Journal list menu
Watching plants’ dance: movements of live and dead branches linked to atmospheric water demand
Corresponding Editor: Dawn M. Browning.
Abstract
Diurnal branch movements in woody plants have only recently been described in detail. While previously only vegetative and reproductive structures have been known to move on hourly timescales, imaging technologies such as terrestrial laser scanning and near-surface repeat digital photography provide a means of remotely monitoring plant movements at high enough temporal and spatial resolution to capture rhythmic movements of woody material. Virtually, nothing is known about the range of species and ecosystems in which woody movements might occur or what causes these movements. We report that diurnal woody branch movements occur in a number of tree and shrub species across a broad range of abiotic conditions. We examined detailed branch movements in one species, creosote (Larrea tridentata), and found that branch movements were highly correlated with humidity, air temperature, vapor pressure deficit, and stem water potential: all factors related to plant water status. We also found that live and dead branch movements were distinct in the timing of their movements and in the abiotic conditions with which they were most correlated. Changes in dead branch position were most correlated with humidity, with these movements consistently lagging 1–2 h behind changes in humidity. Live branch movements were also highly correlated with vapor pressure deficit and humidity but went from lagging 1–2 h behind changes in these abiotic conditions in summer to being nearly in sync in winter. We believe that this is the first study that (1) documents diurnal branch movements in creosote, (2) differentiates between the movements of live and dead branches, and (3) relates environmental data to these movements. We hope these findings encourage other researchers to more closely examine imagery from their sites for evidence of branch movements, which may provide deeper insights into water and solute movements in plants and physiological responses to water stress.
Introduction
Although movements of vegetative and reproductive plant organs are well documented (Darwin and Darwin 1880), diurnal movements of woody branches have only recently been described in detail (Puttonen et al. 2016, Zlinszky et al. 2017). There is mounting evidence that branch movements may occur in many woody species, but previous studies only documented movements in relatively well-watered conditions over short time frames (Puttonen et al. 2016, Zlinszky et al. 2017). It is unclear how frequently this phenomenon occurs in nature, whether movement patterns persist across long time periods, what the potential drivers of branch movements may be, and what role these movements may play, if any, in organismal to ecosystem feedbacks.
Canopy size, shape, branching architecture, and orientation influence plant physiological processes and interactions with other organisms and the environment (Norman and Campbell 1989). In particular, the architecture of branches and leaves affects light penetration, self-shading, transpiration rates, rainfall interception, stemflow, and plant microclimate (Valladares and Pugnaire 1999, Falster and Westoby 2003, Iida et al. 2005, Niinemets 2010). Because woody canopy shape, size, and architecture affect so many facets of a plants’ life, daily or sub-daily branch movements have the potential to continuously modify plant–organism and plant–environment interactions.
Changes in environmental conditions can trigger a range of non-woody plant movements. Daily cycles of light can elicit movements of leaves and flowers, using mechanisms such as changes in turgor pressure, circadian hormonal signaling and gene expression, and asymmetric growth or cell expansion (Atamian et al. 2016, Apelt et al. 2017). The consequences of light-induced movements range from reduced photoinhibition or herbivore damage to increased light interception or pollinator visitation (van Doorn and van Meeteren 2003). Leaves can also move rapidly in response to temperature changes, often to shelter delicate tissues from extreme heat or cold, enhance photosynthetic uptake, or increase water conservation (Smith 1974, Ludlow and Björkman 1984, Comstock and Mahall 1985, Gamon and Pearcy 1989, Nilsen 1991). Differences in stem and leaf water potential can drive or enhance leaf movements, altering rates of photosynthesis, stomatal conductance, and photoinhibition (Nilsen 1987, Kao and Forseth 1992, Xu et al. 2009). The mechanisms of non-woody movements vary widely between species, and only a fraction of these mechanisms may be realized in woody tissues, especially dead wood.
Existing sensor networks can be leveraged to explore the connections between branch movements and environmental conditions across a wide range of ecosystems. In two previous studies of rapid branch movements, researchers used high-resolution terrestrial laser scanning (TLS) to track overnight branch movements in two European silver birch (Betula pendula) trees (Puttonen et al. 2016) and nocturnal movement patterns in several other horticultural tree species growing in well-watered conditions (Zlinszky et al. 2017). Terrestrial laser scanning techniques can yield extraordinarily detailed, three-dimensional point clouds. These systems do not require external lighting but do work best in still air conditions. However, TLS equipment can be expensive, data sets prohibitively large, and analyses computationally intensive. Visible-spectrum cameras provide a cheap, easy-to-use alternative to monitoring branch movements. Research networks such as PhenoCam, EuroCam, and AUSCam have accumulated years of time series imagery. Cameras in these networks collect repeat imagery of static scenes, often at hourly frequency. They have been installed in natural, experimental, agricultural, laboratory, greenhouse, and urban environments (Richardson et al. 2007, Nichols et al. 2013, Petach et al. 2014). However, repeat digital photography has most often been used to relate canopy reflectance to carbon uptake and phenological transition dates (e.g., leaf-out and senescence). Cameras require external flashes to capture images in dark or nighttime settings and only produce two-dimensional images (unless stereo placement and post-processing is used). Still, multiple studies have tracked leaf movements using photographic techniques that could be translated to branch monitoring (Biskup et al. 2007). Many cameras have been co-located with meteorological, soil, and stem sensor networks, data from which could be coupled with observations of diurnal branch movements.
To better characterize the occurrence and potential consequences of branch movements in woody species, we first present a survey of near-surface repeat digital photographs from the PhenoCam network. We then focus on one species, in particular the desert shrub creosote (Larrea tridentata), to (1) quantify branch movements in both live and dead branches, (2) identify the potential abiotic and/or biotic drivers of these movements, and (3) discuss potential plant–environment feedbacks of these movements. We address these goals across a range of environmental conditions and across daily-to-seasonal timescales. The small stature, canopy structure, and dramatic branch movements of creosote made images of this species particularly easy to analyze. In addition, the extreme variability of semiarid environments in which creosote lives provided a greater range of conditions under which to study the triggers and ramifications of branch movements. We hypothesized that (1) dead branches would be more sensitive to changes in atmospheric moisture and temperature while live branch movements would be more sensitive changes in stem water potential and atmospheric demand, and (2) branch movements would affect plant microclimate, namely soil temperature.
Methods
Cross-site survey of woody plant movements
We surveyed PhenoCam imagery from cameras installed at the National Ecological Observatory Network (NEON) sites (NEON 2021). Over the past decade, NEON sites have been established to represent a variety of biomes, species, and environmental conditions (Keller et al. 2008). PhenoCams at these sites are placed to capture canopy, understory, and streamside images, providing a range of angles from which to potentially view branch movements (Elmendorf et al. 2016). NEON PhenoCams take up to four images per hour, increasing the probability of capturing fast branch movements. Special attention was paid to imagery from around dawn and dusk and during both humid and dry time periods. Unfortunately, the position of some cameras made branch movements hard to see or track over time. Other cameras were programmed to take photographs less frequently than once per hour, making smooth, continuous branch movements harder to follow. Thus, our survey simply highlights the diversity of species and ecological contexts in which branch movements can be observed. Lack of inclusion does not indicate a lack of branch movement at a given site, only that we did not detect movements in the images surveyed, during the time periods we surveyed. We attempted to identify live and seemingly dead branches, including fallen logs, within each camera scene.
Case study: branch movements in creosote
Site description
We more extensively documented branch movements at a creosote shrubland within the Sevilleta National Wildlife Refuge and Long-Term Ecological Research site in central New Mexico, USA. This study site has been operational since 2007 and has been an Ameriflux core site (US-Ses) since 2013 (D’Odorico et al. 2010, He et al. 2010, Anderson-Teixeira et al. 2011, Petrie et al. 2015). The vegetation at the site is dominated by creosote (L. tridentata), with sparse grasses (Bouteloua spp., Pleuraphis jamesii, and Scleropogon brevifolius), scattered forbs (Machaeranthera pinnatifida, Townsendia annua, and Gutierrezia sarothrae), and cacti (Opuntia macrocentra).
All data for our study of creosote branch movements were collected between 31 July 2015 and 5 December 2015. This study period encompassed a range of abiotic conditions. The growing season in this area of the northern Chihuahuan Desert is bimodal, with a short growing season in the spring (March–April) followed by a hot and dry period (typically May–June), and the main growing season occurring from mid-July to early October. The first half of our study period encompassed the main growing season (31 July–September). During this time, the air temperature was 23.2° ± 4.7°C (mean ± standard deviation [SD]) and the median volumetric soil water content was 9.7% ± 0.5%. During the last half of the study period (October–December), the growing season gave way to a wet winter, with the average air temperatures dropping to 9.8° ± 7.0°C and the median volumetric soil water content rising to 15.9% ± 3.3%.
In February 2011, nearly five years before our study, this site experienced an extreme cold event, with temperatures dropping to −30°C. Many creosote, evergreen shrubs, suffered extreme (>90%) canopy dieback. Although there was very little creosote mortality (Ladwig et al. 2019), the shrubs were left with a unique crown of dead branches. In the ensuing years, shrub canopies regrew from the base of each plant. All dead creosote branches described in this study remain connected to the central stem of the plant, where the living branches also originate, but are visibly distinct from their living counterparts. They have no vegetative growth or living tissue from the stem tip to the stem base where the branch enters the soil. Dead branches are dry, brittle, missing most or all of their bark, have deep cracks, and have no measurable xylem water.
Repeat digital photographs
For this study, we positioned three Moultrie Game Spy I-60 cameras (EBSCO Industries, Inc., Birmingham, Alabama, USA) to photograph creosote shrubs growing within 5 m of the main site instrumentation where we measured air temperature and humidity. These focal creosote were located above soil pits where we measured soil temperature and were further instrumented with stem psychrometers. Our cameras took one photograph per hour throughout the study period. At night, scenes were illuminated with built-in infrared camera flashes that were automatically activated once the ambient lighting dropped below a certain level. Within each scene, we selected multiple branch points (branch tips or nodes) which we could distinguish in photographs throughout the study period and during both day and nighttime conditions. We placed white plastic balls on several additional branches so that their positions could be more easily tracked. Branches ranged from 0.4 to 0.8 m in total length and were located within 6 m of each camera. All scenes and tracked branches are shown in Appendix S1: Fig. S1 and described in Appendix S1: Table S1.
Branch movements were quantified using the y-pixel-coordinate of branch points within each hourly photograph. Branches flexed along their entire length, from base to tip, bending in x-, y-, and z-coordinate space. This constantly changing branch geometry, as well as the density of branches at the base of the shrub, made it impossible to track the trajectory of entire branch lengths in still photographs. Our analysis focuses on changes in each branch point’s vertical (y-pixel-coordinate) position within each photograph time series because vertical, “up-and-down,” movements were the most apparent in the imagery. To standardize this measurement for branches of different lengths and at different distances from the camera, we z-scored the y-coordinate time series of each branch point, a metric we call Branch Position. This means that each branch’s mean Branch Position was centered at zero and a score of +1 indicates that the branch was positioned one standard deviation higher than average. Positive Branch Position indicates that a branch is oriented more skyward, while negative Branch Position indicates that a branch is closer to the ground. A larger absolute value of Branch Position indicates that the branch moved further from its average vertical position. We visually assessed live branches to confirm that branch growth was minimal over the course of the study period. Stationary objects were tracked within each scene to detect wind interference and ensure that cameras did not drift significantly over the course of the study period. We removed outlier Branch Position points (<0.01% of data), filled gaps of less than 6 h using a spline method, and smoothed all data to decrease noise.
Meteorological data
We measured relative humidity and air temperature with an HMP45C Vaisala temperature/relative humidity probe (Vaisala Instruments, Helsinki, Finland) and used these values to calculate vapor pressure deficit. Incoming photosynthetically active radiation was measured with a Kipp & Zonen LI-190 PAR sensor (LI-COR, Lincoln, Nebraska, USA). These sensor data were continuously measured at 10-Hz frequency and stored as 30-min averages. Precipitation, recorded as a 30-min sum, was measured using a TE525 Texas Electronics 6” tipping bucket rain gage (Texas Electronics, Dallas, Texas, USA).
From 7 August 2015 to 15 August 2015, we measured stem water potential using a stem psychrometer installed at the base of a living creosote stem. After 15 August, we installed an automated PSY1 stem psychrometer (ICT International, Armidale, New South Wales, Australia) at the base of a living stem on a different creosote growing approximately 3 m from the first. Stem water potential was calculated as the difference in wet bulb and dry bulb thermocouple temperatures, recorded every 30 min, and corrected for ambient air temperature (Dixon and Tyree 1984). All psychrometers were calibrated using standardized saline solutions in the laboratory before installation.
We measured soil temperature and soil water content in four soil profiles at 2.5-, 12.5-, 22.5-, 37.5-, and 52.5-cm depths. Two “covered” profiles were located under creosote canopies, and another two “uncovered” profiles were located in bare canopy interspaces without shrub or grass cover. Soil water content was measured with CS-616 water content reflectometers at each profile depth (Campbell Scientific, Logan, Utah, USA). Soil temperature was measured with thermocouple probes at each depth (T-107; Campbell Scientific) and additionally with dielectric water potential sensors at 22.5- and 37.5-cm depths (MPS6; Decagon Devices, Pullman, Washington, USA). Soil temperature and water content were measured every 5 min and recorded as 30-min averages. Other variables such as air pressure and wind speed, were measured but are not shown here due to the lack of correlation with branch movements (linear correlation with Branch Position: R2 = 0.007 and R2 = 0.03, respectively).
Data processing
Data manipulation and statistical analyses were conducted with R 3.5.0 (R Core Team 2020). In order to disentangle the relative importance of multiple abiotic factors, which all displayed some degree of diurnal periodicity, we calculated cross-correlations between Branch Position and abiotic variables at differing hourly lags (from −3 to +3 h) using the ccf() function from the R package stats (R Core Team 2020). Cross-correlation fits were calculated on both daily and seasonal timescales. To assess how well branch movements were correlated with abiotic factors on daily timescales, we calculated lagged correlations within 5-d rolling windows along the entire time series. There were a total of 118 of these 5-d windows included in our analysis. Within each 5-d window, the correlation between up to 120 hourly data points per data time series was compared. We also calculated the correlation between abiotic variables and Branch Position across the entire time series, using all hourly data from July to December to fit cross-correlation models. This approach helped us determine whether variables were correlated across the entire study period or only within certain seasons, and whether the lag between the factors changed seasonally.
While multiple environmental variables display diurnal periodicity, we hypothesized that only some have the potential to drive branch movements. This is particularly true for dead branches, which lack the living cells needed to sense sunlight, produce signaling hormones, or create xylem water potential gradients. In order to differentiate between factors that were simply co-correlated with branch movements and those that were correlated and potentially driving branch movements, we assessed the linear relationship with each abiotic factor at multiple lags. Fig. 1 illustrates four potential outcomes of this analysis. These figures use hypothetical data, not real data from our analyses. In each case, there would be a strong, statistically significant correlation between the predictor and response signals. In the first panel, the predictor and response signals are completely synchronized. As the background color of the lower correlation panel indicates, the strongest cross-correlation between the two signals occurs at a 0-h time lag. The black line in the lower correlation panel shows that the cross-correlation between the two signals is 100%. We may expect to see this signal whether the predictor signal elicits an immediate reaction in the response signal. The second panel illustrates a potential scenario when the response consistently lags behind the predictor signal. Here, the strongest cross-correlation between the two signals (averaging ˜95% correlation) occurs when the predictor time series is shifted back (earlier) in time. We might expect to see this pattern when the response signal is reacting directly to the predictor variable but takes some time to occur. This scenario could occur if the predictor signal was driving the response signal. The third panel illustrates a potential scenario where both signals are highly correlated, but because the response signal consistently leads the so-called predictor variable in time, the predictor signal could not be causing the response signal. Finally, the fourth panel illustrates a scenario where the response signal does not consistently change with the predictor signal over time. We assume this scenario might occur if there is no relationship between the predictor and response signals, or if the predictor signal directly elicits a change in the response signal, but only under certain conditions.
We are using this lagged correlation analysis as a first step in narrowing down the list of possible causal factors related to branch movements. We think this framework is useful when investigating this newly discovered phenomenon in a natural setting where many abiotic and biotic factors display similar diurnal and seasonal periodicity. We emphasize that these analyses do not by themselves indicate direct causation between abiotic drivers and branch movements.
Results
Cross-site survey of woody plant movements
Using time lapse photographs from NEON sites and the PhenoCam Network, we found evidence of diurnal woody branch movements in a range of species and ecosystems, from temperate woodlands to boreal forests and arid shrublands (Table 1, Fig. 2; Video S1). In some species, we observed live and dead branches moving synchronously, with live and dead branches moving upward and downward in tandem. In other species, branches moved asynchronously or without discernable diurnal patterns. In general, all species moved more vertically than in other directions. Humid conditions amplified movements in most species.
| NEON domain | Location | Latitude | Longitude | Elevation | Camera placement | Branch movements | |
|---|---|---|---|---|---|---|---|
| Live | Dead | ||||||
| Northeast | Bartlett Experimental Forest, NH | 44.0639 | −71.2874 | 285 | Top-of-tower | Yes | n/o |
| Bartlett Experimental Forest, NH | 44.0639 | −71.2874 | 285 | Mid-tower | Yes | Yes | |
| Harvard Forest, MA | 42.5369 | −72.1727 | 359 | Top-of-tower | Yes | n/o | |
| Harvard Forest, MA | 42.5369 | −72.1727 | 359 | Mid-tower | Yes | n/o | |
| Hop Brook, MA | 42.4718 | −72.3296 | 203 | Stream gauge | Yes | Yes | |
| Mid-Atlantic | Blandy Experimental Farm, VA | 39.0337 | −78.0418 | 162 | Top-of-tower | Yes | n/o |
| Blandy Experimental Farm, VA | 39.0337 | −78.0418 | 162 | Mid-tower | Yes | n/o | |
| Posey Creek, VA | 38.8933 | −78.1468 | 293 | Stream gauge | Yes | n/o | |
| Smithsonian Conservation Biology Institute, VA | 38.8929 | −78.1395 | 364 | Mid-tower | Yes | n/o | |
| Smithsonian Environmental Research Center, MD | 38.8901 | −76.5600 | 30 | Mid-tower | Yes | n/o | |
| Southeast | Flint River, GA | 31.1854 | −84.4374 | 27 | Stream gauge | n/o | Yes |
| Ordway–Swisher Biological Station, FL | 29.6893 | −81.9934 | 56 | Mid-tower | Yes | Yes | |
| Atlantic Neotropical | Rio Cupeyes, PR | 18.1135 | −66.9868 | 164 | Stream gauge | Yes | Yes |
| Guanica Forest, PR | 17.9696 | −66.8687 | 136 | Top-of-tower | Yes | n/o | |
| Guanica Forest, PR | 17.9696 | −66.8687 | 136 | Mid-tower | Yes | Yes | |
| Great Lakes | Crampton Lake, WI | 46.2111 | −89.4783 | 518 | Stream gauge | Yes | n/o |
| Steigerwaldt Land Services, WI | 45.5089 | −89.5864 | 476 | Top-of-tower | Yes | n/o | |
| Steigerwaldt Land Services, WI | 45.5089 | −89.5864 | 476 | Mid-tower | Yes | Yes | |
| Treehaven, WI | 45.4937 | −89.5857 | 474 | Mid-tower | Yes | n/o | |
| UNDERC, MI | 46.2339 | −89.5373 | 529 | Top-of-tower | Yes | n/o | |
| UNDERC, MI | 46.2339 | −89.5373 | 529 | Mid-tower | Yes | Yes | |
| Prairie Peninsula | Kings Creek, KS | 39.1051 | −96.6034 | 339 | Stream gauge | Yes | n/o |
| McDiffett Creek, KS | 38.9443 | −96.4420 | 376 | Stream gauge | Yes | n/o | |
| The Univ. of Kansas Field Station, KS | 39.0404 | −95.1922 | 330 | Mid-tower | Yes | n/o | |
| Appalachians and Cumberland Plateau | Great Smoky Mountains National Park, TN | 35.6890 | −83.5020 | 589 | Top-of-tower | Yes | n/o |
| Great Smoky Mountains National Park, TN | 35.6890 | −83.5020 | 589 | Mid-tower | Yes | Yes | |
| LeConte Creek, TN | 35.6904 | −83.5038 | 578 | Stream gauge | n/o | Yes | |
| Mountain Lake Biological Station, VA | 37.3783 | −80.5248 | 1177 | Top-of-tower | Yes | n/o | |
| Mountain Lake Biological Station, VA | 37.3783 | −80.5248 | 1177 | Mid-tower | Yes | Yes | |
| Walker Ranch, TN | 35.9595 | −84.2804 | 274 | Stream gauge | Yes | n/o | |
| Ozarks Complex | Black Warrior River, AL | 32.5415 | −87.7982 | 23 | Stream gauge | Yes | n/o |
| Dead Lake, AL | 32.5417 | −87.8039 | 36 | Top-of-tower | Yes | n/o | |
| Dead Lake, AL | 32.5417 | −87.8039 | 36 | Mid-tower | |||
| Lenoir Landing, AL | 31.8539 | −88.1612 | 10 | Top-of-tower | Yes | n/o | |
| Lenoir Landing, AL | 31.8539 | −88.1612 | 10 | Mid-tower | Yes | n/o | |
| Mayfield Creek, AL | 32.9597 | −87.4081 | 93 | Stream gauge | Yes | Yes | |
| Talladega National Forest, AL | 32.9505 | −87.3933 | 167 | Mid-tower | n/o | Yes | |
| Tombigbee River, AL | 31.8534 | −88.1589 | 10 | Stream gauge | n/o | Yes | |
| Central Plains | Rocky Mountain National Park, CASTNET, CO | 40.2759 | −105.5460 | 2751 | Mid-tower | Yes | n/o |
| Southern Plains | LBJ National Grassland, TX | 33.4012 | −97.5700 | 279 | Mid-tower | Yes | Yes |
| Pringle Creek, TX | 33.3786 | −97.7823 | 255 | Stream gauge | Yes | n/o | |
| Northern Rockies | Yellowstone National Park, WY | 44.9535 | −110.5391 | Mid-tower | Yes | n/o | |
| Southern Rockies and Colorado Plateau | Como Creek, CO | 40.0350 | −105.5449 | 3036 | Stream gauge | Yes | Yes |
| West St Louis Creek, CO | 39.8914 | −105.9154 | 2920 | Stream gauge | n/o | Yes | |
| Desert Southwest | Santa Rita Experimental Range, AZ | 31.9107 | −110.8355 | 999 | Mid-tower | Yes | n/o |
| Sycamore Creek, AZ | 33.7491 | −111.5069 | 644 | Stream gauge | n/o | Yes | |
| Great Basin | Red Butte Creek, UT | 40.7839 | −111.7979 | 1696 | Stream gauge | Yes | n/o |
| Pacific Northwest | Abby Road, WA | 45.7624 | −122.3303 | 390 | Top-of-tower | Yes | n/o |
| Abby Road, WA | 45.7624 | −122.3303 | 390 | Mid-tower | Yes | n/o | |
| Martha Creek, WA | 45.7912 | −121.9320 | 354 | Stream gauge | Yes | n/o | |
| McRae Creek, OR | 44.2596 | −122.1656 | 880 | Stream gauge | n/o | Yes | |
| Wind River Experimental Forest, WA | 45.8205 | −121.9519 | 368 | Mid-tower | Yes | Yes | |
| Pacific Southwest | Upper Big Creek, CA | 37.0597 | −119.2575 | 1133 | Stream gauge | n/o | Yes |
| Lower Teakettle, CA | 37.0058 | −119.0060 | 2149 | Mid-tower | Yes | Yes | |
| Taiga | Caribou Creek—Poker Flats Watershed, AK | 65.1540 | −147.5026 | 233 | Mid-tower | Yes | Yes |
| Caribou Creek at Poker Flats, AK | 65.1531 | −147.5025 | 229 | Stream gauge | Yes | Yes | |
| Delta Junction, AK | Mid-tower | Yes | Yes | ||||
Note
- Each camera’s NEON domain, site name, latitude, longitude, and elevation are noted, as well as an indication of whether periodic live or dead branches’ movements were observed (yes) or not observed (n/o). NEON, National Ecological Observatory Network.
Case study: branch movements in creosote
At our creosote case study site, we tracked 18 creosote branches in hourly photographs for 126 d between 31 July and 4 December 2015. The end of our study period was cut short by a series of snowstorms that covered the shrubs in snow and fogged the cameras for several weeks. The branches we tracked all displayed cyclical daily movements throughout the entire study period (Video S2). Using trigonometric methods, we estimate that some branches moved more than 20 vertical centimeters per day during this study period. A timeline and photograph montage illustrating 48 typical hours of branch movement is shown in Fig. 3.
Creosote branches were typically oriented higher (skyward, steeper angle) at night and lower (groundward, shallower angle) in the day. The most common diurnal pattern of branch movement we observed was downward movement (decrease in branch angle) initiated at dawn, with branches reaching their lowest height midday, and upward movement (increase in branch angle) starting in the afternoon or evening, with maximum height reached just before dawn each day. These diurnal movements were often correlated with the diurnal and weekly–biweekly cycles of relative humidity, air temperature, vapor pressure deficit, and stem water potential (Fig. 4). Creosote branches maintained a steeper angle, on average, in the wet winter months than in the hotter monsoon months (Fig. 5). Surprisingly, branch movements in creosote did not track seasonal patterns in stem water potential (Fig. 6).
Comparing live and dead creosote branch movements
Branch movements of live and dead creosote branches were highly correlated to one another on diurnal and seasonal timescales (Fig. 5). Live branches, however, consistently moved before dead branches throughout the day. Live branches started to droop earlier at dawn and also stabilized and started raising earlier in the afternoon or evening (Fig. 4). Over the whole study period, live branches moved, on average, 1 h before dead branches (Table 2). The average cross-correlation coefficient between live and dead Branch Position within all 5-d rolling windows was 84.0% ± 6.1% (mean ± SD), and the average time lag was −1.0 ± 0.8, meaning that live branches changed position ˜1 h before dead branches. When comparing live and dead Branch Position with a single cross-correlation model, which included all data from the entire study period, the correlation was 71.9% with a −1.0-h lag, meaning that live branches changed position an hour before dead branches. Although the average lag between live and branch movements was approximately an hour, we do see a slight change in this lag throughout the study period. In the growing season (July–September), live branches moved 1–2 h before dead branches (average of all 5-d window correlations = 82.7%, single model correlation = 71.2%) (Fig. 6). In the winter months (October–December), however, live and dead branch movements were more correlated (average correlation within all 5-d windows = 85.2%, single model correlation = 84.8%) and nearly synchronized (lag decreased to 0–1 h) (Fig. 6). Overall, dead branches moved more than live branches—they drooped closer to the ground and lifted higher toward the sky.
| Environmental factor | Live branches | Dead branches | ||||||
|---|---|---|---|---|---|---|---|---|
| All 5-d windows | All season | All 5-d windows | All season | |||||
| Correlation (%) | Time lag (h) | Correlation (%) | Time lag (h) | Correlation (%) | Time lag (h) | Correlation (%) | Time lag (h) | |
| Relative humidity | 85 ± 6 | 0.6 ± 0.7 | 64 | 1 | 88 ± 6 | 1.8 ± 0.5 | 86 | 2 |
| Vapor pressure deficit | 84 ± 10 | 0.4 ± 0.6 | 86 | 0 | 80 ± 7 | 1.8 ± 0.5 | 68 | 2 |
| Air temperature | 79 ± 19 | 0.2 ± 0.8 | 84 | 0 | 71 ± 17 | 1.4 ± 1.0 | 48 | 2 |
| Stem water potential | 75 ± 15 | −0.3 ± 1.1 | 40 | 0 | 73 ± 17 | 1.0 ± 1.2 | 52 | 1 |
| Photosynthetically active radiation | 68 ± 11 | 2.6 ± 0.8 | 51 | 3 | 47 ± 10 | 1.9 ± 1.8 | 35 | 3 |
Notes
- First, the correlation coefficient (mean ± standard deviation) within every 5-d rolling window is listed along with the time lag that maximized the correlation between Branch Position and the abiotic variable. Second, we report the correlation coefficient and maximized time lag when all data are used in a single model.
Relationships between creosote Branch Position and abiotic factors
Because dead branches lack leaves, they are often easier to distinguish in photographs. However, their movement patterns often differ in timing or direction when compared to their live, leafy neighbors. In creosote, live branch movements were highly correlated with relative humidity, temperature, vapor pressure deficit, and stem water potential on daily timescales (Fig. 4, Table 2). However, the time lag between live branch movements and these abiotic factors differed throughout the study period (Fig. 6). Changes in live branch position were in sync or slightly lagging behind (0- to 1-h lag) most abiotic factors throughout the growing season, and were less consistently correlated with abiotic factors in the winter (Fig. 6).
Within all 5-d windows, Branch Position of live branches was most strongly correlated with relative humidity (85% ± 6%) and vapor pressure deficit (84% ± 10%), moving 0.6 ± 0.7 h after observed changes in relative humidity and 0.4 ± 0.6 h after changes in vapor pressure deficit (Table 2). In the single, all-season cross-correlation model, however, live branch movements were correlated most strongly with vapor pressure deficit and air temperature (86% with a 0-h lag and 84% with a 0-h lag, respectively) (Table 2). Notably, live branch movements in creosote were not highly correlated with stem water potential (Table 2). In 5-d windows throughout the study period, the −0.3 ± 1.1-h lag between these two variables indicates on short timescales, live branch movements (measured at or near terminal branch nodes) often occurred before changes in stem water potential (measured at the branches’ base near the ground).
Dead branches consistently moved 1–2 h after observed changes in relative humidity, air temperature, and vapor pressure deficit throughout the study period (Table 2). Stem water potential of live branches and photosynthetically active radiation were also weakly correlated with dead branch movements (Table 2). However, the dead branches we tracked have no measurable stem water potential or living cells with which to sense sunlight. Although we found statistically significant correlations between dead branch movements and these factors, biologically, they cannot drive the branch movements. Therefore, these model results (Table 2) serve as signposts: “significant” correlations as weak and time lags as variable as between dead branches and stem water potential or sunlight reflect spurious correlations with factors that have similar diurnal periodicity but are not directly causing branch movements.
Creosote Branch Position and soil temperature
We compared creosote Branch Position with changes in soil temperature beneath creosote canopies. Fig. 7 illustrates the difference between soil temperature under creosote canopies vs. soil temperature in unshaded bare ground (ΔTsoil) during the months of August and November, when soil temperature data were available. Soil temperatures beneath creosote were 1.15° ± 0.7°C (mean ± SD) cooler than in intercanopy spaces in August and 0.21° ± 0.6°C in November. There was a diurnal pattern to this temperature difference (Fig. 7). Soils beneath creosote canopies were slightly warmer than soils in intercanopy spaces right after dawn. During each day, ΔTsoil increased, with maximum under-canopy cooling occurring a few hours before sunset. In August, changes in ΔTsoil occurred 3.3 ± 0.3 h after changes in live Branch Position and these factors were fairly well correlated (68% ± 7% average correlation across all 5-d windows in August). In November, the correlation between live Branch Position and ΔTsoil was slightly weaker (57% ± 12%) and the time lag was shorter (0.85 ± 0.37 h).
In contrast, the average correlation between dead Branch Position and ΔTsoil was only 50% ± 14% in August and 45% ± 14% in November. In August, changes in ΔTsoil occurred 1.9 ± 0.3 h after changes in dead Branch Position, but in November, changes in dead Branch Position occurred 0.48 ± 0.58 h before changes in ΔTsoil.
Discussion
Cross-site survey of woody plant movements
We documented diurnal and sub-diurnal branch movements in multiple woody species across a broad range of ecosystems. These observations, along with the findings of Puttonen et al. (2016) and Zlinszky et al. (2017), show that many species are capable of rapid, quantifiable branch movements. In the case of creosote, these movements seem to occur in response to abiotic conditions, fitting the definition of plant behavior (Karban 2008). Identifying the drivers and repercussions of these movements in different species may be an exciting new field of study, one aided by open data and the prevalence of highly instrumented study sites worldwide.
We were able to mine photographs from an existing public depository, the PhenoCam network, to retroactively document branch movements across a spectrum of ecosystem monitoring sites (NEON), despite the fact that these cameras were not originally installed for this purpose. Although digital photographs yield lower resolution data than the TLS techniques employed in previous research, we were able to use them to remotely monitor branch movements across many sites at high frequency (hourly) over long time periods (months to years) and readily distinguish between live and dead branches. Digital cameras are cheap, easy-to-use, pervasive, and non-invasive instruments with which to study plant movements. Factors such as wind or intense rain can obscure images, but this drawback is common across many sensors, including TLS. Cameras deployed for biological monitoring are often co-located with meteorological stations, flux towers, and other sensor arrays. The prevalence of camera imagery with associated environmental sensor data makes photographs an ideal medium to further study the relationship between plant movements and abiotic factors.
Case study: branch movements in creosote
We documented branch movements of one species, creosote, over the course of several months. We found subtle but consistent differences in the timing of live and dead branch movements. While gross patterns of movement were the same in all branches—raising skyward at night and drooping groundward in the day—there was a consistent temporal lag between live and dead branches. We then leveraged co-located site instrumentation to correlate the movements of live and dead branches with potential abiotic drivers. In both cases, the moisture content of the air seemed to the most likely driver of branch movements. Based on our results, we think dead creosote branches swell and dry in response to the humidity (or wetness) of the air, while live branches respond more to changes in vapor pressure deficit (or dryness) of the air. These factors are co-correlated, but the differing reactions reveal how this wood reacts when wood fibers are exposed to the atmosphere vs. when internal water content is controlled by living stomata, intact vessel elements, and protective bark tissues.
Dead branches move regularly, even though they cannot sense light, produce intercellular hormones, or transport water or solutes through intact vessel elements. This substantially narrows down the list of possible drivers of dead branch movements. In creosote, we found that dead branch movements consistently tracked relative humidity, with a 1- to 2-h lag, at daily and seasonal timescales (Table 2). According to our conceptual framework, the fact that dead Branch Position is highly correlated with relative humidity and dead branch movements consistently lag ˜2 h behind changes in relative humidity makes relative humidity a likely candidate for causing subsequent dead branch movements. This suggests that creosote wood has a structure that causes it to passively flex up (skyward) and down (groundward) when exposed to changing humidity. Wood is known to bend in response to changing humidity (Armstrong and Christensen 1961). In fact, balsam fir twigs, stripped of bark, are sold in some regions as “weather sticks” to predict incoming storms. Deformation patterns differ between species of wood, influencing which species we use as building materials (structural timbers, furniture-grade woods) and wood products (composite boards, paper) (Zhou et al. 1999). However, this kinetic behavior has previously only been associated with cut timber, not wood (live or dead) that is still part of a living plant (Holstov et al. 2015). In our continental survey of branch movements and at our study site, we observed movements in dead branches of living plants, dead woody plants, and fallen logs. Like cut timber, this dead material has more open pores and cracks and less protective bark than live wood. These exposed surfaces can interact with the moisture content and temperature of water, soil, and air, causing different planes of the wood to passively flex, just like in timber. Further study of branch movements may provide important insights in the mechanical properties of different species’ wood, impacting economically important fields such as silviculture, engineering, and material sciences.
We were surprised to find that live creosote branch movements were not related to seasonal patterns of stem water potential but responded primarily to atmospheric water demand (humidity and vapor pressure deficit) (Table 2). At least in this species, atmospheric water potential experienced at the stomata seemed to be a more important driver of live branch movements than water potential within the stem. In the absence of leaves, bark, and other living tissues, we would expect live and dead wood on the same plant to behave in more or less similar patterns. Therefore, the differences in the response of these tissues are likely attributable to biotic control over water loss. In creosote, the correlation and time lag between live branches and vapor pressure deficit were more variable than between dead branches and humidity (Fig. 6, Table 2). These patterns also changed seasonally, indicating differing stomatal behavior in different seasons. Other desert shrubs change leaf angles seasonally, optimizing photosynthetic and water conservation capabilities in different environmental regimes (Comstock and Mahall 1985).
Branch Position and plant–environment feedbacks
Woody plant architecture and non-woody plant movements have repercussions on plant–plant, plant–animal, and plant–environment feedbacks. In order to investigate how fast branch movements may alter these feedbacks, we compared branch position with soil temperature under the canopies of creosote. We found that soil under creosote canopies was briefly warmed in the early mornings, then shaded and cooled (by more than 3°C in hot months) for the rest of the day, relative to soils in the intercanopy area. The movement of live, leafy branches was strongly correlated with these cooling effects, suggesting that branch movements may play a role in controlling canopy and soil microclimates in creosote shrublands. Canopy shading has been shown to reduce soil water evaporation in other desert shrubs and trees (Tracol et al. 2011, Royer et al. 2012). Sub-daily branch movements in desert shrubs may enhance canopy shading, increasing soil water retention in the hottest, driest conditions.
Desert shrub canopies play other roles in ecosystem feedbacks. In creosote, branch orientation and leaf inclination reduce heat and water stress to foliar tissues (Ezcurra et al. 1991, 1992). The canopy size and angle of branches play an important role in stemflow, capturing nutrients through dry deposition and funneling them to the base of the plant, localized wind patterns, and light penetration into the canopy (Martinez-Meza and Whitford 1996, De Soyza et al. 1997, Whitford et al. 1997, Devakumar et al. 1999, Johnson and Lehmann 2006). Larrea species vary in both architecture and branch angles, depending on their latitude and habitat, suggesting that this genus shows plasticity in its architecture in order to adapt to arid conditions. (Ezcurra et al. 1991). In general, desert and cerrado shrubs and saplings with higher branch orientation and straight stems have higher stemflow (Wang et al. 2013, Honda et al. 2015, Levia et al. 2015, Zhang et al. 2017). In addition, branches that have been exposed to different (wind-induced) movements display differing flexibility and oscillations when exposed to wind later in life. So branch flexibility may be adaptive (Sellier and Fourcaud 2005). It is unknown how daily changes in branch position may affect these and other biotic and abiotic environmental feedbacks.
Implications and Conclusions
The assumption that woody plants have static architecture permeates many areas of scientific theory and methodology. We encourage fellow scientists to consider diurnal branch movements in future study designs. Anecdotally, we found that differences in branch position within a single day changed total canopy volume and the resulting biomass estimations of creosote individuals by over 20% when using volume:biomass allometric relationships. This diurnal difference in canopy volume could affect the remote-sensed size and position of woody plants measured using drone or TLS techniques.
We recommend further study of branch movements at sites with PhenoCams, especially those we have identified in Table 1, by researchers familiar with the environmental context of those sites. We suggest that future studies of branch movements attempt to incorporate a range of individual plant conditions (dead, alive, healthy, sick) and age class and environmental conditions to better understand daily and seasonal variation in woody plant architecture. While our continental survey focused on daytime images of plants, previous studies used nighttime TLS point clouds (Puttonen et al. 2016, Zlinszky et al. 2017). Twenty-four-hour observations of branch movements are preferred to capture the full diurnal cycle of branch movements. For researchers using optical sensors (e.g., cameras), we recommend the use of flashes to illuminate nighttime scenes and camera placement close enough to target plants to capture individual branches. Using automated systems to track branch movements over long study periods may help us understand plant physiology and stress adaptation better in a variety of species and habitats. Beyond simply being an interesting phenomenon, these movements may provide insight into daily changes in stress behavior and environmental interactions previously thought to only change over the course of entire seasons or plant lifetimes.
Acknowledgments
Alesia Hallmark, Marcy Litvak, Robert Pangle, and Gregory Maurer designed the study, calibrated instrumentation, processed incoming data streams, and contributed to analyses. Alesia Hallmark and Marcy Litvak led the writing of the manuscript. All authors contributed to drafts of the final manuscript. We would like to thank technicians Margaret Schluter and Jonathan Furst for extensive support in the design of this project in the laboratory and in the field. We would also like to thank Amy Bennett, Tim Ohlert, and other members of the Collins Think Tank for their enthusiasm and reading early versions of this manuscript. Dr. William Pockman contributed invaluable knowledge about creosote bush physiology and donated the psychrometers and CR7 data logger. Drs. Scott Collins, Chris Lippitt, and Andrew Richardson provided valuable feedback on early drafts. We acknowledge funding for this project from the National Science Foundation Sevilleta LTER, 1440478, and from the Ameriflux Management Project, Department of Energy Subcontract Number 7074628. The National Ecological Observatory Network is a program sponsored by the National Science Foundation and operated under cooperative agreement by Battelle.
