Climate models predict that water limited regions around the world will become drier and warmer in the near future, including southwestern North America. We developed a large-scale experimental system that allows testing of the ecosystem impacts of precipitation changes. Four treatments were applied to 1600 m2 plots (40 m × 40 m), each with three replicates in a piñon pine (Pinus edulis) and juniper (Juniper monosperma) ecosystem. These species have extensive root systems, requiring large-scale manipulation to effectively alter soil water availability. Treatments consisted of: 1) irrigation plots that receive supplemental water additions, 2) drought plots that receive 55% of ambient rainfall, 3) cover-control plots that receive ambient precipitation, but allow determination of treatment infrastructure artifacts, and 4) ambient control plots. Our drought structures effectively reduced soil water potential and volumetric water content compared to the ambient, cover-control, and water addition plots. Drought and cover control plots experienced an average increase in maximum soil and air temperature at ground level of 1-4° C during the growing season compared to ambient plots, and concurrent short-term diurnal increases in maximum air temperature were also observed directly above and below plastic structures. Our drought and irrigation treatments significantly influenced tree predawn water potential, sap-flow, and net photosynthesis, with drought treatment trees exhibiting significant decreases in physiological function compared to ambient and irrigated trees. Supplemental irrigation resulted in a significant increase in both plant water potential and xylem sap-flow compared to trees in the other treatments. This experimental design effectively allows manipulation of plant water stress at the ecosystem scale, permits a wide range of drought conditions, and provides prolonged drought conditions comparable to historical droughts in the past – drought events for which wide-spread mortality in both these species was observed.
The focus of this study was to determine the effects of rainfall manipulation on our two target tree species. Therefore, the analysis of the water relations of these trees was an essential component of the project. Sap-flow within each individual target tree was monitored through the use of Granier probes. These monitoring efforts provided a window on processes such as transpiration and the night-time re-filling of the xylem tissue. Drought tolerance and adaptation strategies were also explored by comparing differences in sap-flow rates across treatment types and between species.
In total, our study site consisted of 12 experimental plots located in three replicate blocks that varied in slope % and aspect. Slope varied from 0-2% in experimental plots situated in level portions of the site, with steeper grades ranging from 6-18% for plots established on hill-slopes. Soil depth across the site ranged from 20 to ≥ 100 cm, with shallower soil depths occurring on hill-slopes where depth to caliche and/or bed-rock was only 20-30 cm in some instances.
The study utilized four different experimental treatments applied in three replicate blocks. The four experimental treatments included 1) un-manipulated, ambient control plots, 2) drought plots, 3) supplemental irrigation plots, and 4) cover-control plots that have a similar infrastructure to the drought plots, but remove no precipitation. The three replicated blocks differed in their slope and aspect. One block of four plots was located on south facing slopes, one on north facing slopes, and one in a flat area of the landscape.
Experimental Treatment Design
To effectively reduce water availability to trees, we installed treatments of sufficient size to minimize tree water uptake from outside of the plot. Thus, we constructed three replicated drought structures that were 40 m × 40 m (1600 m2). We targeted a 50% reduction in ambient precipitation through water removal troughs that covered ~50% of the land surface area. Drought plot infrastructure was positioned to insure that targeted Piñon pine and juniper were centrally located within each drought plot to provide the maximum distance between tree stems and the nearest plot boundary. Each drought and cover-control plot consists of 27 parallel troughs running across the 40 m plot. Each trough was constructed with overlapping 3ft ×10 ft (0.91 m × 3.05 m) pieces of thermoplastic polymer sheets (Makloron SL Polycarbonate Sheet, Sheffield Plastics Inc, Sheffield, MA) fixed with self-tapping metal screws to horizontal rails that are approximately waist height and are supported by vertical posts every 2.5-3.5 m. The plastic sheets were bent into a concave shape to collect and divert the precipitation off of the plot. The bending and spacing of the plastic resulted in 0.81 m (32 in) troughs separated by 0.56 m (22 in) walkways.
Individual troughs often intersected the canopy of trees because of their height. The troughs were installed as close to the bole of the tree as possible without damaging branches in order to maximize the area covered by the plastic across the entire plot. An end-cap was attached to the downstream edge of the trough to prevent water from falling onto the base of the tree. The end-caps were 81 cm × 30 cm and made with the same plastic as the troughs. Each end-cap was fixed to the trough with a 75 cm piece of 20 gauge angle iron cut to match the curve of the bottom of the trough and held in place with self-tapping screws. The plastic junctures were then sealed with acrylic cement (Weld-On #3 epoxy, IPS Corp., Compton, CA). The middle of the end-cap was fitted with a 3 in (7.62 cm) PVC collar to allow water to flow through. A piece of 3 in (7.62 cm) PVC pipe or suction hose (used when the bole of a tree was directly below trough) was then attached to the downstream side of the end-cap, enabling water to flow into the trough on the other side of a tree. End-caps were also placed at the downhill end of the troughs on the edge of the plot and fitted with 90o fittings to divert water down into a 30 cm2 gutter (open on top) that ran perpendicular to the plot. Collected water was then channeled from the gutter into adjacent arroyos for drainage away from the study area.
We built cover-control infrastructures to investigate the impact of the plastic drought structures independent of changes in precipitation. This was necessary because of the high radiation environment in central New Mexico, in which the clear plastic troughs can effectively act as a greenhouse structure. The cover-control treatment had the same dimensions as the drought plots with one key difference. The plastic was attached to the rails in a convex orientation so precipitation would fall on top of the plastic and then drain directly down onto the plot. The cover-control plots were designed to receive the same amount of precipitation as un-manipulated ambient plots, with the precipitation falling and draining into the walkways between the rows of troughs. Cover-control plots were constructed between June-21-07 and July-24-07; drought plots were constructed between August-09-07 and August-27-07. The total plastic coverage in each plot is 45% ± 1% of the 1600 m2 plot area due to the variable terrain and canopy cover. A direct test of the amount of precipitation excluded via the plastic troughs was performed over a 2-week period during the summer monsoon season of 2008. Two rainfall collection gutters (7.6 cm width, 6.1 m length) were installed in a perpendicular arrangement across four plastic drought structures and four intervening open walkways. One gutter was located below the troughs (~0.6 m above ground), and the other was located just above (~1.35 m) and offset, to determine the interception of rainfall by the troughs. Rainfall totals collected via the perpendicular gutters were measured using Series 525 tipping bucket rain gauges (Texas Electronics, Dallas, TX).
Our irrigation system consisted of above-canopy sprinkler nozzles configured to deliver supplemental rainstorm event(s) at a rate of 19 mm hr-1. Our irrigation system is a modified design of the above-canopy irrigation system outlined by Munster et al. (2006). Each of the three irrigation plots has three 2750 gal (10.41 m3) water storage tanks connected in parallel. These tanks were filled with filtered reverse osmosis (RO) water brought to the site with multiple tractor-trailer trucks. During irrigation events, water is pumped from the tanks through a series of hoses that decrease from 7.62 cm (3 in) main lines out of the tank to 2.54 cm (1 in) hoses attached to 16 equally-spaced sprinklers within the plot. Each sprinkler is 6.1 m (20 ft) tall (2-3 m higher than mean tree height), and fitted with a sprinkler nozzle that creates an even circular distribution of water with a radius of 5 m on the ground. Due to the varying topography, sprinklers located downslope (if unregulated) would receive more pressure than those at the top of a hill and thus spray more water. To mitigate this problem, each sprinkler line was fitted with a pressure gauge and variable globe valve (inline water spigot with precise regulation) equidistant from the top of the sprinkler. Each sprinkler line was then set so that the pressure gauges were equal, thus ensuring equal distribution of water throughout the plot, regardless of elevational differences. The irrigation systems were tested in October 2007 (2 mm supplemental), and full applications (19 mm) were applied in 2008 on 24-June, 15-July, and 26-August. During the 24-June event, we deployed six ~1 m2 circular trays across one of the irrigation plots to test the spatial variation of the wetting. Data from this test indicated that on average, collection trays received 19.5 (± 2.5) mm of water.
Site Abiotic Monitoring
We utilized Campbell Scientific dataloggers to continuously monitor and record abiotic conditions and physiological measurements across the site. All systems were connected to a solar-powered wireless network with NL100 relays (Campbell Scientific, Logan, UT). Plots were instrumented with CR-1000 dataloggers (Campbell Scientific, Logan, UT). Each CR-1000 datalogger was accompanied by AM25T and AM 16/32 multiplexers to expand sensor measurement capacity (Campbell Scientific, Logan, UT). The south facing block of experimental plots (the intensive physiology block) was extensively instrumented with sensors to measure both abiotic and plant physiological parameters. The non-intensive experimental plots in the north facing and flat blocks were initially only instrumented to monitor abiotic conditions.
Plant Physiological Response
Multiple physiological characteristics of ten target trees (five piñon and five juniper) within each of the intensive measurement plots were continually monitored by automated sensors. Stem sap-flow (JS) was measured using Granier heat dissipation sap flow sensors installed in 2007 in each intensive physiology plot within the south aspect block. Trees in north facing (plots 5-8) and flat blocks (plots 1-4) were not instrumented with sap-flow sensors during the 2007 or 2008 seasons. All target trees had two 10mm Granier sap-flow sensors installed in the outermost sapwood (Granier 1987). Each sensor used the traditional two probe heated and unheated reference design (Granier 1987), with two additional probes located 5 cm to the right side of the primary probes to correct for axial temperature gradients in the stem (Goulden and Field 1994). We found that this compensation for axial temperature gradients is critical to reduce measurement noise resulting from the open-canopy and high radiation environment of this ecosystem. In addition, stems were wrapped with reflective insulation (Reflectix Inc., Markleville, IN) in an effort to shield sap-flow probes from short term ambient temperature fluctuations and direct solar irradiance. Sapflow (JS) was calculated according to the methods outlined in Granier (1987) and Goulden and Field (1994). Sapwood depth was generally greater than 10 mm on the majority of instrumented trees, thus only a small % of measurements required a correction due to sensor installation in non-functional stem heartwood (see Clearwater and others 1999). All data from sap-flow sensors was recorded using Campbell Scientific AM16/32 multiplexers and CR1000 dataloggers (Campbell Scientific, Logan, UT).
Statistical tests for treatment differences in stem sap-flow (JS) were analyzed using a linear-effects mixed model with repeated measures. For the repeated measures analysis, an autoregressive first order [AR(1)] covariance structure was utilized. Differences between groups and/or treatment means were deemed significant at a threshold α-value of p = 0.05. All mean values are reported with ± 1 S.E.
Data Processing and Quality Assurance & Control (QA-QC)
Data processing and QA-QC were performed using either Matlab (The Mathworks, Inc.) or Microsoft Office 2010 Excel (Microsoft Corporation) software. All raw and/or processed data traces were visually plotted and inspected for noisy, erroneous, or out of range data points or sensors traces. All removed data points had a “NaN” value assigned. Despite this QA-QC review and data cleaning, all data sets should still be evaluated for outliers, etc., as standard outlier statistical tests were not performed.
Additional notes: The Sapflow Js data-set contains 15 minute interval data from 2006 thru 2012. Data Qa/Qc has been performed on these files. PJ day refers to days since start of project (i.e., 1/1/2006). PJ Timestamp denotes/records each 15 minute interval entry from 1/1/2006.
The treatment classes provided in the file are as follows; ambient (1), drought (2), cover control (3), and irrigation (4). The experiment used plot aspect as the blocking factor. There are 3 different replicate blocks and block classifications designated in the files; flat aspect (1), north aspect (2), and south aspect (3). This will be obvious when viewing the files.
The remaining cols in the data frame are the 15 minute sap-flow data for each tree in a particular plot. This type of variable is commonly referred to as sapflow density, and it is represented by the symbol/abbreviation JS, in units of (g/m2 s).
Tree numbers are always grouped by species as follows (regardless of plot); Trees 1-5 are original Pinus edulis, Trees 6-10 are original Juniper monosperma. When one of these original trees died, an additional tree in the plot was added to retain an adequate sample size over time (i.e., multiple years+). These additional trees are grouped as follows; Trees 11-15 are “replacement” Pinus edulis, Trees 16-20 are “replacement” Juniper monosperma. “Replacement” is used here in a more restricted sense, as these additional trees have their separate and unique tree designation number.
So, in differing plots you will have differing numbers of trees depending on; 1) the number of trees for which data was collected, and 2) how many additional “replacement trees” had to be designated due to mortality (or partial mortality) of original trees. Many plots have n=10 trees, based on the original T1-T5 & T6-10 designation, as these particular plots did not experience mortality. However, a plot like P10 has a total of n=16 trees. In P10, the original T1-5 & T6-T10 trees are listed, a replacement Pinon (T11) is listed, and five additional/replacement junipers (T16-T20). In some cases you will see data present at the same time for both original and replacement junipers (plots 6 & 10). This is fine, as juniper experiences a slow/partial canopy dieback, so we monitored the original and replacement trees at the same time in these two plots. Finally, we only provide data on trees for which data was collected (so for example, in some instances you may only have n=4 cols of data for a particular species in a particular plot).