tudy System: Location.
We installed infrastructure at the Sevilleta National Wildlife Refuge in central New Mexico, at the northern edge of the Chihuahuan Desert. The experiment is a central component of the Sevilleta Long-Term Ecological Research Program (SEV-LTER), funded by the US National Science Foundation (sevlter.unm.edu). Annual precipitation is 200–300 mm, the majority of which is summer monsoon rainfall.
Study System: Climate.
Water is the most vital resource in drylands, and has high temporal variability at local and regional scales. SEV precipitation is weakly bimodal. Soil moisture accumulates in winter, particularly at higher elevations where snowfall is significant. Warm, dry conditions in late spring create a pulse of snowmelt moisture at high elevation, while severely depleting soil moisture at low elevation. Then, from July through September, the North American Monsoon drives localized convective storms that contribute ~60% of mean annual precipitation. Long-term minimum precipitation is <4mm every month, and precipitation reconstructions using Sevilleta tree-ring record reveal a history of severe drought every 6,070 years since the 17th century. Climate models predict higher winter and summer annual temperature, more frequent and intense El Niño events, declines in winter/spring precipitation and more variable monsoon rainfall. Dryland water availability is determined not only by precipitation inputs but also by the strong influence of temperature on evaporative demand. The Standardized Precipitation Evapotranspiration Index (SPEI), accounts for the influence of temperature on water demand. More negative SPEI values signify drier and hotter conditions. SPEI thus influences the amount and duration of soil moisture. Since 1900, mean SPEI has declined in our region, while variance in SPEI has increased since 1980, a scenario of dual change in climate mean and variance. Climate warming affects trends in long-run mean SPEI, while altered precipitation amplifies variance in SPEI.
Design: Stochastic precipitation variance.
To increase variance in precipitation stochastically, without changing the mean, we paired plots and amplified their precipitation regimes. Specifically, every water year, plots within a pair are randomly assigned to either a 50% decrease or 50% increase in precipitation. This is achieved during our annual MVE flip party, where participants flip coins to determine whether paired plots will retain the same treatments or treatments will be flipped within the pair.
We chose 50% as our target manipulation to amplify variance because a 50% change produced a 66% increase in the coefficient of variation (CV) of precipitation at a site in southern New Mexico, matching projected regional climate extremes (Gherardi and Sala 2015a, b). However, because Gheradi and Sala (2015a) were not testing for antecedent effects, they applied regular alternation between high and low rainfall years, which has the effect of reducing stochasticity, potentially even below naturally occurring levels. Our treatment created stochasticity through random assignments of which plot received extreme high or extreme low precipitation in a given year. Extreme wet and dry years were achieved by covering plots with roof panels that intercepted precipitation year-round using a modified version of a prior design (Gherardi & Sala, 2013). Water was captured from shelters with gutters, stored in tanks, then delivered to the paired plots via solar-powered pumps (Rudgers et al. 2023).
Design: Increasing mean aridity.
To reduce the long-run mean precipitation, we intercepted 25%, a moderate forcing within range of likely climate futures. Control plots received similar shelters but with inverted panels that allowed precipitation through. Plots receiving both reduced mean and increased variance randomly alternated between 75% less net precipitation (-25% for mean - 50% for variance) or 25% more (-25% for mean + 50% for variance). Simulations of historical Sevilleta LTER met data projected that 25% rainfall reductions would reduce SPEI by ~19% while 50% deviations will increase the coefficient of variation (CV) of SPEI by ~53%, without altering its mean under feasible replication (22 plots simulated per biome).
Other design elements.
Replication was uneven to account for higher variability among plots in the increased variance treatment, consistent with initial modeling efforts using our long-term climate data. All 30 plots per ecosystem were hydrologically isolated via aluminum flashing installed to 20 cm depth using a gas-powered trencher, and were co-located with existing meteorological stations.
Sensor network.
In a subset of 18 plots per site, we installed sensors to track soil moisture and temperature at three depths (12.5, 22.5, 37.5 cm; TEROS 11 sensors, Meter Group, 2365 NE Hopkins Ct., Pullman, WA 99163, +1.509.332.2756, metergroup.com). The sensor manuals are available here:
https://publications.metergroup.com/Manuals/20587_TEROS11-12_Manual_Web.pdf
Some values have missing data at ~XX% error rate due to sensor or wireless network temporary failures, rodent damage to wires and cables, or installation of replacement sensors.
Gherardi, L. A., & Sala, O. E. (2013). Automated rainfall manipulation system: a reliable and inexpensive tool for ecologists. Ecosphere, 4(2), 1–10. doi:10.1890/es12-00371.1
Gherardi, L. A., & Sala, O. E. (2015a). Enhanced interannual precipitation variability increases plant functional diversity that in turn ameliorates negative impact on productivity. Ecology Letters, 18(12), 1293–1300. doi:10.1111/ele.12523
Gherardi, L. A., & Sala, O. E. (2015b). Enhanced precipitation variability decreases grass- and increases shrub-productivity. Proceedings of the National Academy of Sciences of the United States of America, 112(41), 12735–12740. doi:10.1073/pnas.1506433112
Rudgers, J. A., Luketich, A., Bacigalupa, M., Baur, L. E., Collins, S. L., Hall, K. M., Hou, E., Litvak, M. E., Luo, Y., & Miller, T. E. (2023). Infrastructure to factorially manipulate the mean and variance of precipitation in the field. Ecosphere, 14(7), e4603. https://doi.org/10.1002/ecs2.4603
