Understory Vegetation Response to Thinning Pinyon-Juniper Woodlands

ABSTRACT Portions of the western United States, extending from west Texas to eastern Oregon, are thin to densely populated with juniper (Juniperus spp.) woodlands. Managing tree densities, such as through mechanical thinning, has numerous ecological implications for rangeland watersheds and wildlife habitat. The objective of this study was to determine vegetation response to juniper removal in a pinyon (Pinus edulis Engelm)- and juniper (Juniperus monosperma Engelm)- dominated landscape in north-central New Mexico. Herbaceous cover and standing crop were collected from six 1.00- to 1.35-ha catchments situated within the New Mexico State University (NMSU)-Santa Fe Ranch, Santa Fe County, New Mexico, before treatment and at 5 and 10 yr following juniper removal. About 70% of mature juniper trees were cut in 2009 in three of the catchments (treated) while the remaining catchments were left as controls (untreated). Herbaceous cover and standing crop were measured in 2009 (pretreatment), 2014, and 2019 (post treatment) to test our hypothesis that juniper reductions would significantly increase herbaceous standing crop. After 10 yr, gravel, grass, and forb cover were comparable in the treated and the untreated, whereas bare soil cover was significantly decreased in the treated (30% ± 2.5) compared with the untreated (42% ± 2.5). Litter cover in the treated was higher (18% ± 2.6) than the untreated (5% ± 2.6). Herbaceous standing crop significantly increased in the treated compared with the untreated. Standing crop of grass in the treated was two to three times greater than the untreated. Forb standing crop was not influenced by treatment but increased over time in both treatments. Study findings provide land managers with critical information regarding one-seed juniper clearing effects on herbaceous vegetation response in the warm-climate rangeland ecosystems of the southwestern United States.


Introduction
Expansion of woody vegetation cover in rangelands has been observed globally the past century ( Archer et al. 2017 ; Van Auken 20 0 0 ). Contributing factors that led to the recent expansion of woody plants include fire suppression, climatic variations, unsustainable grazing, and higher levels of atmospheric CO 2 ( Archer et al. 2017 ). Pinyon (Pinus edulis) −Juniper ( Juniperus spp.) (P-J) woodland encroachment is common throughout the western United States Romme et al. 2009 ). This portion of the continent has seen an increase in density and spatial cover of P-J trees, covering 24 million ha ( Miller and Wigand 1994 ). In New Mexico alone, P-J and juniper-only woodlands cover an estimated 4 million ha ( O'Brien 2003 ). Establishment of one-seed Juniper ( Juniperus monosperma Engelm) into degraded grassland represented the majority of woodland expansion within north-central New Mexico ( Jacobs et al. 2008 ).
The expansion of woody vegetation can positively and negatively impact ecosystems. Juniper woodlands provide essential ecosystem services including biodiversity, aesthetics, and economic products ( Romme et al. 2009 ), such as firewood, fencing posts, and feedstock for commercial energy production   Juniper woodlands also provide habitat for wildlife ( Belsky 1996 ;Dobkin and Sauder 2004 ;Neff et al. 2009 ) and possibly act as a carbon sink ( Neff et al. 2009 ; Barger et al. 2011 ). Conversely, juniper encroachment may decrease understory herbaceous and shrub cover and decrease biodiversity by reducing available resources for plant growth such as water and nutrients ( Jameson 1967 ;Vaitkus and Eddleman 1987 ;McPherson and Wright 1990 ;Ernst and Pieper 1996 ;Brockway et al. 2002 ) and alter hydrologic processes ( Owens et al. 2006 ;Petersen and Stringham 2008 ;Mollnau et al. 2014 ;Zou et al. 2014 ;Ochoa et al. 2018 ) and soil nutrient cycling ( Bates et al. 2002 ). However, the impacts of juniper encroachment vary depending on the environment, abiotic factors, and spatial and temporal scale.
In New Mexico, impacts of woodland expansion led land managers to reduce woodland stands using various methods, such as bulldozing, chaining, thinning, and prescribed burning ( Evans 1988 ). Reducing woodland density has been done for multiple objectives, including improving herbaceous production for livestock, increasing streamflow on a watershed, reducing fire severity, improving soil health, and enhancing wildlife habitat Kerns and Day 2014 ;Bombaci and Pejchar 2016 ;Williams et al. 2018 ;Fick et al. 2022 ).
The objective of this study was to assess herbaceous vegetation response to one-seed juniper ( Juniperus monosperma Engelm) removal in P-J woodlands of north-central New Mexico. Several studies ( Schott and Pieper 1985 ;Pieper 1990 ;Jacobs and Gatewood 1999 ;Mason et al. 2009 ;Garduno et al. 2010 ;Fernald et al. 2022 ) have evaluated vegetation overstory-understory relationships in New Mexico. However, this study looked at relationships spanning a decade in multiple catchments to monitor juniper removal effects on forage response. This experiment sought to answer overstoryunderstory relationship questions at spatial (multicatchment) and temporal (a decade) scales relevant to land management. It was hypothesized that herbage standing crop would increase by reducing juniper density.

Study site
The study area was located on the New Mexico State University (NMSU) −Santa Fe Ranch, located approximately 16 km northwest of Santa Fe, New Mexico (latitude 35 °44 13.86 N, longitude 105 °04 53.89 W). Historically, the study site was dominated by warm-and cool-season grasses before shrub and woody species encroached ( USDA-NRCS 2021 ). Soils at this site are moderately deep to very deep (50 −203 cm deep), and the surface and underlying layers are gravelly or very gravelly loam, sandy loams, and fine sandy loam ( USDA-NRCS 2021 ). The study site elevation ranged from 1 939 to 1 977 m. The variation of slopes was such that the bottom of the valley ranged from 2% to 5%, while the slope on the hillsides varied from 20% to 50%. Annual precipitation in this area averaged 273 mm over 14 yr (from 2006 to 2019) and ranged from 208 to 368 mm ( Figs. 1 and 2 ). It occurred predominantly during two periods: in the form of rainfall from May through October and as snow from December through March. About 75% of the annual precipitation at the study site was from high-intensity, short-duration thunderstorms that typically occur between March and October. Precipitation and temperature data were retrieved from the National Oceanic and Atmospheric Administration National Centers for Environmental Information Climate ( www.ncdc.noaa.gov/cdo-web/ ). The precipitation data were downloaded from a weather station located about 1 km north of the study site, while the temperature data were downloaded from a weather station located 14 km south of the study site. The daily records from 2006 to 2019 were used to evaluate mean an- . Precipitation data were retrieved from a weather station, located 1 km north from to the study site.

Treatment
A paired catchments approach was used to investigate herbaceous vegetation response to juniper removal in six catchments ranging from 1.00 to 1.35 ha in area. Tree basal area and frequency were measured in all catchments in 2009, pretreatment. Average basal area of the untreated catchments was 1 402 (m 2 /ha) and 1 222 (m 2 /ha) for the treated catchments, while tree density was 247 tree/ha and 229 tree/ha in the untreated and the treated, respectively. Juniper tree frequency was 76% to 82%, while pinyon (Pinus edulis) was 18% in all sites. Gambel oak tree (Quercus gambelii) was found in two treated catchments with 6% tree frequency. In 2009, about 70% of juniper trees were cut and removed in three randomly selected catchments (hereafter, treated), while the remaining three were left as control catchments (hereafter, untreated) ( Fig. 3 ). Juniper trees were cut using chainsaws between May and July in 2009. The tree boles were removed while branches were used to fill in ravines to slow runoff and erosion. Branches were also scattered across the site after cutting.
Cattle grazing on the specific study sites was limited throughout the duration of the study due to the distance (about 3.3 km) from watering points. Grazing was only allowed for a few years in the surrounding pastures, except in 2019 when cattle of the NMSU Chihuahuan Desert Rangeland Research Center were brought to the study site due to drought-driven, low-forage production in the southern portion of the state. No specific stocking density inside the catchments was documented; it was assumed that because of the extended distance from the study site to the watering points, they did not access the site frequently and had minimal to no impact on the herbaceous standing crop.

Vegetation sampling
Herbaceous cover and standing crop were evaluated in April 2009 (pretreatment) and then in October 2014 and October 2019 (post treatment) in the treated and untreated catchments. Three 50-m transects were established and used to evaluate the vegetation dynamics all 3 yr. Each transect was installed perpendicular to the direction of water drainage, beginning on the south face and The precipitation and temperature data were retrieved from different weather stations from Santa Fe, New Mexico. The precipitation data were from a weather station located about 1 km north from the study site, while the temperature data were from a weather station located about 14 km south from the study site.
ending on the north face of the catchments traversing the valley bottom. The point intercept method was used to determine ground cover ( Bonham 1989 ;Elzinga et al. 1998 ) by dropping a pin flag vertically, without guidance, from the same height each time noting the category of interest (bare soil, gravel, litter, grass, and forb) at each point. This occurred along the transect at a 1-m interval, giving 50 points per transect. Cover classes included soil, gravel (more than 1-inch diameter), litter, grass, and forb.
Herbaceous standing crop was determined by functional group (grass and forb) using the direct method of clipping ( Bonham 1989 ). At five 0.25 m 2 quadrats, clippings were made at 1 inch above ground level within 5 m of the transect tape, with a 10m interval, thus producing a total of five quadrats per transect. All plant species of the same functional group were combined to get total dry matter that was measured by drying samples at 55 °C −65 °C for 48 h.

Data analysis
Proc Mixed was used for repeated measures analysis of variance using SAS software ( SAS Institute. 2013 ) to compare surface cover and standing crops between the treated and untreated catchments. Treatment type (treated, untreated, df = 1); time period (2009,2014, and 2019, df = 2); and their interaction (df = 2, with the error term df = 8) were placed in the model statement. Time of sampling was the repeated variable, and site nested with treatment was the subject. The experimental unit was the catchment, and the observation unit was the transect. All variables were averaged from each catchment, and repeated measures were defined by replication within the catchment by treatment. The level of statistical significance was set at 0.05 for all variables.

Results
Annual precipitation during the length of the study (2009 −2019) was variable. Annual precipitation was above the 14-yr average in 6 yr and below the 14-yr average in 5 yr. However, the annual precipitation of the 5 low precipitation yr was not below 75% of the 14-yr average (see Fig. 1 ). Total summer and fall precipitation contributed the most to the annual precipitation of the years where they received above-normal precipitation (see Fig. 2 ). The fall season of 2013 was colder, and the winter of 2019 was warmer (see Fig. 2 ). In general, annual weather events were within the natural range of variability.
Bare soil cover differed by the interaction between treatment and time ( P = 0.0241). Bare soil cover was slightly less in the treated than the untreated in the second sampling yr (52.7% ± 2.5 and 57.3% ± 2.5, respectively) and significantly lesser in the third sampling yr. At the end of the study, bare soil cover in the treated was 12.4% lower than the untreated (29.8% ± 2.5 and 42.2% ± 2.5, respectively). Gravel cover differed neither between treatments ( P = 0.374) nor across years ( P = 0.334). Litter cover differed by the interaction between treatment and time ( P = 0.0126). Litter cover was greater in the treated compared with the untreated ( P = 0.0070). In yr 5, litter cover in the treated was 12% higher than the untreated (18.4% ± 2.6 and 5.8% ± 2.6, respectively) and remained greater in the treated compared with the untreated in yr 10. Grass cover was influenced by time ( P = 0.0042). Grass cover varied across the sampling period in both the treated and untreated. It ranged from 20% ± 3.8 to 35.1% ± 3.8 in the treated and from 23.6% ± 3.8 to 29.8% ± 3.8 in the untreated. Forb cover differed across years ( P = 0.0 0 01). Forb cover increased post treatment over time in both the treated and untreated compared with the pretreatment condition. However, there were no differences between treatments ( P = 0.1520) ( Fig. 4 ; Table 1 ).
Grass standing crop varied by the interaction between treatments and times ( P = 0.0331). Post treatment, grass standing crop was twofold greater in the treated than untreated in yr 5 (633 kg/ha ± 53 and 305 kg/ha ± 53, respectively) and threefold greater in yr 10 in the treated (330 kg/ha ± 53) compared with the untreated (112 kg/ha ± 53). Forb standing crop was increasing between yr 5 and 10 ( P = 0.0054). Post treatment, forb standing crop in the treated was 22 kg/ha ± 10 in yr 5 and was 44 kg/ha ± 10 in yr 10, while in the untreated, forb standing crop was 34 kg/ha ± 10 in yr 5 and remained constant the following yrs ( Fig. 5 ; Table 2 ).

Table 1
P value for ground cover between the untreated catchments and treated catchments, regarding the main effect (treatment and time) and the interaction between them. Asterisks ( * ) indicate significant main effect and the interactions at P < 0.05.

Response variable
Ground cover

Discussion
The objective of this study was to determine the effect on herbaceous vegetation from thinning one-seed juniper trees in small catchments of P-J woodland. Results supported our hypothesis that reducing tree density would increase herbage standing crops, particularly grass standing crop. However, thinning of tree stands did not appear to increase vegetation cover. Grass cover and forb cover were similar in thinned and unthinned stands in the post-treatment years. However, herbaceous cover was responsive to the year effect, which likely resulted from differing amounts and timing of annual precipitation during the growing seasons in yrs 5 and 10.
Previous studies found an increase in herbaceous cover after reducing tree density ( Bates et al. 20 0 0 ;Miller et al. 2005 ;Owen et al. 2009 ;Ray et al. 2019 ) by improving effective rainfall (less interception), increasing available sunlight for the understory, and reducing competition on soil water and nutrients ( Ernst and Pieper 1996 ;Bates et al. 20 0 0 ;Brockway et al. 2002 ;Huffman et al. 2013 ). However, other studies have shown no significant increase on understory cover on treatment sites ( Clary 1971 ;Evans and Young 1985 ;Huffman et al. 2008 ;Garduno et al. 2010 ;Rubin and Roybal 2018 ). It has been reported that other factors that drive changes to understory cover following tree thinning include treatment type, soil properties, long-term weather patterns, and pretreatment vegetation composition ( Clary 1971 ;Vaitkus and Eddleman 1987 ;Pieper 1995 ;Jacobs and Gatewood 1999 ;Huffman et al. 2013 ;Monaco and Gunnell 2020 ). We suggest that soil properties, particularly soil depth, may have limited herbaceous cover response to tree thinning. A measurable increase in herbaceous cover after reducing juniper trees on deep soil, as found on our sites, may not be detected in semiarid environments. In shallow soils, juniper competition for available soil water and nutrient is stronger, which limits herbaceous understory growth ( Breshears et al. 1998 ;Miller et al. 2005 ). Thus, reducing conifers on shallow soil sites generally results in larger relative increases in herbaceous foliar cover ( Brockway et al. 2002 ).
Bare ground increased in both sites in yr 5, probably because of the dry yr in 2014. Bare ground declined in the thinned treatment in yr 10 because of greatly increased litter cover compared with the untreated. Our findings are in agreement with other studies ( Brockway et al. 2002 ;Fulé et al. 2002 ;Miller et al. 2014 ). The increase in litter potentially provides positives impacts by protecting soil surfaces, thus, reducing runoff and erosion ( Evans 1988 ;Ernest et al. 1993 ). However, litter cover does not seem to enhance herbaceous understory establishment 10 yr after thinning in this woodland environment.
Five and 10 yr after thinning, grass standing crop in the treated was greater than the untreated. Standing crop of grass ranged from twofold to threefold greater in the treated than the untreated. Our findings for standing crop of grass are consistent with previous studies from New Mexico ( Pieper 1995 ;Brockway et al. 2002 ) and Oregon ( Bates et al. 20 05 , 2019 , 20 0 0 ). We assumed the increase of standing crop of grass in the treated is due to a decrease in the competition from trees for soil water and soil nutrient resources. Previous studies found that decreasing competition from juniper trees freed up resources for understory vegetation, including light, nutrients, and increased available soil water ( Jameson 1967 ;Breshears et al. 1998 ;Bates et al. 2002 ;Roundy et al. 2014 ). Although herbaceous understory cover in the treated may not produce a detectable increase in ground cover, grasses tillering increases; thus, there is higher measured standing crop per unit area. For example, a study in northern mixed-grass prairie by Gates et al. (2017) found that fire increased biomass of perennial grass but had no significant effect on basal cover.
The effectiveness of thinning on forb standing crop was not apparent. Standing crop of forbs in the treated remained comparable with the untreated in yr 5 and yr 10. Our findings are in agreement with previous studies Ernst-Brock et al. 2019 ). Bates et al. (2019) stated that perennial forb yield may or may not respond to fire treatment in juniper woodland. It has been mentioned that the perennial forb response to woodland reduction can be related to the type of treatment and plant composition before treatment Roundy et al. 2014 ;Bates and Davies 2017 ;Bates et al. 2019 ;Ernst-Brock et al. 2019 ). Year effects were significant to forb standing crop. We assumed the fluctuation of forb standing crop is related to precipitation amount and timing of precipitation during the growing season.

Implications
This study provides evidence that reducing juniper trees by thinning catchments in northern New Mexico can promote a twofold to threefold increase in grass standing crop. Thus, Juniper tree control in this area can have a long-term effect on the herbaceous standing crop, consequently providing more grazing management flexibility to land managers through increasing grazing capacity and forage for livestock production. Increasing herbaceous standing crop could improve the catchment health and water quality by reducing soil erosion. Even though the thinning treatment positively affects herbaceous standing crop in the long term, the thinning treatment alone could not increase herbaceous cover in this area where one-seed juniper trees have encroached; therefore, including other practices with thinning treatment may be required to accelerate restoration of herbaceous understory and improve rangeland health.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.