|dc.description.abstract||Rakiura Māori (New Zealand’s southern-most group of indigenous peoples) have harvested the chicks of burrow-nesting sooty shearwaters (Tītī; Puffinus griseus) for generations. As part of the harvest process some families have maintained annual harvest diaries, some of which date back to the 1950s. Throughout this thesis I investigated how these diaries could be used to determine if data obtained from the sooty shearwater harvest can predict shifts in the Southern Oscillation Index (SOI). To answer this question, I had to first calculate indices of the harvest which take into account human effort. Then I had to determine if those indices could in fact predict upcoming values of SOI. I next determined what oceanographic factors explained at-sea behaviour of sooty shearwater adults, and if those variables also explained variation in the harvest indices. I finally determined if there were any relationships between oceanographic variables and upcoming values of SOI.
In Chapter 2, I used generalised boosted regression models, a machine learning algorithm, to calculate a harvest index that takes into account factors which could impact the numbers of birds taken on any given hunt. The r2 of predicted versus observed values were between 0.59 and 0.90 for the nanao (first half of the season, when chicks are harvested from burrows during the day) and 0.67 and 0.88 for the rama (second half of the season, during which chicks are harvested from the surface at night). Exploration of the controlling factors of the models reveals that ‘day of season’ plays an important role in predicting daily harvest during the second half of the season (the rama). The nightly tally in the rama peaks approximately half-way through (10 – 15 days in), which is likely related to the timing of birds emerging from burrows to fledge. The models also suggest that data from the rama (when chicks are 100 – 120 days old) may be the most suitable for long-term monitoring of populations of sooty shearwaters due to consistencies in calculated harvest indices between diaries. Nanao harvest indices, though less consistent, showed similar patterns to those of the rama. When compared to the harvest indices calculated by general linear models by Clucas et al. (2012) I found that the agreement between both indices was r =0.56 and r =0.77 for the nanao and rama, respectively. Although harvest indices have been created in the past, this represents a re-analysis of these data with a different statistical technique, plus a new diary to test the relationships under new circumstances. The use of machine learning to correct for extraneous factors (e.g., hunting effort, skill level or weather) and create standardised measures could be applied to other systems such as fisheries or terrestrial resource management. The harvest indices calculated in this study were then used to examine relationships between chick quantity, quality and Pacific Climate indices.
In Chapter 3, I demonstrate that shifts in sooty shearwater (Puffinus griseus) chick size and abundance occur in advance of shifts in the climate indices used to indicate upcoming El Niño Southern Oscillation events. I used indices of the harvest calculated from “muttonbirding” diaries collected by the Rakiura Māori of New Zealand from 1957 to 2010. I compared these indices to SOI (pressure differential between Tahiti and Darwin), Niño4 Index (sea temperature anomaly in Eastern equatorial Pacific), Oceanic Niño Index (ONI; sea temperature anomaly in the central equatorial Pacific), and Pacific Decadal Oscillation (PDO; a decadal shift in Northern Pacific sea temperatures which affects El Niño). Spearman correlations showed that La Niña events tended to occur after those harvest seasons with relatively high success and chick size, whereas El Niño events tended to occur after harvest seasons with relatively low success and chick size. Values of SOI and ONI before the harvest were highly correlated with PDO values averaged for two years before the harvest. Mean SOI values one year before the harvest tended to be higher when two year mean PDO values were high. Similarly, ONI tended to be lower when two year mean PDO values were high suggesting a complex interaction between the three indices and the harvest data. Generalised boosted regression models show that chick size alone is able to predict shifts in SOI from 0 – 12 months after the harvest, and can also predict ONI occurring 0 – 12 months and 14 – 24 months after the harvest. A model that included chick size, two-year average PDO (prior to the harvest), and one year averages of SOI and ONI (prior to the harvest), was able to predict shifts in SOI 13 – 20 months after the harvest, but not able to predict ONI. The relationships shown here closely follow those found in previous studies however they are now confirmed using multiple diaries. Similarly, the finding that chick size predicts shifts in SOI is novel and the strength of the relationship is high. I conclude that the birds are being affected by oceanographic conditions that are also precursors to shifts in SOI and ONI, and that there is a complex interaction between PDO prior to the harvest, the harvest indices and SOI. The location and timing of adult birds at the time they are provisioning chicks leads to potential mechanisms. In order to pinpoint these mechanisms, I first sought to determine which oceanographic features influence the adult at-sea behaviour, and the harvest indices.
A few studies have examined the use of the sooty shearwater (Puffinus griseus) harvest as a way of monitoring the Southern Ocean ecosystem, however we have yet to determine the particular physical oceanographic factors that explain annual variation in the harvest. In Chapter 4, I used a step-wise approach with generalised boosted regression models. I first explored the relationships between oceanographic factors and broad at-sea behaviours including foraging trip time and foraging site choice. I then computed the core foraging areas in the nearshore and offshore regions for sooty shearwaters based on archival geolocation tag data obtained from Shaffer et al. (2006) and determined which particular physical oceanographic characteristics defined this region. Finally, I tested if oceanographic conditions in the foraging areas could explain the variation in the harvest indices. Total trip time was the behaviour that was predicted best, with a Pearson’s r value of 0.56; other behaviours (time en route to foraging, time foraging, and time returning from foraging) had r values of 0.05, 0.38 and 0.20 respectively. Important predictors of the total time at sea model were significant wave height and wind speed en route and returning, charnock parameter (a constant representing surface roughness) and surface pressure within the foraging regions. Models of foraging sites (delineated by the 50% kernel density) performed well, with an area under the receiver operating characteristic (ROC) value > 0.95. For both nearshore and offshore models, sea surface temperature was the best predictor. For the offshore models, wind differential en route to foraging areas was found to be important. Physical predictors of the chick size harvest index were found to be most important in the sub-Antarctic water region, where significant relationships with wind speeds, and significant wave height were identified. This result is likely related to findings from the behavioural models in that turbulent conditions are altering the foraging behavior of adult birds, which has an effect on the size and quantity of chicks. Winds have been shown to affect sooty shearwater flight in past studies however, this chapter identifies oceanographic characteristics that have colony level effects. By understanding the relationships between these oceanographic characteristics and the harvest indices it may be possible to build an understanding of what controls population levels of sooty shearwaters around New Zealand, and help to highlight them as monitors of ecosystems. The information here was then used to perform a targeted study on the oceanographic conditions that not only affect sooty shearwaters, but also predict upcoming values of SOI.
Although analysis has shown that sooty shearwater (Puffinus griseus) harvest data are able to predict El Niño events by approximately 4 - 14 months, it is unclear what oceanographic parameters that affect sooty shearwaters are the mechanistic cause for prediction of El Niño events. However, using information from Chapter 3 (i.e., when the data are able to predict SOI), and Chapter 4 (i.e., oceanographic regions which are important to sooty shearwater adults in the breeding season), it is possible to perform a focused study on potential connections between the harvest data and SOI. In Chapter 5, I examined the relationship between fluctuations in the time series of chick size index of sooty shearwaters calculated in Chapter 3, oceanographic conditions in the South Pacific Ocean region around New Zealand during the breeding season, and values of Southern Oscillation Index (SOI) from 24 months before and after the peak chick size (March). Gridded spatial models of values of SOI show that from 1 to 12 months after March, the oceanographic regions which best explain variation in the time series are Southeast of New Zealand along the Polar and sub-Antarctic fronts and in the southern regions of the sub-Antarctic water zone. From 12 to 24 months after peak chick size, the region which best explains variation in the time series is the Tasman Sea between Australia and New Zealand. Lagged Spearman correlations of oceanographic parameters show that within the sub-Antarctic water and core foraging areas of sooty shearwaters, positive significant relationships exist with wind speed, significant wave height and charnock (a measure of ocean surface roughness) parameters, on the same time scale as the chick size index. A model that combined the significant parameters from both offshore regions and the nearshore foraging area of sooty shearwaters, had a Pearson’s correlation of r > 0.8 for SOI values from 0 to 14 months after peak chick size. It seems that a combination of parameters and areas best explain the variation in the SOI data, however the most important variables are those that represent general turbulence in the sub-Antarctic water and Polar front regions (i.e., wind speed, and significant wave height).
The relationships between SOI, and sooty shearwaters (and their chicks) highlights this species as more than just a monitor of Ocean ecosystems, but also as a predictor of important climate events. Sooty shearwaters, like all seabirds, integrate oceanographic resources in such a way that colony level (e.g., chick size and quantity) effects are apparent when changes occur in Ocean systems. These Ocean changes could reflect precursor events to large scale climate shifts. In this case, a poorly understood region (the Southern Ocean), has been identified as having links to the formation of El Niño and La Niña events by following sooty shearwaters from the colony (via chick size and quantity) to breeding season foraging regions. This thesis has not only identified a potentially important region for the formation of El Niño, but has also created an index that may be able to predict El Niño events nearly two years before they occur, as is the case for the predicted upcoming 2014 event. In the future, lessons that are learned from sooty shearwaters can be applied to climate modelling efforts to help further examine potential mechanisms of the El Niño Southern Oscillation, which affects weather patterns around the world.||