Spring phenological variability promoted by topography and vegetation assembly processes in a temperate forest landscape

Highlights

• Land Surface Phenology of 35⁺ years of Landsat observations revealed fine variation.

• Greenup patterns related to complex relationships between topography and vegetation.

• Topographic position relative to valley bottoms was strongest driver.

• Foundational species white oak explained most species-level variation in phenology.

Abstract

Plant phenological processes significantly impact ecosystem function and services across multiple ecological scales and are widely viewed to be among the most sensitive indicators of global environmental change. Remote sensing has crucially expanded our understanding of phenological variability. Yet, we continue to lack a complete mechanistic understanding of phenology and its variability and drivers, which is important to the development of predictive models, especially under continued environmental change. We combined field inventories and Land Surface Phenology (LSP) approaches, using 36 years of Landsat phenological observations, to characterize the degree to which long-term spring greenup patterns are shaped by topography, vegetation, and topographically structured vegetation assembly processes within a dissected forest landscape in southeastern Ohio. We found temporal and spatial variability among the field samples where greenup patterns displayed rapid change (18 total days) over relatively short distances (<1 km). Slope position explained the most variation (35%), where the bases of hills displayed the latest timing in spring greenup. However, we found that differences in terrain aspect and slope influenced canopy diversity, height, and composition of forest stands, influencing plant community processes that support heterogeneity in spring leaf-out timing. Understanding how forest phenology is shaped by direct and often complex interacting processes that influence the distribution of species assemblages supports new insight into phenological variability and, importantly, the management of forest ecosystems facing continued environmental change.

Introduction

Vegetation phenology significantly influences diverse ecosystem processes across multiple ecological scales, from local organismal interactions (Halupka and Halupka, 2017; Heberling et al., 2019; Royo and Stanovick, 2019; Singer and Parmesan, 2010; Visser et al., 2004) to global biogeochemical cycles (Morisette et al., 2009; Richardson et al., 2013). Vegetation phenology is also among the most sensitive indicators of global environmental change, supporting improved understanding of climate change impacts on terrestrial ecosystems (Fitter and Fitter, 2002; Morisette et al., 2009; Root et al., 2003). As a result, vegetation phenology is increasingly studied, particularly in light of advancing remote sensing technologies that repetitively resolve phenological patterns, aka, Land Surface Phenology (LSP), across multiple spatial and temporal scales (Cleland et al., 2007; Morisette et al., 2009; Nagai et al., 2016). However, despite increased research capacity, we continue to lack adequate mechanistic understandings of vegetation phenology and its variability and drivers (Chmura et al., 2019), limiting the development of predictive models in the face of continued environmental change (Basler, 2016; Richardson et al., 2013).

The timing of spring leaf development among temperate forest ecosystems has been shown to be particularly sensitive to temperature (Linkosalo et al., 2006; Polgar et al., 2014; Polgar and Primack, 2011; Vitasse et al., 2009). At large spatial scales, phenological patterns remain largely consistent with climatic variation across broad latitudinal and elevational gradients (Fitzjarrald et al., 2001; Hopkins, 1918). However, spatial patterns often reveal significant fine-scale variation as well (Elmore et al., 2012; Fisher et al., 2006; Liang et al., 2011; Melaas et al., 2013). Thus, phenological patterns likely also incorporate a variety of possibly interacting local processes (Chmura et al., 2019). For example, local topography influences microclimatic variation, such as processes contributing to cold air drainage into small valleys, which has been shown to delay phenological timing of forest vegetation in comparison to surrounding uplands (Fisher et al., 2006; Schuster et al., 2014). Individual tree species, which also frequently sort according to local topography (Desta et al., 2004; Hix, 1988; Hix and Pearcy, 1997; Martin et al., 2011), can display large interspecific variation in phenological timing as well (Delpierre et al., 2017; Denéchère et al., 2019; Lechowicz, 1984; Murray et al., 1989; Richardson et al., 2006). There also exists some variability in leaf phenology that can be explained by differences in tree height (Osada and Hiura, 2019; Seiwa, 1999), likely serving as a proxy for both individual tree response to site conditions (Bassow and Bazzaz, 1998) as well as differences between juvenile and adult age classes (Augspurger and Bartlett, 2003; de Souza and da Costa, 2020). Because spatial variability likely incorporates a greater range of direct and indirect processes at more local scales, as opposed to more direct climatic forcing across large regions, models at these scales should account for the combined effects of local topography on the composition of forest vegetation and how these factors interact to influence phenological behavior.

Significant efforts have identified the dual importance of topography and forest stand characteristics in explaining variation in LSP behavior (Isaacson et al., 2012; Kraus et al., 2016; Liang et al., 2011; Misra et al., 2018; Reaves et al., 2018; Xie et al., 2015). However, few studies have interpreted spatial variation in light of interactions between topography and forest vegetation composition (Reaves et al., 2018; Xie et al., 2015). First, many LSP studies continue to incorporate generalized models of broad forest types in lieu of detailed species-level information resolving the ecophysiological variation of forest stands (Isaacson et al., 2012; Xie et al., 2015). Second, modeling efforts should also ideally employ statistical procedures that account for the often causative and correlative factors among spatial drivers (Bassow and Bazzaz, 1998; Schemske and Horvitz, 1988). For example, oak (Quercus spp.) dominated forests among dissected landscapes of the Central Hardwood Region of the United States often display strong topographically-mediated vegetation gradients that underlie variation in the fundamental ecological strategies of component tree species (Fralish, 2003). Here, drought-tolerant oaks dominate drier site conditions on southwest-facing hillslopes and ridgetops and transition to mixed mesophytic species assemblages, adapted to withstand increased competition, on opposing mesic northeast-facing hillslopes and bottomlands (Adams et al., 2019; Desta et al., 2004; Martin et al., 2011). Understanding how these environmental and organismal processes (and interactions) contribute to phenological patterns could help improve forest management practices in supporting plant diversity in light-structured environments (Heberling et al., 2019; Royo and Stanovick, 2019), especially in predicting the potential consequences of anticipated environmental change (Iverson et al., 2019b; McEwan et al., 2011).

We designed a study to determine how variation in the timing of spring greenup is explained by topography and vegetation assembly processes across a topographically complex and diverse forest landscape in southeastern Ohio, USA. We used LSP methods, combining multi-temporal Landsat observations from the years 1984 to 2020 (including an 8-day temporal resolution for a given sensor among overlapping scenes), to retrieve long-term average dates of spring greenup across a series of field inventories. Phenological curves fitted to multidecadal Landsat observations were used to resolve long-term climatological average spring greenup dates, insensitive to annual climate anomalies (Fisher et al., 2006; Melaas et al., 2013) and at spatial resolutions (30 m) complementary to local management activities (Adams et al., 2019; Iverson et al., 2017). Field inventories elucidated variation in tree species richness and community mean wood density (WD), the ratio of oven-dry mass to total green volume. We used WD to collectively represent variation in assemblage and functional composition of individual tree stands (Chave et al., 2009; Stahl et al., 2013). Finally, data from digital elevation models and LiDAR were used to quantify key topographic variables (elevation, slope percent, and aspect) and mean vertical canopy height.

Our layered approach used regression modeling and variance partitioning procedures to determine the extent to which site-to-site variation in spring greenup is explained by topography, vegetation, and topographically-structured vegetation assembly processes. Next, we incorporated path analysis to precisely determine how topographic features influence phenological variation by mediating changes in forest stand vegetation. Finally, we isolated the relative importance of individual species in explaining forest stand phenological variation, based on relative dominance profiles and contributions towards community-level functional composition. Together, we illustrate how spring leaf phenology is shaped by direct and often complex interacting processes across tightly integrated forest ecosystems.