7. Discussion
7.1 Spatially and temporally intensive phenological observations
Compared to most traditional direct observations, this dissertation contributes a unique phenological dataset with three exclusive characteristics: 1) high density observations both over time and space; 2) intensive spring and autumn phenological observations; and 3) an urban woodlot study site. This phenological dataset employs direct phenological observations with very high spatial and temporal resolutions at both the community and species level. Thus, this dataset provides a level of detail relating to phenological development throughout the season that is greater than the more commonly reported date-specific phenophases (Donnelly et al., 2006; Estrella & Menzel, 2006; Schwartz et al., 2006; Xu & Chen, 2013). In addition, the intensive autumn phenological observations and research are in contrast to many phenological studies focusing primarily on spring phenological observations (Badeck et al., 2004; Donnelly et al., 2006;
Schwartz et al., 2006; Caffarra et al., 2011). Finally, the observations were conducted in an urban woodlot, so they are distinct from the observations in rural forests, botanical gardens, or agricultural fields.
This phenological dataset covers the entire durations of both spring and autumn leaf phenology from autumn 2007 to autumn 2012. Traditionally, direct observations
usually record the date or dates on which one or a few specific phenophases occur. For example, phenological data from the European International Phenological Network included the dates of leaf unfolding, flowering, leaf coloring, etc. (Menzel & Fabian, 1999). Similarly, first leaf unfolding data were employed from the China Meteorological Administration phenological network (Xu & Chen, 2013). The first leaf and first bloom of cloned lilac and honeysuckle were used to build spring indices models (Schwartz, 1997). The phenophases used in most phenological studies are recorded on specific dates and are discrete from each other, whereas in this study I used a detailed observation protocol consisting of 24 spring phenophase levels and 8 autumn phenophase levels. These phenophase levels represent the progression of development from buds visible through to leaves fully expanded in spring and from leaf coloration through to leaf fall in autumn, thus providing a spring and autumn phenological profile for each season.
Phenological studies traditionally reported single dates for each phenophase. For example, Matsumoto et al. (2003) used the date of budding (20% buds open) and leaf- fall (80% leaves fall) for Ginkgo biloba L. in Japan from 1953 to 2000, and Gordo and Sanz (2009) employed a plant and animal phenological dataset in Spain for the period 1943-2003, which also reported the date of individual phenophases, such as flowering, leaf unfolding, fruit ripening. Both studies reported a single calendar date at which each phenophase occurred. In contrast to this date-specific record for a particular phenophase, here I subdivided each of six spring phenophases (Liang et al., 2011), and two autumn phenophases into four levels, thus achieving more accurate and detailed phenological information for each phenophase through the use of intensive observations and recordings. Such high resolution data are crucial for detecting relationships between
phenological development and environmental factors, building phenological progression models, and bridging the gaps between ground visual phenological data and remote sensing data. Having a range of data representing each phenophase allows more of the variability around each phenophase to be captured and presented, which results in more accurate modeling and allows direct comparisons with remote sensing data, which is available in a similar timeframe.
In addition to the large number of phenophases observed in this study, the observation frequency was also more intensive than traditional methods. In general, direct observations usually only report the day of the year when critical phenophases occur, whereas in this study the progression of the entire season (spring and autumn) was recorded by taking bi-daily phenological observations. For example, in a study conducted on deciduous trees by Delpierre et al. (2009) in France, phenological observations were conducted weekly and only two leaf coloration phase levels (10% yellow leaves and 90% yellow leaves) were recorded. Similarly, Morin et al. (2009) employed the dates of leaf unfolding from 18 North American temperate tree species at three locations in the United States. Again, high resolution data collection permits greater accuracy for 1) subsequent phenological model building, by portraying the whole progress of phenology and 2) remote sensing data validating, by scaling up from individual tree phenology to community phenology.
Combined with temporally intensive observations, spatially intensive observations also were conducted, in line with other studies conducted by Prof. Schwartz’s group (Liang & Schwartz, 2009; Hanes & Schwartz, 2011; Liang et al., 2011). Spatially
intensive observations can be used to: 1) fully examine community level as well as individual level phenology; and 2) build ground observation metrics corresponding to satellite pixels to fill the gaps between ground visual observations and remote sensing phenology.
Although high resolution data collection was a main focus of this study, the length of the time series was limited due to the nature of a PhD dissertation. Long time series of 30+ years are more common in places like Europe and Asia, where phenological
recording has been ongoing for many years (decades to centuries) (Sparks & Carey, 1995; Walther et al., 2002; Menzel et al., 2005). For centuries, phenology has been observed and recorded by naturalists and widely used for managing agricultural activities, and for presenting the calendar of seasons in both Europe and Asia (Schnelle & Yang, 1965; Zhu & Wan, 1973). Since phenology has been recognized as a useful indicator of recent climate change (Schwartz, 2003; Rosenzweig et al., 2008), the need to identify and establish suitable data sets has become a priority. However, in order to draw meaningful conclusions relating to climate change from phenological observations, the length of the data set is important. The longer the data series, the more confident one can be regarding the relationship between climate and phenology.
European countries and China established nationwide phenological networks in the mid-20th century (Chen, 2003; Menzel et al., 2006) and most phenological studies employed spatially extensive observations (Schwartz & Hanes, 2010; Caffarra et al., 2011; Xu & Chen, 2013) and across study areas, much larger than the ones in this study. There are tradeoffs between observation intensity/density and length/scale in
phenological research. Here, spatially and temporally intensive observations were conducted to further understand the relationship between phenological progression throughout the season (especially autumn phenological development) and environmental factors, thus providing more detailed information than one date per phenophase.