In eastern North America, cucumber-root (Medeola virginiana) is a widely distributed perennial forest herb that has been used as an ecological indicator of white-tailed deer (Odocoileus virginiana) browsing due to its predictable responses to deer exclusion (i.e., increased height, abundance). However, cucumber-root is less likely to occupy sites with high concentrations of soil manganese (Mn), which may limit its utility as an indicator under limiting soil conditions. We examined responses of cucumber-root total counts and flowering abundance to deer exclusion, competitive release from surrounding vegetation, and soil application of dolomitic limestone to determine the relative effects of these treatments over 7 years (2014-2021). Prior to treatment, initial total and flowering abundance were best explained by soil extractable Mn concentration. Post-treatment, fencing best explained increases in total counts, but flowering abundance was most affected by soil extractable Mn and pH. Initial soil Mn concentrations determined the effectiveness of dolomitic limestone application; microplots with moderate to high soil Mn (> 6 cmol_c kg^-1) had increased flowering with increased pH, while flowering decreased on microplots with initially low soil Mn concentrations (< 6 cmol_c kg^-1). We suspect changes to soil chemistry from liming affected plant stress, but that stress was either alleviated or intensified depending on initial soil Mn concentrations. Herbivory is an important driver of plant abundance across our study area but flowering response, a critical component of plant demography, seems to be driven by soil Mn. Cucumber-root may have limited utility as an indicator because soil chemistry mediates flowering responses to deer exclusion.

Study Area and Study Design

The study area was in central Pennsylvania on the Rothrock and Bald Eagle State Forests in the Ridge and Valley Physiographic Province (Fig. 1a). Both state forests are primarily forested with contiguous, oak-dominated hardwood stands classified as red oak (Q. rubra) – mixed hardwood forest (Zimmerman et al. 2012). Overstory species composition varies by elevation, with red oak, chestnut oak (Q. montana), black oak (Q. velutina), and scarlet oak (Q. coccinea) being the dominant overstory trees on ridgetops and red maple (Acer rubrum) and yellow poplar (Liriodendron tulipifera) dominant at lower elevations.

From 100 potential locations across both state forests that span 198 km², 24 sites were randomly selected for plot establishment (Fig 1b). Plots consisted of a network of eleven 14.63 m diameter subplots (1/60^th ha; 1/24^th acre; 168 m²) oriented around a center subplot (subplot 1) at specified azimuths (Fig. 1c). Each subplot contained a single nested circular 3.30 m diameter microplot (1/750^th ha; 1/300^th acre; 13.5 m²) 3.66 m due east of each subplot center (Fig 1c). Microplots were numbered the same as the subplots in which they were nested.

Treatment Applications

Microplots 5-11 at each location received one of seven potential treatment combinations (fence alone, herbicide alone, lime alone, a two way combination of lime + fence, lime + herbicide, fence + herbicide, or a three-way combination lime + fence + herbicide). The fence-only treatment was always at microplot 5, but the remaining treatments were randomly assigned to microplots 6-11. Microplots 1-4 served as controls and had no treatments applied.

Following a vegetation inventory (described below), 1.8 m tall polyvinyl deer fences were installed around all microplots assigned a fence treatment in the summer of 2014 (27 May–14 August) and they were maintained yearly. At the same time as fence installation, pelletized dolomitic limestone (CaMg(CO₂)₃) was applied by hand to each microplot assigned a lime treatment at a rate of 5 Mg ha^-1 (6.7 kg per microplot). A one-time 5 Mg ha^-1 application was chosen to raise soil pH and increase Ca and Mg cation concentrations to offset soil chemistry changes attributed to acid deposition (Drohan & Sharpe 1997) while remaining consistent with other studies (e.g., Moore et al. 2008; Long et al. 2012).

Over an 11-day period from 17 to 28 August 2015 (following the 2015 summer vegetation monitoring season), microplots assigned to an herbicide treatment were broadcast sprayed with 1.9L of a non-targeted herbicide, oil–water emulsification mix of 1.5% v v^-1 ester triclopyr and 3.0% v v^-1 amine glyphosate combined with a 1.5% v v^-1 emulsifier and 0.05% v v^-1 non-ionic surfactant. All understory vegetation in the microplot was sprayed to minimize the variation of the treatment application, and all sprayed vegetation was affected due to the non-selective nature of the herbicide. Our previous work showed that ericaceous vegetation was reduced in response to herbicide treatment (Begley-Miller et al. 2024).

Vegetation Data Collection

Cucumber-root counts in each microplot were recorded across all 24 locations in 2014 (pre-treatment), and again in 2021 (7 years after treatment). Despite their rhizomatous nature, we considered ramets as “individuals” and recorded total counts (flowering and non-flowering combined), and flowering counts (flowering or fruiting individuals only) accordingly.

Soil Sampling and Chemical Analysis

In 2014 (pre-treatment) and 2016 (two years post-lime application), soil samples were collected by morphological soil horizon to a depth of 40cm from microplots 1, 5, 6, 7, 8, 9, 10, and 11 (plus microplots 2–4 when time allowed) at each location. We pooled the Oa and A horizons together for analysis because the A horizon was thin and indistinguishable from the Oa horizon. Soil pH values were measured from each sample using 0.01M CaCl₂ with a 1:5 soil-to-solution ratio (The Pennsylvania State University Agricultural Analytical Services Laboratory; Hendershot et al. 2008).

For 2014 samples, the USDA-NRCS Kellogg Soil Survey Laboratory (KSSL) in Lincoln, NE extracted Ca (cmol_c kg⁻¹) at unbuffered soil pH using ammonium acetate (NH₄OAc) and displacement with 2 M KCl (Hendershot et al. 2008; Soil Survey Staff 2014). They extracted Mn (mg kg⁻¹) separately using 1 N KCl (Long et al. 1997; Long et al. 2011; Soon et al. 2008), and we converted Mn values to cmol_c kg⁻¹ for consistency. For 2016 samples, Ca, Mg, K, Mn, and Al were extracted from 5g of soil for mineral horizons and 1.25g of soil for organic horizons using 1M NH₄Cl with a 6-hour manual vacuum. Detection of elements from the extraction was with ICP-OES with one blank and one standard soil run per 22 samples. 

To ensure consistency in our assessment of soil chemistry, we only included soil parameters from the uppermost organic horizon (the Oa/A horizon) in the analysis for both years (2014 and 2016). Due to the different extraction methods for cations and metals used in 2014 and 2016, we only assessed changes to soil chemistry after application of dolomitic limestone using 2016 control v. treatment data. We report 2014 soil chemistry data in cmol_c kg^-1 and 2016 soil chemistry data in mg L^-1.