6.1. Climatic Interpretation
The January (= summer) temperature estimates (Figure 5
b) generated by transfer functions based on a modern pollen rain-temperature data set [8
] need to be interpreted cautiously, since they are based on the current disposition of vegetation. We, therefore, refer to them as ‘temperature equivalents’ since they indicate the January temperature the pollen assemblages suggest if the same type of climatic regime prevailed.
The early (17,000 to c. 12,500 BP) pollen assemblages of the two lowland site profiles are typical of areas with abundant exposed rock and tundra vegetation. The only modern analogues are high elevation sites above 500 m where peat does not form and cold temperatures (January temperatures are ≤6 °C in the tundra zone), frost heave, and high winds induce stunted vegetation cover. The pollen transfer function reconstructions for the early part of the macrophyllous forb stage corroborate this estimate and show no pronounced shift in temperature until about 12,000 BP (Figure 5
b). Tundra, therefore, covered the island surface from the sea level to at least 120 m elevation.
A transition from macrophyllous forb tundra to grassland-shrubland began c. 14,000–13,000 BP and was associated with a trend toward wetter conditions in the two bog sites, but not the well-drained HSP site. At HSP, first Coprosma
shrubland and then Myrsine divaricata
rose to dominance, with shrub cover peaking at c. 11,000 BP. This may be taken for the more general trend on well drained slopes. Phlegmariurus varius
, which is a clubmoss common in low shrubland, and Hymenophyllum
, which is a filmy fern more abundant in the current pollen assemblages above c. 100 m [24
], were prominent at all sites over this period (Figure 5
f). High percentages of Myrsine
pollen on the present landscape are inevitably associated with high percentages of Dracophyllum
]. Therefore, the absence of the latter is significant, and indicates that forest climates have not been established. The major increase in shrubland noted at HSP was only weakly registered at the two bog sites. Equivalent January temperatures are estimated to have risen by c. 1 °C shortly after 12,000 BP. Between 11,000 and 10,500 BP, equivalent January temperatures were within 1 °C of the current temperature. The timing of this warming c. 12,000 BP is consistent with the end of the Antarctic Cold Reversal cooling event between 14,700 and 13,000 BP [39
The grassland-wetland stage (present at all sites between 10,000 and 9000 BP) is characterised by the lowest percentages of shrub types in the post 12,500 BP period. The shrubland associate Phlegmariurus varius
ferns continued to be well represented. The reassertion of grassland cover and decline of shrubs is consistent with a lowering of elevational vegetation zones. The pollen transfer functions support such an interpretation since they show an equivalent temperature decline of c. 1 °C. The wetland index peaked, which could also be related to less evapotranspiration under cooler climates. However, vertical peat growth slowed at all three sites to a minimum and peat humification increased (Figure 5
c–e) and this suggests an alternative explanation.
Vertical peat growth is determined by the balance between vegetation productivity and breakdown [40
]. Most peat decay takes place in the few centimetres of the upper aerated acrotelm. If peat vertical growth was a faithful proxy for peat mass accumulation into the unaerated catotelm, steady peat accumulation would result in a linear relationship between peat height and age. However, compaction and slow anaerobic decay with increasing age and depth in the profile reduces the apparent vertical annual increment and upper peat layers. Therefore, they usually show an apparent acceleration of vertical growth [41
]. Normally, a wetter peat will grow faster, all things being equal, because organic matter spends less time in the aerated acrotelm, which is thinner because of the higher water table. However, if vegetation productivity is low, even though the acrotelm is thin, the peat spends a greater amount of time within it, and, thus, decomposes more. Moreover, if winters and overnight temperatures are warm, peat breakdown increases as higher temperatures favour accelerated loss of organics [42
]. Therefore, although the pollen sequences and temperature transfer functions developed from them, at first sight, seems to support a cooler summer climate, we prefer an interpretation based on changes in cloud cover and wind direction.
The peat surfaces became wetter after 13,000 BP, and much wetter between 10,500 and 9000 BP, and peat grew slowly (Figure 5
c,d). Atmospheric humidity across the whole period (12,500 to 9000 BP) must have increased, as the filmy fern Hymenophyllum
that is poorly represented in the modern pollen rain below 100 m [24
], but common in the cloudier uplands, was abundant at all sites (Figure 5
f). Rather than temperature decreases being the driver of these ecological changes, increased cloud immersion with accompanying low light, increased humidity and decreased transpiration, and wetter substrates [43
] may have restricted shrub growth. Observations made on cloud bases over Campbell Island in the Perseverance Harbour basin [10
] show that cloud immersion has a linear trend with altitude (Figure 6
Wind direction has a strong effect on cloud base and mean daily temperature (Table 3
). When prevailing winds were west to southwest, cloud bases were high (all peaks were clear in almost 60% of observations) and temperatures were cool (5.0 °C) (Table 3
). Winds from the north-west quarter had cool temperatures but low cloud bases (peaks clear only in 21% of observations). Winds from the north and east led to cloud bases being low but temperatures were 3 °C warmer. Low cloud cover decreases warming during the day by reflecting solar radiation, but prevents heat loss overnight. The combination of northerly source air and prevention of overnight heat loss leads to markedly warmer temperatures. As plant growth depends on bright light and daytime warmth, the same daily mean temperature can have radically different consequences. Cloudiness at these latitudes, therefore, will not cool the land relative to the ocean but will suppress woody plant growth. Mild winters, in particular, are, paradoxically, a physiological stress factor for woody plants in oceanic environments since they break dormancy and encourage respiration and growth during a season when ground water levels are high and light levels are minimal [44
]. For instance, Dracophyllum
growth on Campbell Island is substantially reduced during warmer than normal winters [45
]. In turn, loss of woody cover decreases evapotranspiration and increases soil moisture. This feedback cycle further reduces the suitability of soils for shrubs and encourages the spread of graminoids and cushion plants. Consistent with our interpretation is the lack of any resurgence of a cool climate and tundra indicators such as the macrophyllous forbs. If the early Holocene wetland-grassland phase represents a cloudiness and soil moisture induced depression of the limit to shrub-low forest on the island, it may have been substantial. The difference in the tree line altitude across the middle of the Southern Alps in New Zealand is c. 200 m, the cloudier, western slopes have lower elevation tree lines [46
]. In terms of mean annual temperature, this is equivalent to more than a 1 °C depression, which shows that cloudiness could potentially offset a warming of this magnitude.
The near simultaneous expansion of Dracophyllum
scrub across our sites argues for an abrupt change in the climate regime and a nearly 2 °C increase in equivalent January temperatures is reconstructed over the next 1000 years. However, the adjacent ocean was cooling from 9000 BP onward after an early Holocene peak [9
]. Therefore, as the cooler equivalent temperatures reconstructed for the previous period were largely induced by cloudiness, it follows that the subsequent ‘warming’ reflects a reversion to clearer skies with increased southwesterly winds and, therefore, a real cooling of mean annual temperatures over the island. The establishment of Dracophyllum
trees (as distinct from Dracophyllum
shrubs) at the lower HSB site cannot have occurred before 9000 BP and we suggest not until 6200 BP when the grass and Hymenophyllum
percentages fell to levels consistent with present day pollen rain from forested sites (Figure 5
f). The timing is similar at the higher elevation MHB site. The Campbell Island fossil wood occurrences support this interpretation [48
]. Buried Dracophyllum
wood was collected along an elevational transect 65 to 100 m above the shrub line at c. 210 m in sheltered Southeast Harbour 3.5 km to the southeast of MHB, with no wood before 6000 BP.
The vegetation consequences of the interaction between wind and clear skies is complex. Although shrubs, and Dracophyllum
in particular, become common at all three sites. They are not well aligned and each Dracophyllum
curve has a different trajectory over the Holocene (Figure 4
). The two lowland sites (HSP and HRB) achieve high Dracophyllum
percentages at 6200 BP and, from then on, show marked fluctuations—especially the HRB—but no overall trend. In contrast, the upland MHB site after high values between 7500 and 5500 BP, undergoes a fluctuating decline in response to grassland expansion. A sharp rise in monolete fern spores at 3000 BP at this site shows winds may have increased in strength in the late Holocene (Figure 4
) since ground ferns are likely to be favoured by reduced competition from wind-stunted shrubs and grasses. A similar trend is seen in the late Holocene at Rocky Bay [12
These differing Holocene long term trends in Dracophyllum between the upland and lowland sites are yet a further indication of how the separate components of the climate system can have markedly differing effects on sites depending on their altitude and exposure. The pollen-based temperature reconstructions (which, after 9000 BP, likely reflect actual January temperatures) do not show, in either case, a marked Holocene trend nor do sea surface temperatures in the adjacent ocean. Dracophyllum is, therefore, most likely responding directly to wind speed and bright sunshine hours. The upland MHB site lies in a saddle exposed to winds from the north-western and south-western sectors, while the lowland HSP site is sheltered. Any increase in southern sector winds is likely to subject MHB to strong, cold winds and reduced Dracophyllum growth, whereas HSP will experience clearer skies and increased growth.
Evidence for a direct wind effect comes from the absence or decline in the course of the Holocene of shrubs in the most wind-exposed sites on Campbell Island. West-facing Hooker Cliffs (90 m asl) north of the island, never had significant shrub representation. The south-facing Rocky Bay cliff-top (130 m a.s.l, c. 6 km west of MHB) lost its initial Coprosma
shrub cover in the course of the Holocene [12
]. The fossil wood transect from Southeast Harbour lacks wood between 2300 and 1000 BP, which coincides with an interval with wind-blown silt and stones in nearby cliffs [49
] and a marked decline in Dracophyllum
pollen at MHB, and is attributed to an episode of intense windiness [48
]. Increased windiness is recorded in sediments from an Auckland Island fiord over the same interval [50
]. We attribute this pattern to the increased prevalence of southwesterly winds bringing clearer skies and, thus, favouring woody growth in sheltered areas but stronger, cooler winds eliminate woody communities on exposed sites.
While, from c. 5000 BP, the general Dracophyllum
trend is strongly negative at MHB and positive at HSP. There is more alignment from 2000 BP. Both sites show a Dracophyllum
peak at or around 2000 BP, low centered on 1500 BP, and peak values between 1000 and 600 BP, and a decline thereafter. The periods of falling or low Dracophyllum
values are well aligned with glacial advances in the Southern Alps of New Zealand, where there are two groups of closely spaced advances between 2000 and 1400 BP, and 700 and 200 BP [51
]. These are broadly consistent with the timing of the Dark Ages Cool period (DAC: 1550–1250 BP), the Mediaeval Climate Anomaly (MCA: 1150–750 BP), and the Little Ice Age (LIA: 650–100 BP).
6.3. Southern Ocean Context
There are only a few terrestrial peat records in the southern ocean region with which to compare those from Campbell Island. The Auckland Island group 150 km to the north lies in the same ocean waters and has a similar vegetation cover [5
]. While the chronology is not as secure as in the Campbell Island peats, the general pattern is the same, with the major trees, Dracophyllum longifolium
and Metrosideros umbellata
, not abundant until after c. 9000 BP. Acceleration of vertical peat growth after 6000 BP at two sheltered sites at Dea’s Head on the main island, and restriction of shrubs and small trees after ca. 9600 BP by intense windiness on the exposed Enderby Island site [5
], supports the inferences made for Campbell Island. Evidence from wind-blown minerogenic input on Stewart Island immediately south of the New Zealand mainland points to stronger westerly winds after 5500 BP [54
] as does isotopic and stratigraphic evidence from a lake in the far south and in the wind shadow east of the main ranges [55
]. Further north on the New Zealand mainland, there are indications that tree lines were similarly inhibited by weaker westerly airflow. In the southeastern and central South Islands, tree lines did not reach current elevations until well after 9000 BP [56
] and, in the northwestern South Island, not until 9500 BP [57
A number of peat sites in Patagonia, Tierra del Fuego, and the Falkland Islands group between 55 and 51° S have vegetation records comparable to those of Campbell Island [58
]. The initial timing of peat growth in the higher latitudes of Patagonia and Tierra del Fuego is similar to that of Campbell Island, commencing between 17,500 and 12,000 BP. The vegetation transitions from steppe-moorland to Nothofagus
parkland by 12,000–11,000 BP, and then to closed Nothofagus
forest during the early to mid-Holocene. Of particular relevance to the Campbell Island sequence is the late arrival of closed Nothofagus
forest at upland sites close to the tree line. At Paso Garibaldi (54°43′ S, 500 m a.s.l), a closed forest was not present until after 8700 BP [64
] and, at Las Contornas mire (54°41′ S, 420 m a.s.l), not until 6500 BP [63
]. At Port Howard, Falkland Islands (51°20’S, 100–130 m a.s.l), the current vegetation of low shrubs did not replace herbaceous associations until 8000–7000 BP [66
]. These patterns are confirmed in a recent review of vegetation change in Patagonia [68
], which shows Nothofagus
abundance not reaching current levels until 7000 BP in central Patagonia and 4000 BP in southern Patagonia.
The climatic explanation for the late establishment of forest or shrubland in southern South America differs from that given here for Campbell Island. In southern South America, SSTs were as warm or warmer than by 12,000 BP. However, weaker westerly flow brought more arid conditions and, thus, prevented the Nothofagus forest expanding, while, in the New Zealand sub-Antarctic regions, mist and low cloud associated with weaker westerlies suppressed the taller woody vegetation. Expansion of forest in the early to mid-Holocene in both regions was ultimately driven by increased windiness. However, in southern South America, rain-bearing westerlies delivered more precipitation and were pushed further inland, while, in the New Zealand sub-Antarctic regions, the south-westerly to westerly orientation of the wind flow brought higher cloud-bases and more spells of bright sunshine.
Recent tree line and scrub coverage on Campbell Island has been largely unresponsive to regional climate change. Around the turn of the 19th century and for several decades after, SSTs in the immediate region were as much as 0.5 °C below the 20th century average rising to 0.5 °C above 1970 to 1990 AD [69
]. Although there has been vigorous regrowth of forest and shrubland across grassland induced by fire during the late 19th to early 20th century farming era [34
], there has been no indication of elevational shifts in response to temperature [35
]. This is consistent with our findings that past temperature shifts alone in these highly oceanic situations are insufficient to alter tree line dynamics, which has been noted for the mountains of mainland New Zealand [3