MODELOS ANCOVA

Differences of surface soil chemistry were shown on a relatively short timescale on the soil chronosequence and restoration trajectory. The younger Transect 3 soils usually showed greater chemical and biological activity with higher Soil MBC and mineral nitrogen, compared with the older Transect 1. Soil pH was more acidic with soil age, and reached a plateau in the mid-later stage of soil chronosequence. This was mainly due to the acid producti on (organic and inorganic acids) following the decomposition of accumulated organic materials. Similar patterns of soil chemistry, in particular soil pH, had been shown in previous long-term soil chronosequence studies in a similar climate at the Haast chronosequence (Eger et al., 2011); as well as in a million-year timescale chronosequence study in tropical island forest ecosystem in Hawaii (Hedin et al., 2003). Higher soil pH and EC in Transect 3 could also be the result of receiving more sea spray with high basic cations (Whipkey et al., 2000), as it is located closer to the current shoreline. The foliage acts like a collector for the sea spray, rain will wash the deposition down into the soil (this is discussed more in Chapter 6). Inputs of significantly larger amount of Na, Mg, and K were evident in early study (Smith et al., 2016) (Table 5.3); and substantial Na enrichment was also found by Eger et al. (2011). Higher soil pH and MBC probably promoted soil nitrogen mineralization, resulting in higher ammonium-N. Lower nitrate-N in M1 might be due to lower pH which is unfavourable to the nitrification processes (Clough et al., 2004).

In terms of soil microbial biomass P, Turner et al. (2013) indicated its increasing importance to either the soil organic P or biomass P (plant and microbial) alongside the long-term ecosystem development in Franz Josef chronosequence. This partly explains the low soil MBP in the younger M3 in the present study. However, the reasons that MBP concentration was higher in R3 and U3 compared to M3 remained unclear. This could be because of the different vegetation hence different microbial communities between the three plots (Yin et al., 2016; Zak et al., 2003), as R3 and U3 have mixed grass cover with abundant fine grass roots. Unlike in Turner et al. (2013), there was not a further correction of potential contributions of fine leaf fragments and roots to the microbial P measured by the fumigation method in the present study. This could lead to potential overestimation of microbial P in the surface soils. Additionally, it is worth noting that surface soil biological properties, in particular microbial properties are highly sensitive to changing environmental conditions such as temperature and moisture. Thus, results from a one-time sampling event in the present study might not reveal long- term changes in soil biological properties. Long-term study of effects of vegetation restoration and composition on soil microbial P dynamic would be a valuable property to study.

Surface soil P fractions in the present study mainly agreed with the conceptual model proposed by Walker and Syers (1976), as the younger (less weathered) soils at Transect 3 had overall larger soil P

pools than the older (more weathered) soils at Transect 1. Similar results of soil P factions on long- term soil chronosequences and ecosystem development had been reported in many studies (e.g. Crews et al., 1995; Parfitt et al., 2005; Eger et al., 2011; Izquierdo et al., 2013; Chen et al., 2015). On the other hand, the concentration of P (org) and proportional importance of P (org) to P (tot) had been significantly improved by the restoration of native plants, compared to unplanted mixed grassland (Table 5.4 and Figure 5.3). There was not comparable studies re garding forest restoration in this super- humid climate of the West Coast, New Zealand. Similar pattern of soil inorganic and organic P transformations was observed in a 40-year Pinus sylvestris natural revegetation site in North-western Russia (Celi et al., 2013).

However, different results were reported relating grassland afforestation in dry high country areas of Canterbury, New Zealand, indicating that lower organic P concentration in the topsoil of pine stands compared with adjacent grassland (Davis and Lang, 1991; Condron et al., 1996; and Chen et al., 2000). They accounted this for the enhanced mineralization of soil organic P under pine. Differences observations between studies could possibly due to: (i) different degrees of organic material decomposition because of the different ages of restoration or afforestation (De Schrijver et al., 2011; Zhang et al., 2016); (ii) multi-species restoration versus single species afforestation resulting in different strategies of P mining and P dynamics (Oelmann et al., 2011; Rosling et al., 2016); and (iii) different environmental conditions (dry versus wet climate) (Chen, Condron, & Xu, 2008). In the present study, it could be extrapolated that organic P pool in restoration soils would approach the level in the mature forest soil and got stabilised in the future.

Low inorganic P fractions, including primary apatite P (Ca) and secondary mineral P (Fe/Al), in the mature forest soils could be due to more intense weathering, which is promoted by diverse mature vegetation. Released P might have been subsequently transformed into soil organic P, immobilized in the biomass, or lost via leaching and runoff (Hedin et al., 2003). Losses of soil nutrients in more weathered soils were also reflected in lower total Al and Fe at the mature forest plots (Table 5.5). In this scenario, the restoration of native plant species at the oldest restoration site (R1) had substantially promoted the weathering of soil minerals, but these weathered minerals were not lost yet. Thereby, significant high concentrations of both oxalate and citrate-dithionite extractable Al and Fe were found in the R1 surface soils. In the present study, Al and Fe contents in the rocks or gravels were not included in the analyses, but more than half (50%-75%) of surface materials were gravels at the younger Transect 3 (see profile description in the Appendix D.1). This could possibly explain that lower concentrations of total Al and Fe found at the younger Transect 3, compared to the older Transect 1. Although soil occluded P fractions were not significantly different between plots, they still become slightly more important as soils age (from Transect 3 to Transect 1). This is probably because of the

97 relatively small soil age gap between transects at the present study. However, occluded P fractions were proportionally less important in restoration plots, compared to mature and unplanted plots at both transects (Figure 5.3). This is partially attributed to promoted P transformation by restored vegetation, as reflected in increased importance of organic P. Similar results were also observed in a 300-yr post-landslide tropical forest development at Puerto Rico (Frizano et al., 2002). They attributed this to the potential release of P from occluded pool by soil biota. However, Zhang et al. (2016) found an increase in occluded P fraction in the middle (90-yr old) forest successional stage, but decreased to the late (ca. 400-yr old) successional stage at South China. They indicated, although increased soil microbial activity in the middle successional stage had promoted the release of inorganic P into soil solution, but might quickly be precipitated by Al and Fe mineral in the favourable pH environment of tropical soils and become occluded.