University of Nebraska - Lincoln DigitalCommons@University of Nebraska - Lincoln USGS Staff -- Published Research US Geological Survey 2012 Landscape controls on the timing of spring, autumn, and growing season length in mid-Atlantic forests Andrew J. Elmore University of Maryland Center for Environmental Science, [email protected] Steven M. Guinn University of Maryland Center for Environmental Science Burke J. Minsley United States Geological Survey, [email protected] Andrew Richardson Harvard University, [email protected] Follow this and additional works at: https://digitalcommons.unl.edu/usgsstaffpub Part of the Earth Sciences Commons Elmore, Andrew J.; Guinn, Steven M.; Minsley, Burke J.; and Richardson, Andrew, "Landscape controls on the timing of spring, autumn, and growing season length in mid-Atlantic forests" (2012). USGS Staff -- Published Research. 514. https://digitalcommons.unl.edu/usgsstaffpub/514 This Article is brought to you for free and open access by the US Geological Survey at DigitalCommons@University of Nebraska - Lincoln. It has been accepted for inclusion in USGS Staff -- Published Research by an authorized administrator of DigitalCommons@University of Nebraska - Lincoln. Global Change Biology (2012) 18, 656–674, doi: 10.1111/j.1365-2486.2011.02521.x Landscape controls on the timing of spring, autumn, and growing season length in mid-Atlantic forests ANDREW J. ELMORE*, STEVEN M. GUINN*, BURKE J. MINSLEY† andANDREW D. RICHARDSON‡ *Appalachian Laboratory, University of Maryland Center for Environmental Science, 301 Braddock Rd, Frostburg, MD USA, †United States Geological Survey, Crustal Geophysics and Geochemistry Science Center, Denver, CO USA, ‡Department of Organismic & Evolutionary Biology, Harvard University, Cambridge, MA USA Abstract The timing of spring leaf development, trajectories of summer leaf area, and the timing of autumn senescence have profound impacts to the water, carbon, and energy balance of ecosystems, and are likely influenced by global climate change. Limited field-based and remote-sensing observations have suggested complex spatial patterns related to geo- graphic features that influence climate. However, much of this variability occurs at spatial scales that inhibit a detailed understanding of even the dominant drivers. Recognizing these limitations, we used nonlinear inverse mod- eling of medium-resolution remote sensing data, organized by day of year, to explore the influence of climate-related landscape factors on the timing of spring and autumn leaf-area trajectories in mid-Atlantic, USA forests. We also examined the extent to which declining summer greenness (greendown) degrades the precision and accuracy of observations of autumn offset of greenness. Of the dominant drivers of landscape phenology, elevation was the strongest, explaining up to 70% of the spatial variation in the onset of greenness. Urban land cover was second in importance, influencing spring onset and autumn offset to a distance of 32 km from large cities. Distance to tidal water also influenced phenological timing, but only within ~5 km of shorelines. Additionally, we observed that (i) growing season length unexpectedly increases with increasing elevation at elevations below 275 m; (ii) along gradi- ents in urban land cover, timing of autumn offset has a stronger effect on growing season length than does timing of spring onset; and (iii) summer greendown introduces bias and uncertainty into observations of the autumn offset of greenness. These results demonstrate the power of medium grain analyses of landscape-scale phenology for under- standing environmental controls on growing season length, and predicting how these might be affected by climate change. Keywords: forest phenology, growing season length, inverse modeling, remote sensing Received 31 May 2011; revised version received 21 July 2011 and accepted 4 August 2011 greenhouse gas concentrations and global warming Introduction (Baldocchi & Wilson, 2001; Chapin et al., 2008). How- One of the largest, most easily predicted, and already ever, if belowground resources become more limiting documented biological impacts of climate change on or if warmer temperatures stimulate heterotrophic res- temperate ecosystems is a lengthening of the growing piration, the impact of a longer growing season will be season (Menzel & Fabian, 1999; Field et al., 2007). In reduced (Gu et al., 2003; White & Nemani, 2003; Moris- summer-active/winter-dormant systems, the timing of ette et al., 2009; Noormets, 2009; Richardson et al., spring and autumn has a profound impact on the 2009b, 2010; Zhao & Running, 2010). Understanding water, carbon, and energy balance of forests, grass- landscape and climatic controls on the trajectory of leaf lands, and agricultural fields (Parmesan & Yohe, 2003; area development, summer maturity, and senescence is Liang & Schwartz, 2009; Morisette et al., 2009). In par- therefore key to understanding the impact of feedback ticular, a longer growing season, which has already processes and ecosystem response to climate change. been detected in remote sensing time series for some The nature of phenological patterns in spring and regions (Myneni et al., 1997; Zhang et al., 2007), might autumn is dependent on the scale of observation, which be expected to increase carbon uptake by temperate for- strongly determines the suite of environmental parame- ests, providing an important negative feedback to ters influencing phenological timing. At continental scales, growing season length is predicted well by Correspondence: Andrew J. Elmore, Appalachian Laboratory, 301 climatic gradients associated with latitude, longitude, Braddock Rd., Frostburg, MD 21532, USA, tel. +1 (301) 689 7124, and elevation (Hopkins, 1918; Fitzjarrald et al., 2001). fax +1 (301) 689 7200, e-mail: [email protected] Common garden experiments have also demonstrated 656 © 2011 Blackwell Publishing Ltd LANDSCAPE CONTROLS ON GROWING SEASON LENGTH 657 a direct effect of photoperiod (as expressed by latitudi- Coarse resolution sensors such as the Moderate Resolu- nal differences) on spring phenological timing for some tion Imaging Spectrometer (MODIS) are inadequate for species (Lechowicz, 1984). At smaller scales, spatial pat- detecting fine-scale variability in the onset or offset of terns are likely more complex, incorporating a greater greenness due to microclimatological patterns that are range of processes. For example, urban heat islands obscured by the large pixel sizes. Recently, methods (Imhoff et al., 2010) have been shown to lead to an ear- have been developed to calculate average phenology at lier start of spring in some areas of the world (Zhang medium spatial resolution (Fisher et al., 2006). These et al., 2004; Fisher et al., 2006), but not all (Gazal et al., methods involve analyzing stacks of many medium-res- 2008). Climate moderation due to large water bodies olution satellite images, organized by day of year has been related to a later spring onset of greenness, (DOY) and discarding the year of acquisition. Measure- and even microclimatic patterns related to cold air ments of vegetation from each image are then used to fit drainage into small valleys have been shown to influ- an average phenology curve. The resulting average date ence the timing of the spring onset (Fisher et al., 2006). of the onset of greenness has been validated against Fine scale variability in phenological timing can be due field data (Fisher et al., 2006), spatially aggregated and to species effects (Liang et al., 2011), with the phenol- compared against measurements from lower-resolution ogy of individual tree species consistently either pre- data (Fisher & Mustard, 2007), and compared with pre- ceding or lagging other species despite up to 3 weeks dictions from phenology models (Fisher et al., 2007). inter-annual variation (Richardson & O’Keefe, 2009). Such observations provide valuable information regard- Remote sensing data are widely considered to pro- ing spatial patterns in the average land surface phenol- vide a source for consistent phenological measurements ogy, but do not directly lend themselves to monitoring across space and time. However, observed trends in the effects of climate change over time. growing season length are highly sensitive to the obser- The objectives of this paper were to (i) develop and vation and algorithm used to identify the start and end quantify the uncertainty of a phenology curve-fitting of the growing season (White et al., 2009). Of the many algorithm that accounts for declining greenness studies investigating changes in phenology (e.g. Myne- throughout the summer growing season, here termed ni et al., 1997; Schwartz et al., 2006), the vast majority ‘greendown’; (ii) apply this algorithm at a spatial grain have focused on the start of spring or have used the appropriate for the study of individual forest stands; same methodology to identify both ends of the growing and (iii) use the derived estimates of spring onset and season (Robeson, 2002; Christidis et al., 2007; Vitasse autumn offset of greenness and growing season length et al., 2009). In general, studies that have examined both to explore landscape patterns in phenological timing. spring and autumn timing have concluded that both Due to mixed pixel effects, we saw the need for obser- are important to global change research due to differing vations of growing season length made at a signifi- effects on ecosystem metabolism (Ahas et al., 2002; cantly smaller spatial grain than what is routinely used