Signatures of autumn deluges revealed during spring drought in a semi-arid grassland

Increases in extremely large precipitation events (deluges) and shifts in seasonal patterns of water availability with climate change will both have important consequences for ecosystem function, particularly in water-limited regions. While previous work in the semi-arid shortgrass steppe of northeastern Colorado has demonstrated this ecosystem’s strong sensitivity to growing season deluges, our understanding of ecosystem responses to deluges during the dormant season is limited. Here, we imposed experimental 100 mm deluges (~ 30% of mean annual precipitation) in either September or October in a native C4-dominated shortgrass steppe ecosystem to evaluate the impact of this post-growing season shift in water availability during the autumn and the following growing season. Soil moisture for both deluge treatments remained elevated compared with ambient levels through April as spring precipitation was atypically low. Despite overall low levels of productivity with spring drought, these deluges from the previous autumn increased aboveground net primary production (ANPP), primarily due to increases with C4 grasses. C3 ANPP was also enhanced, largely due to an increase in the annual C3 grass, Vulpia octoflora, in the October deluge treatment. While spring precipitation has historically been the primary determinant of ecosystem function in this ecosystem, this combination of two climate extremes – an extremely wet autumn followed by a naturally occurring spring drought – revealed the potential for meaningful carryover effects from autumn precipitation. With climate change increasing the likelihood of extremes during all seasons, experiments that create novel climatic conditions can provide new insight into the dynamics of ecosystem functioning in the future.

Methods
Methods

Site Description

Our study site was in the North American shortgrass steppe at the USDA Central Plains Experimental Range in northeastern Colorado, USA (N 40.8422°, W 104.7156°). Mean annual temperature is 8.6°C (Lauenroth and Sala 1992), and mean annual precipitation is 342 mm with only ~20% of annual precipitation occurring during the autumn season (Sept. – Nov.; Hermance et al. 2015). Growing season precipitation (May – Aug.) accounts for majority (~75%) of the annual total. Long-term ANPP is ~75 g m-2 with the majority comprised of perennial C4 grasses (~75%, primarily Bouteloua gracillis and Bouteloua dactyloides) and the remainder including perennial C3 grasses, cool-season annual grasses, forbs, and cacti (Hoover et al. 2021). Soil texture at the study site is sandy clay loam (Sala et al. 1992), and although the site has a history of moderate cattle grazing, the area where this experiment was conducted has been protected from large ruminant grazers since 2011 (Post and Knapp 2020).

Experimental Design

At the beginning of Sept. 2021, we established thirty 1-m2 plots separated by at least 3 m. Plots were randomly assigned to one of two deluge treatments or the ambient group (AMB), which received only ambient precipitation. Both deluge treatments included the addition of 100 mm of water applied over six days during the autumn of 2021 – either in Sept. (SEP) or in Oct. (OCT). The timing of these deluges, early and late fall, captured partial vs. complete senescence of the dominant grasses aboveground. Using long-term precipitation data from this site (Hoover and Derner 2020), this amount of precipitation is equivalent to two of the wettest autumn months on record (Sept. 1989 and 2013), each of which received approximately 100 mm of precipitation.

            Water was applied between Sept. 15-20 and Oct. 20-25 for the SEP and OCT treatments, respectively, using a hand-held watering wand attached to a flow meter and water pump. Over the week, 8 mm of water was added on the first day, 24 mm on the second, 18 mm on the third, and then 50 mm was applied on day 6. This water application pattern is similar to a 7-day high rainfall period observed in Sept. 1989, where rain accumulated for seven days with 50 mm recorded on the final day. To prevent overland flow off the plots, water was applied several times each day with only 4-8 liters of water added at a time. Added water was tested for nitrogen levels (American Agricultural Laboratory, Inc., Nebraska, USA) and was below US EPA drinking water standards. Besides the deluge events, all plots received ambient precipitation (which was well below normal – see Results).

Measured Responses

            From Sept. 2021 – Aug. 2022, we monitored soil moisture, soil CO2 efflux (soil respiration), and canopy greenness. From Sept. – Nov. 2021 and mid-Mar. – June 2022, measurements were made weekly, but during the winter, measurement frequency was more variable, ranging from biweekly to monthly intervals. This schedule allowed us to capture key phenological stages including autumn senescence and spring green-up with higher resolution. In addition, we also measured soil nitrogen availability during two separate sampling periods – the autumn and the spring – to assess both treatment and potential carryover of soil nitrate (NO3-) and ammonium (NH4+). Finally, above-ground net primary production (ANPP) was measured in mid-June to capture early-season plant growth responses.

            Soil moisture (volumetric water content, % VWC) was measured using a 20 cm handheld soil moisture time-domain reflectometry probe (Campbell-Scientific Hydrosense II), which integrates soil moisture across the top 20 cm of soil. Because most of the root biomass in this ecosystem is within the upper 20 cm of soil (Milchunas and Lauenroth 1989; Gill et al. 1999; Nelson et al. 2004), this measurement provides a reliable indicator of water available for the dominant grasses. We also used a Sentek Diviner probe (Diviner 2000, Sentek Pty Ltd.) on a subset of the plots to observe soil moisture dynamics to greater depths (0-60 cm with measurements at 10 cm increments).

            We monitored canopy phenology with digital repeat photography to estimate canopy greenness (following the methods of Post and Knapp 2019; 2020; Hoover et al. 2022). For each photograph, we placed an iPhone camera directly above a movable 50 x 50 cm frame positioned in the corner of each plot; this image was then cropped to only include the interior area of the frame. The cropped photos were then analyzed with the R package EBImage (Pau et al. 2010), which calculates the average green chromatic coordinate (GCC) index (Filippa et al. 2016). The GCC index accounts for variation in image lighting by computing the greenness relative to the total brightness of each pixel as green / (red + blue+ green) (Filippa et al. 2016).

            To measure soil CO2 efflux, we installed PVC collars (10 cm in diameter, n=5 per treatment) in bare areas between grasses at the beginning of September (2.4 cm belowground, 2 cm aboveground). If any plant growth occurred within the collars during the measurement period, it was removed (clipped to the base) before measurement. We used a 6400-09 soil flux chamber attached to an LI-6400XT (LiCor., Inc, Lincoln NE, USA) to measure soil respiration. Measurements were taken between 8:30hr – 12:30hr local time at ambient CO2 concentration, humidity, and temperature

            Nutrient availability was assessed using Plant Root Simulator probes (PRS, Western Ag Innovations Inc., Saskatoon, Canada) in both the autumn of 2021 and the spring of 2022. These 15 cm x 3 cm probes, buried in the soil with minimal disturbance, utilize either a cation or anion exchange membrane. For the autumn period, one set of probes (anion and cation) was inserted on Sept. 9 and removed on Dec. 3, 2021, in 23 plots (n=8 for SEP and OCT, n=7 for AMB). During the spring period, three sets of probes per plot were inserted on Mar. 16, 2022, and removed on June 22 in the same 23 plots. After removal, samples were washed using deionized water and sent to Western Ag Innovations for extraction. Probes were analyzed for ammonium (NH4+) and nitrate (NO3- ) with a Technicon Autoanalyzer. Values were reported in µgrams 10 cm-2 11 weeks-1.

            Finally, ANPP was estimated in mid-June (June 14-16) in all plots to capture early-season growth (before potential mid- and late-summer rains fell). We harvested all aboveground vegetation within two 0.1 m2 subplots to ground height, sorted by functional group (C3 grass, C4 grass, forb, or annual C3 grass). Vegetation was then dried at 60 °C for 48 hours before being weighed to the nearest 0.01g. The previous year’s growth was easily distinguishable from the current year's growth and was excluded.