Disturbance often increases local-scale (α) diversity by suppressing dominant competitors. However, widespread disturbances may also reduce biotic heterogeneity (β diversity) by making the identities and abundances of species more similar among patches. Landscape-scale (γ) diversity may also decline if disturbance-sensitive species are lost. California’s vernal pool plant communities are species-rich due in part to two scales of β diversity: (1) within pools, as species composition changes with depth (referred to here as vertical β diversity), (2) between pools, in response to dispersal limitation and variation in pool attributes (referred to here as horizontal β diversity). We asked how grazing by livestock, a common management practice, affects vernal pool plant diversity at multiple hierarchical spatial scales. In terms of abundance-weighted diversity, grazing increased α both within local pool habitat zones and at the whole-pool scale,  as well as γ  at the pasture scale without influencing horizontal or vertical β diversity. In terms of species richness, increases in α diversity within habitat zones and within whole pools led to small decreases in horizontal β diversity as species occupancy increased. This had a dampened effect on species richness at the γ  (pasture) scale without any loss of disturbance-sensitive species. We conclude that grazing increases species richness and evenness (α) by reducing competitive dominance, without large disruptions to the critical spatial heterogeneity (β) that generates high landscape-level diversity (γ).  

Site selection:

Our study took place at Rancho Seco (38.34˚ N, -121.11˚ W), a 458.10-ha conservation site in Northern California. Rancho Seco is located on a high-terrace alluvial formation that hosts Northern Hardpan Vernal Pools on Redding Gravelly Loam and Corning Complex soils (USGS SoilWeb) (Figure 1). The climate is Mediterranean with an average annual precipitation of 526.2 mm per water year (1 Oct – 30 Sep, CIMIS Weather Station, 21-year avg. 1997-2018, Fair Oaks, CA). Annual plants germinate with the first significant fall rains (generally Oct.-Nov.) and flower as the rainy season ends (Apr.-May), and seeds are dormant through the dry summers.  Our study included the last 2 years of a multi-year drought:  2014-15 (39.06 cm, 75.27% of 21-year avg.), the slightly wetter year of 2015-2016 (43.60 cm, 82.83% of 21-year avg.), and the extremely wet year of 2016-2017 (93.06 cm, 176.84% of 21-year avg.) (based on the Oct 1-Sep 30^th  water year, CIMIS Weather Station, 1997-2018, Fair Oaks, CA). Pool standing water depths vary greatly both between pools and within pools between years. The pools at our site ranged from maximum water depths of  0.00 (no water) to 38.00 cm  across all pools across all three wet seasons (Oct-May).  The site includes a 20.9 ha pasture, where grazing has been in place for 150 years, and the current regime is 1 Animal Unit (AU) per 2.4 ha, where AU is defined as the forage demand of a 450 kg-cow. While typical stocking rate varies greatly by region (Herrero-Jáuregui & Oesterheld 2017),this regime is moderate for conservation grazing  and typical for vernal pool landscapes in this area (Marty 2015, George et al. 2016). In montane vernal pool landscapes, this stocking rate may be higher (1 AU/1.68 ha) (Merriam 2017). This site also includes an adjacent ungrazed pasture of 24.35 ha from which cattle were removed 40 years ago when a fence was built to delineate property management boundaries.

In winter 2014, we selected 14 pools each from the grazed and ungrazed areas that spanned two soil types, Corning Complex and Redding Gravelly Loam, (USGS SoilWeb) and a range of pool characteristics affecting plant communities, including, size, shape and slope around the pool perimeter (Gerhardt and Collinge 2003). We matched each grazed pool with an ungrazed pool with as many similar key characteristics as possible (Appendix S-1: Table S1).  

We were interested in the effects of grazing at the pasture scale in addition to the pool (4-6800 m²) and local (<1 m²) scale. To achieve this, we chose a site in which grazing was applied at the pasture level rather than in a spatially random pattern, typical of many grazing experiments. We therefore expected to see some spatial autocorrelation across the whole site driven by vegetation differences between the grazed and ungrazed pastures.  Within grazing treatments, however, we also wanted to ensure that the similarity between any set of pools (horizontal β diversity) that we observed were not simply due to their spatial proximity. To determine whether spatial autocorrelation needed to be accounted for in our analyses, we conducted a partial Mantel test using spatial coordinates of each pool centroid. After accounting for grazing treatment, we found no significant spatial pattern in community composition (Mantel statistic based on Pearson’s product-moment correlation = 0.09, P = 0.10; Appendix S1: Figure S1). Thus, we can rely on our multivariate analyses to assess differences in plant composition that are not confounded by spatial proximity.

Vegetation Sampling:

We followed established sampling methods for vernal pools that stratify based on vertical habitat zones and randomly sample within each zone (Marty 2005, Gerhardt & Collinge, 2007). In early spring 2015, after the pools dried down and before forb taxa were identifiable, we delineated three vertical habitat zones (inundated, transition, and upland) by recording slope and visual differences (e.g., soil color and texture, algal mats, water marks, and matted litter/vegetation) that indicated differences in inundation time. Two water lines were visible in each pool—one low-elevation distinct line indicating inundation throughout the season, and another, fainter high-elevation line suggesting  peak wet-season inundation level . We delineated the lowest point in the pool up to the inner line as the ‘inundated’ zone and the area between the two lines as the ‘transition’ zone.  We delineated the ‘upland zone’ as the area outside the basin within 5-m of the upper edge of the transition zone, beyond which we expect little interaction with the vernal pool ecosystem (Marty 2005, Harpel 2008). Biweekly from March-May, we visited each pool and tracked the phenology of forb species. When we determined that a pool had reached ‘peak flowering’ in which the majority of forbs were blooming and identifiable, we placed quadrats in randomly chosen locations within each zone. Each quadrat was 50x50 cm, divided into 100 5 x 5 cm squares. We recorded the number of cells in which each species occurred. Each year, new locations were randomly chosen for the quadrats within each habitat zone in each pool. Due to the short phenological sampling window, we were limited to three quadrats per zone in each pool (9 quadrats per pool, 216 quadrats/year total).