Arctic tundra plant phenology and greenness across space and time

Abstract

The Arctic is warming at twice the rate of the rest of the planet with dramatic consequences for Northern ecosystems. The rapid warming is predicted to cause shifts in plant phenology and increases in tundra vegetation productivity. Changes in phenology and productivity can have knock-on effects on key ecosystem functions. They directly influence plant-herbivore and plant-pollinator interactions creating the potential for mismatches and changes in food web structure, and they alter carbon and nutrient cycling, which in turn influence feedback mechanisms that couple the tundra biome with the global climate system. Improving our understanding of changes in tundra phenology and productivity is therefore critical to projecting not only the future state of Arctic ecosystems, but also the magnitude of potential feedbacks to global climate change. In this thesis, I combine observations from ground-based ecological monitoring, satellites and drones (also known as unmanned aerial vehicles or remotely piloted aircraft systems) to investigate how tundra plant phenology and productivity are changing across space and time, and to test how observational scales influences our ability to detect these changes. Spring plant phenology is tightly linked to temperatures, and advances in spring phenology are one of the most well documented effects of climate change on global biological systems. With rapid and near-ubiquitous Arctic warming, the absence of consistent trends in tundra spring phenology among sites suggests that additional environmental factors may exert important controls on tundra plant phenology. Indeed, further to temperature, snowmelt and sea-ice have been reported to strongly influence tundra phenology. Yet, the relative influence of these three factors has yet to be evaluated in a single cross-site analysis. In Chapter 2, I tested the importance of local average spring temperatures, local snowmelt and the timing of the drop in regional spring sea-ice extent as controls on variation in spring leaf out and flowering of 14 plant species from long-term records at four coastal sites in Arctic Alaska, Canada and Greenland. I found that spring phenology was best explained by snowmelt and spring temperature. In contrast to previous studies, sea-ice did not predict spring plant phenology at these study sites. This contrasting finding is likely explained by differences in the scale of the sea-ice measures employed. While many previous studies used descriptors of circum-polar sea-ice conditions that serve as aggregate measures for global weather conditions, I tested for the indirect effects of sea-ice conditions at a regional scale. My findings (re)emphasize the importance of snowmelt timing for tundra spring plant phenology and therefore highlight the localised nature of some of the key drivers of tundra vegetation change. Discrepancies between conventional scales of observation and underlying ecological processes could limit our ability to explain variation in tundra plant phenology and vegetation productivity. In the remote biome, ground-based monitoring is logistically challenging and restricted to comparably few sites and small plot sizes. Multispectral satellite observations cover the whole biome but are coarse in scale (tens of meters to kilometres) and uncertainties persist in how trends in vegetation indices like the Normalised Differential Vegetation Index (NDVI) relate to in situ ecological processes. Recent advances in drone technologies allow for the collection of multispectral fine-grain imagery at landscape level and have the potential to bridge the gap in observational scales. However, collecting high-quality multispectral drone imagery that is comparable across sensors, space and time remains challenging particularly when operating in extreme environments such as the tundra. In Chapter 3 of this thesis, I discuss the key error sources associated with solar angle, weather conditions, geolocation and radiometric calibration and estimate their relative contributions to the uncertainty of landscape level NDVI measurements at Qikiqtaruk in the Yukon Territory of Canada. My findings show that these errors can lead to uncertainties of greater than ± 10% in peak season NDVI, but also demonstrate they can be accounted for by improved flight planning, meta-data collection, ground control point deployment, use of reflectance targets and quality control. Satellite data suggest that vegetation productivity in the Arctic tundra has been increasing in recent decades: the tundra is greening. However, the observed trends show a lot of variation: although many parts of the tundra are greening, others show reductions in vegetation productivity (sometimes known as browning), and the satellite-based trends do not always match in situ records of change. Our ability to explain this variation has been limited by the coarse grain sizes of the satellite observations. In Chapter 4, I combined time-series of multispectral drone and satellite imagery (Sentinel 2 and MODIS) of coastal tundra plots at my focal study site Qikiqtaruk to quantify the correspondence among satellite and drone observations of vegetation productivity change across spatial scales. My findings show that NDVI estimates of tundra productivity collected with both platform types correspond well at landscape scales (10 m – 100 m) but demonstrate that the majority of spatial variation in NDVI at the study sites occurs at distances below 10 m and is therefore not captured by the latest generation of publicly available satellite products, like those of the Sentinel 2 satellites. I observed strong differences in mean estimates and variation of vegetation productivity between the dominant vegetation types at the field site. When comparing greening observations over two years, I detected differences in the amount of variation amongst years and a within-season decline in variation towards peak growing season for both years. These results suggest that not only the timing, but also the heterogeneity of tundra landscape phenology can vary within and among years, and if lowered by warming could alter trophic interactions between species. The findings presented in this thesis highlight the importance of the localised processes that influence large-scale patterns and trends in tundra vegetation phenology and productivity. Localised snowmelt timing best explained variation in tundra plant phenology and drone imagery revealed meter-scale heterogeneity in tundra productivity. Research that identifies the most relevant scales at which key biological processes occur is therefore critical to improving our forecasts of ecosystem change in the tundra and resulting feedbacks on the global climate system.