ∆ 0-1 m soil N stock estimates (F5,17 =

3.5, P = 0.02; Table 2.2 and Figure 2.4b) while absolute adjusted ∆ 0-1 m soil N stocks were unrelated to age (Table 2.2). There was also an interaction between age and life zone for relative ∆ 0-1 m soil C and N stocks (C: F5,17 = 2.9, P = 0.047; N:

F1,17 = 3.7, P = 0.02; Table 2.2 and Figure 2.4cd). After excluding the oldest pasture

(70 years) in the Tropical Premontane rain forest life zone, the interaction effect was insignificant (C: F5,16 = 1.8, P = 0.16; N: F1,16 = 1.1, P = 0.4) and the effect of age

became similar for all life zones (C: F1,21 = 9.06, P = 0.01; N: F1,17 = 5.66, P = 0.03).

The 70-year-old pasture had higher relative ∆ 0-1 m soil C and N stocks than any of the other pastures >54 years old and was the only pasture >54 years old in higher elevation life zones. Within the same life zone the 70-year-old pasture was not an outlier.

Comparisons of regression lines of pastures ≥8 years indicated ∆ soil C and N stocks varied by life zone (Table 2.2). All life zones were different from three to five other life zones when comparing ∆ soil C stocks in pastures ≥8 years, but we did not conduct this test for 0-1 m soil N stocks due to the interaction between life zone and pasture age (Table 2.5). Original and adjusted soil C stock estimates in most pastures in the Tropical dry forest and Tropical Lower Montane rain forest life zone were lower than in their reference forests. Pastures in all other life zones (with the exception of one pasture in the Tropical moist forest life zone) had higher original soil C stock estimates than their reference forests (Tables 2.5 and 2.6). In contrast, soil C stock

26 estimates adjusted for tool bias were also lower than their reference forests in the Tropical Premontane rain forest life zone (Table 2.6). Original soil N stock estimates were lower than their reference forest in all pastures in the Tropical dry forest life zone and two pastures in the Tropical Lower Montane rain forest life zone; all other

pastures had higher soil N stocks than their reference forests (Table 2.5). In contrast, adjusted soil N stock estimates were lower than their reference forest in most pastures in the Tropical dry forest, Tropical Premontane and Lower Montane rain forest life zones and in one pasture in the Tropical moist forest life zone (Table 2.6). In the Tropical wet forest life zone, original soil C stock estimates were 1.5, and original soil N stock estimates 1.4 times higher due to conversion, which was relative high

compared to other life zones (Table 2.5).

The “best” BIC-ranked model predicting absolute ∆ 0-1 m soil C stocks included the variables age and life zone (Table 2.7a). Additionally, life zone and age predicted ∆ 0-1 m soil C stocks better than temperature and precipitation alone. After excluding the two youngest pastures, the age and life zone model strongly competed with the best ranked life zone model in predicting ∆ 0-1 m soil C stocks (Table 2.7b). The ∆ 0-1 m soil C and N stocks increased with precipitation (C: r = 0.58, P < 0.01; N: r = 0.72, P < 0.01; Figures 2.3e-f), and decreased with age (C: r = -0.65, P < 0.01; N: r = -0.59, P < 0.01; Appendix 9), confirming the comparison of regression lines and BIC model ranking results.