Abstract
Cover crops have been used to build soil health and improve ecosystem services in agricultural fields and pastures, but they have not been tested in restoration contexts. We conducted two experiments in interim oilfield reclamations at Ft. Berthold Indian Reservation in North Dakota: in 2014 two different perennial grass mixes, with and without an oat cover crop and, in 2015, a single grass mix with and without a cover crop cocktail. To determine whether cover crops speeded site recovery or whether they competed with perennial grasses, all sites were assessed for plant establishment and rangeland health in August 2015. Sites planted in 2014 ranged from 20 (± 10 SD) to 38 (± 0.5) perennial grass plants/m2. Sites planted in 2015 ranged from 6 (± 1.8) to 29 (± 11) plants/m2. Cover crop treatment and grass mix treatments were not significant determinants of perennial grass establishment (p > 0.05). Soil nutrients appeared to drive early revegetation establishment: sites with poor perennial grass establishment had lower levels of phosphorous and higher levels of calcium, iron and manganese. Rangeland health trended towards being greater when a cover crop was planted, but the effect was very small. We will eventually test whether the long-term benefits of cover crops in agricultural systems transfer to restoration, but when cover crops establish at low densities, as we observed in these studies, they may only have small effects in reclamations.
Cover crops have been used in agronomic situations to improve soil, air and water conservation and quality, nutrient scavenging, nutrient cycling and management, and pest control (Delgado et al. 2007). Cover crops can add carbon and improve soil structure (Reicosky and Forcella 1998) and specific functional groups of cover crops can provide nitrogen, weed control or rapidly build up soil organic matter (Snapp et al. 2005). Mixing functional groups may provide multiple environmental services, while also buffering against adverse weather events (Wortman et al. 2012). However, cover crops may not be suitable in water-limited environments (Robinson and Nielsen 2015). Despite their increasing popularity in cropping systems, there is little information about using cover crops in reclamation. Existing research in western North Dakota has shown little effect of an annual oat cover crop in grazed pipeline reclamations (Espeland and Perkins 2013) but a positive effect of a native annual cover crop in roadside reclamations (Espeland and Richardson 2015). Because cover crops can reduce bare soil and build soil structure needed in reclamation projects, they are attractive for reclamation. The contribution of cover crops to oilfield reclamation may be evaluated relatively quickly using rangeland health indicators, which assess ecological processes and rate three attributes of rangeland health: soil/site stability, hydrologic function, and biotic integrity (Pellant et al. 2005).
Western North Dakota has a history of high anthropogenic impacts (Shortridge 1988, Hudson 2003) that most recently include the Bakken oil boom. The region spans both glaciated plains to the north and east of the Missouri River and unglaciated plains to the south and west. The glaciated plains have an agricultural history starting with pre-colonial agriculture of the Mandan and Hidatsa peoples (Flynn and Syms 1996) and are currently in high-value crop production (Hill and Olson 2013). Livestock dominates the agricultural history of the unglaciated plains: a conversion from bison to cattle was complete by 1880 (Barsh 1990). The homestead boom starting in 1903 (Hudson 1976) meant plowing of most flat areas (Redmann 1975) and was short-lived: after the dust bowl of the early 1930s, the majority of these plowed areas converted to rangelands (Hill and Olson 2013). Topsoils (O and A horizons) of the unglaciated plains are shallow, and vegetation in the unglaciated plains of western North Dakota is predominantly perennial grasses and sedges (NRCS 2006). The vegetation is characterized as “northern short grasslands” by the World Wildlife Fund (WWF 2013). This vegetation type has been heavily modified on the global scale, and western North Dakota contains some of the most intact areas.
Underneath the rolling prairie and badlands of western North Dakota lie the Bakken and Three Forks geologic formations that contain extractable petroleum. The Bakken petroleum industry involves the building of 10–25 acre drilling sites that are then shrunk to a minimum footprint (2–10 acres) when drilling is completed and pumping commences. Interim reclamation (topographic engineering and revegetation with largely native plant species) is normally mandated within six months from when pumping begins and is conducted to manage runoff and reduce erosion in the impacted area. Livestock are excluded from these reclamations. The doughnut-shaped reclamation around the pumping location may be destroyed and reshaped numerous times during the life of a location, or may remain intact over its entire lifespan. Although temporary, failed interim reclamation may adversely affect neighboring farms and ranches through erosion and weed introductions.
We have undertaken a long-term research project to determine whether cover crops speed reclamation success (evidenced by seeded perennial grass establishment and rangeland health). In order to provide highly-applicable data to reclamation practitioners, the experiment was seeded by contractors either at the scale of an entire average reclamation or half of a large-sized reclamation. This is an early report on our research. We determine immediate (less than two years post-planting) effects of cover cropping and perennial grass seed mix composition on perennial grass establishment and rangeland health. High early establishment of reclamation seeds is important for long-term restoration success (Uselman et al. 2014). In addition, we would expect any competition between cover crops and reclamation species (i.e., limitation of perennial grasses by cover crops) to occur soon after seeding. Because other research has shown that an oat cover crop is not persistent in western North Dakota under grazed conditions (Espeland and Perkins 2013), we also test whether oats planted as a cover crop persist in ungrazed environments. Persistent cover crops may impart more benefit than non-persistent ones.
Methods
We tracked early plant establishment in revegetation within reclamations in two studies: one planted in 2014 in the glaciated plains, seeded with or without an oat cover crop and measured in 2015; the other planted in 2015 in the unglaciated plains, seeded with an oat cover crop or with a cover crop cocktail, measured in 2015. The studies were conducted in interim reclamations planted at Ft. Berthold Indian Reservation (New Town and Mandaree, ND). Both studies used a seven-species perennial grass mix of: Pascopyrum smithii (western wheatgrass), Nassella viridula (green needlegrass), Elymus trachycaulus (slender wheatgrass), Bouteloua gracilis (blue grama), Bouteloua curtipendula (sideoats grama), Schizachyrium scoparium (little bluestem), and Koeleria macrantha (prairie junegrass). Average growing season precipitation (April–August) in nearby Williston, ND is 23 cm. Growing season precipitation for 2014 was 22.1 cm and was 20.4 cm in 2015 (NOAA National Climate Data Center).
In the 2014 study, two versions of the mix (different densities of component species and different seed amounts per unit area) were planted (Table 1). The two different versions were the federal Bureau of Indian Affairs (BIA) recommended mixture and the local Mandan-Hidatsa-Arikara (MHA) recommended mixture. While the species in each mixture were the same, the seeding rates were different (Table 1). To test differences between the mixes, plus the effect and persistence of an annual oat cover crop, we planned eight reclamations with a factorial design of seed mix and cover crop (simultaneously-seeded annual Avena sativa [oat]) in 2014. Due to a planting error, the plantings were not executed in a completely factorial fashion (Table 2). A variety of planting techniques were used: hydro-seeding, drill-seeding, drill-seeding plus dragged with a chain, and broadcast. Seeding rates were doubled when the broadcast method was employed. Sites were planted in late summer/early fall of 2014. Each site and treatment combination is an experimental unit.
Seeding rates (kilograms live seed per hectare, or KLS/ha) of the perennial grass mixes used in 2014 (seeding rate doubled if broadcast). The oat cover crop was planted at 11.2 KLS/ha. Percent frequency (freq) of species occurrence measured in 2015, numbers in parentheses are one standard deviation. * Significant difference in SWG frequency between the two planting mixes (p < 0.05). MHA: Mandan-Hidatsa-Arikara (MHA) recommended mixture. BIA: Bureau of Indian Affairs (BIA) recommended mixture.
Seeding protocols of the 14 experimental units (sites): timing (planting month and year), perennial grass seed mix (Table 1), cover crop type, and planting method along with perennial grass density (plants/m2) measured in August 2015. * This site was dragged after seeding.
The 2015 plantings were designed to test the long-term effect of a cover crop cocktail versus an oat-only cover crop. Although 12 sites were planted in 2015, we only measured the four sites where planting was performed in June: sites planted later than June had insufficient growing time prior to the August measurement event. We split large sites into two treatments: one with cover crop cocktail and one with an oat cover crop (Table 2). All sites were planted with the MHA perennial grass mix, and all sites were broadcast seeded. Each site and treatment combination is an experimental unit.
The cover crop cocktail sown in 2015 was chosen for species that are drought tolerant and support pollinator and beneficial insects as well as to cover the functional groups suggested by the North Dakota NRCS to build soil health (ARS 2016). Recommended rates (kg live seed per hectare, or KLS/ha for drill seeding were all doubled in 2015 because all sites were broadcast-seeded in 2015. The cover crop cocktail was composed of ten species (functional group, recommended KLS/ha: Pennisetum glaucum (pearl millet [=Cenchurs americanus], late annual grass, 0.4256), Coreopsis tinctoria (plains coreopsis, late broadleaf, 0.28), Glycine max (soybean, not Roundup ready, late legume, 2.1), Avena sativa (oats, early annual grass, 1.82), Pisum sativum var. arvense (field pea early legume, 1.05), Phacelia tanacetifolia (phacelia, early broadleaf, 0.14), Linum usitatissimum (flax, early broadleaf, 0.14), Raphanus sativus (radish, root vegetable annual, 0.112), Brassica rapa var. rapa (turnip [=Brassica rapa], root vegetable annual, 0.112), and Helianthus annuus (sunflower, late broadleaf, 0.112).
Measurements for both studies were conducted in late August of 2015. We recorded plant species frequency and presence as well as rangeland health on transects within experimental units stratified for variation in slope and aspect. We placed one 25-m transect on a flat area within each experimental unit and one transect of the same size on a slope, or placed a second transect in the flat area if the reclamation was entirely flat. Each transect encompassed five evenly spaced plant sampling locations and five evenly spaced soil sampling locations. For each transect, soil samples to a depth of 6 cm were collected with a tulip bulb planter (11-cm tall: base and top diameter 6 and 7.5 cm, respectively) and mixed. These samples were sent to Ward Labs (Kearney, NE) for the S4 analysis package (producers. wardlab.com). At four of the flat, smaller sites (BRU, MLS, DAB, and MAS) soil samples were collected along a single, flat transect. Planted perennial grass frequency was collected following Vogel and Masters (2001). To measure plant density, the number of seedlings/m2 was counted in a 75 by 75-cm frame at each sampling location. To measure species frequency, the presence of each perennial grass species per frame was recorded. Cover crop plant species were recorded as present or absent at the level of the sampling frame as well over the entire experimental unit.
Rangeland health methods follow the Pellant et al. (2005) method of Interpreting Indicators of Rangeland Health (IIRH). Rangeland health assessments consisted of evaluating 17 indicators to assess three attributes: 1) soil and site stability; 2) hydrologic function; and 3) biotic integrity. IIRH assessments are qualitative. Assessments are made by rating study site on the degree of departure from reference conditions: 1) none to slight, 2) slight to moderate, 3) moderate, 4) moderate to extreme, and 5) extreme to total. Reference conditions are based on the ecological site concept (Herrick et al. 2006) that are part of Ecological Site Descriptions (NRCS 2011). Given the visible degree of similarity between the texture and depth of all soils at all study sites to thin loamy soils, we based our expectations for the reference condition on the thin loamy ecological site for Major Land Resource Area 53B (Ecological site ID: R053BY015ND). This Major Land Resource Area 53B covers large portions of North Dakota and South Dakota. Component indicators that make up the assessments are shown in a reference matrix created for the thin loamy ecological site (Appendix 1). To standardize qualitative assessments only one person made all of these assessments, using the reference evaluation matrix to guide all ratings.
We also collected quantitative data to further inform the IIRH. A line-point intercept (LPI) with a point spacing of 50cm along the 25m transect was used to collect vegetation cover, basal cover, and ground cover data (Herrick et al. 2005). Soil aggregate stability samples were collected from each plot and analyzed using a soil stability field kit (Herrick et al. 2001, 2005).
Data Analysis
We generated a correlation matrix for all soil variables at the transect level (n = 24) to choose the data we would analyze in a Principal Components Analysis (PCA). We report correlations where correlation coefficient absolute values exceed 0.6 (Table 3). Soil pH, P, Mn, and Ca were uncorrelated with other variables in the analysis (coefficients less than 0.6). Because of high degrees of correlation between salts and N, S, Cu, Mg, and Na (all positive) and high correlation of N, S, Cu, Mg, and Na with each other (Table 3), we included salts in the analysis but excluded its correlates. The PCA model included pH, salts, P, Fe, Mn, Ca, OM, and plants/m2 to understand how soil nutrients contribute to initial plant density. We examined component axes that explained up to 60% of the variation among transects and report loading values with absolute value greater than or equal to 0.7. We used partial least squares regression to further examine the relationship between planting success (plants/m2) and the subset of constituent soil variables at the site level. Data were analyzed in R v 3.2.5 (R Foundation, Vienna, Austria) using the “plsr” function in “pls” with leave-one-out cross-validation (Mevik and Wehrens 2007). Means are reported ± one standard deviation.
Significant Pearson correlation coefficients (p < 0.05) among soil variables with absolute values greater than 0.6. 1Soluble salts 1:1, mmho/cm. 2NH4OAc, ppm. 3DTPA, ppm.
Indicator values for the 17 indicators of rangeland health were used to come up with attribute ratings for each of the three IIRH attributes: soil and site stability, hydrologic function and biotic integrity. We used a paired comparison t-test to determine whether there were significant differences in IIRH attributes between seeding mixes. We then examined component indicators within attributes to examine specific susceptibilities to treatment effects.
Results
Effect of Perennial Grass Seed Mix Planted in 2014
Even though the MHA (9.5 KLS/ha) and BIA mixes (6.1 KLS/ha) planted in 2014 had very different seeding rates, the number of established plants/m2 in 2015 was essentially the same: 28.5 ± 6.5 plants/m2 and 27.9 ± 8.6 plants/m2 respectively. However, the frequency of slender wheatgrass in established stands differed between the mixes. The frequency of E. trachycaulus (Table 1) was higher (85.0%) in BIA-planted sites compared to MHA-planted sites (73.8%). Planting rate of E. trachycaulus in the MHA mix was twice that found in the BIA mix (Table 1). Frequencies of other species between the two plantings did not differ. All MHA-seeded sites were broadcast-seeded at double rates (Table 2), while the BIA mix was seeded in a variety of ways, including broadcast application. We expected establishment in BIA-planted sites would be more variable due to the different types of seeding technology used, but CVs for plant establishment were similar (0.21 and 0.19 for BIA and MHA plantings, respectively). There was no significant difference in attributes of rangeland health for soil and site stability and biotic integrity in either of the seed mixes.
Effect of Cover Crops Planted in 2014
The presence of oat did not alter plant density measured in 2014: sites planted with oats averaged 28.1 ± 6.1 plants/m2, and sites planted without oats averaged 28.3 ± 8.8 plants/m2. Oats persisted at some sites: OLS and MAS had low frequencies of oat (25% and 40% respectively) while AGB and COY had no oats present.
Sites with and without cover crops remained in the same category rating; sites with no oats were closer to moderate to extreme departure from what is expected for this ecological site (y-axis, Figure 1). The only significant individual indicator was invasive species (t ratio = 1.86; df = 11; p = 0.003). Sites where oats had been planted showed no to slight departure from what would be expected while sites without oats showed extreme to total departure (50–70% canopy cover of invasive species).
Departure from expected reference condition of rangeland health for sites seeded in 2014 with and without an oat cover crop. Black bars represent soil and site stability, gray represent hydrologic function, and white represent biotic integrity.
Effect of Cover Crops Planted in 2015
Perennial grass density averaged 20.6 ± 12.3 plants/m2 at sites where the oat cover crop was planted, and sites planted with the cover crop cocktail averaged 17.3 ± 11.3 plants/m2. All cover crop cocktail species established at all sites. Among samples, species frequencies were (% frequency ± one stdev); P. glaucum (87% ± 12%), C. tinctoria (60% ± 36%), G. max (20% ± 20%), A. sativa (70% ± 30%), P. sativum (17% ± 21%), P. tanacetifolia (13% ± 12%), L. usitatissimum (73% ± 21%), R. sativus (50% ± 36%), B. rapa (40% ± 35%), and H. annuus (33% ± 21%).
Rangeland health assessments showed no significant difference in attributes of soil and site stability and hydrologic function, but there was a very slight, statistically significant difference between the treatments for biotic integrity (t ratio = 1.86; df = 9; p = 0.04) with range health of the cover crop cocktail showing moderate departure from reference condition, while the oats cover crop showed moderate to extreme departure from reference condition. Component indicators did not show any statistically significant differences.
Soil Chemistry and Initial Plant Establishment
Soils were variable across sites. Some soil nutrient concentrations such as K and organic matter doubled depending on the location of collection (Table 4), while nutrients such as salts, P, Cu, and Na varied exponentially. Glaciated (AGB, BRU, COY, DAB, FBIR, MAS, MRS, and OLS) and and unglaciated (EGB, SYW, MBL, and IND) sites were interspersed for soil characteristics (Figure 2). Three of the four unglaciated sites (SYW, EGB, MBL) were positioned at three of the four extremes along the two axes of variation. Table 4 shows the measured range of each nutrient shown in Figure 2.
Low and high values of soil nutrients over all transects. Units as in Table 3.
Principal components analysis results of plant density and soil chemistry by site. A) Distribution of sites: grey circle encompasses AGB, COY, DAB, MAS, MRS, OLS and SYW. B) Plant soil vectors: as phosphorous increases, plants m–2 increase; as salts increase, plants m–2 decrease.
Transects with poor soils had the fewest number of plants (Figure 2). Component axis 1 explained 38.2% of the variation among transects and axis 2 explained 26.9% of the variation. Important loadings for axis 1 were plants/m2 (–0.82), salts (0.75), Fe (0.84), and Mn (0.75). The important loading for principal component axis 2 that further discriminated transects from one another was pH (–0.89). Salts, Fe, Mn, and pH were associated with plant density across the landscape.
When we examined the relationship between soil variables and plant density at the site level, the partial least squares regression model with the lowest prediction error consisted of three components predicting 65.1% of the variance in plant density (root mean standard error of prediction: 6.05). These three components reflected a negative relationship with calcium, iron and manganese content (Component 1: Ca [–1.00]; Component 2: Fe [–1.00] and Mn [–0.19]), and positive effects of soil fertility (Component 3: P [0.96]; Fe [0.22]; Mn [0.15]) on plant density. At the level of reclamation site, Ca, P, Fe, and Mn predicted plant density.
Discussion
We found that soil chemistry appeared to drive initial perennial grass establishment and found no evidence that cover crops competed with (or prevented the establishment of) perennial grasses. A calcareous salt layer is common in Western North Dakota soils and is concentrated at depth (below 6 cm) through chemical and physical weathering such that its effect on plants is minimized (e.g., Espeland and Perkins 2013). As part of site construction, topsoil to a depth of about 15 cm is set aside (J. Morrell, WPX Energy, personal communication), and then it is replaced during reclamation. Our soil samples were restricted to this replaced topsoil layer. Soil nutrients reported in this paper do not necessarily reflect their bioavailability. When the buried calcareous salt layer is mixed in the topsoil, salts are in substantial contact with plant roots. During excavation, this salt layer is apparent and, while laborious to remove during the site construction phase, efforts to do so may greatly improve plant establishment in reclamation. After energy development, high sulfate and salt concentrations in surface soils have been found by other researchers in the northern Great Plains (Rowell and Florence 1993, Berquist et al. 2007). Fertilization, increasing organic matter, and/ or mycorrhizal inoculant would not help plants overcome this limitation (e.g., Espeland and Perkins 2013). We found that plant establishment is also limited by P and other nutrients, as has been found by other researchers in the northern Great Plains (Reinhart et al. 2016.)
There was no difference in establishment of the two perennial grass seed mixes. At this time, there appears to be little benefit of seeding the costlier MHA mix because of a slight gain in hydrologic function. Our understanding of the mechanism that underlies this similarity is restricted by low sample sizes (n = 2 for the seed mix by cover crop combination) and wide variation in soil conditions and seed application technologies. Estimates for the desirable number of plants per m2 in restoration, or, the density above which perennial grasses self thin, vary widely: from 16 plants/m2 (Kulpa et al. 2012) to 301 plants/m2 (Burton et al. 2006). It is possible that both mixes are sufficient to achieve the self-thinning density in this system and that self-thinning began before the measurement year.
Species occurrence (frequency) was similar even when seeding rates were more than doubled: for example, B. gracilis occurred at a 57% frequency in both mixes even though the seeding rates for this species were 1.1 KLS/ha in MHA and 0.2 KLS/ha in BIA. The significant difference in E. trachycaulus was opposite of expected: less occurrence with the higher seeding rate. This could be due to density-dependence or to lower E. trachycaulus. performance in broadcast compared to drill seeding, however neither has been shown to be important to E. trachycaulus establishment in multiple plantings and trials (J. Scianna, NRCS Bridger Plant Materials Center, personal communication). The general phenomenon we observed of species insensitivity to seeding rate may be due to the primary importance of soil chemistry to plant establishment.
When they are persistent, it may take several years for annual cover crops to have an impact (Espeland and Richardson 2015). Avena sativa persisted into the next year at half the sites where they were planted: in both cases when they were planted in summer. During a brief site visit in September 2016, only C. tinctoria and H. annuus cover crop species were observed at reclamations sown in 2015 (John Hendrickson, personal observation). Because soil conditions were so variable and because, in most cases, we planted a single treatment per site, environmental variation may continue to determine site recovery rate more than cover crop treatments. In addition, the cover crop cocktail established at a much lower density than in agricultural applications (Figure 3). Low cover crop densities that could have been caused by soil conditions, low rainfall, and/or inadequate seeding rates (all untested in this study) add to the limited likelihood of observing a strong effect in restoration. Low cover crop densities could have also limited their capacity to compete with seeded perennial grasses.
A) Cover crop cocktail density shown to have a beneficial effect in agricultural research, western North Dakota, USA. Photo: Brett Allen. B) Cover crop density in August 2015 reclamation. Photo: the authors. See Table 1 for species composition of the cover crop cocktail.
The oat cover crop appeared to reduce invasive species in the 2014 plantings, leading to increased biotic integrity. Parallel to this result was a slight increase in biotic integrity in sites planted with cover crop cocktails in 2015, although the component indicator of invasive species was not significant in this case. We should be able to better understand the relative effects of time since planting and oat versus cocktail cover cropping on biotic integrity and specifically invasive species cover by continuing to monitor these reclamations and by planting more sites in subsequent years. It should be noted that IIRH defines invasive species as those that can become dominant in either the short- or long-term without intervention; our most common “invasive species” were Bassia scoparia (kochia [= Kochia scoparia]), Chenopodium album (lambsquarters), and Bromus tectorum (cheatgrass).
Given the dramatic benefits of cover crops in agriculture (Reicosky and Forcella 1998, Snapp et al. 2005, Delgado et al. 2007), we expected to see a stronger effect of cover crops on perennial grass establishment and rangeland health in interim reclamations in the short term. We will continue to monitor these sites to determine whether benefits develop over time, and we will add new plantings to increase our statistical power. Our data show that substantial variation in soil chemistry, even within the same ecological site type, can limit the establishment of desirable seeded species. Among-site variation in the northern Great Plains has been shown to have greater consequences for perennial grass establishment and growth compared to reclamation treatment techniques (Rinella et al. 2012, Rinella et al. 2016). In addition, reclamation after significant soil disturbance may pose greater challenges to the establishment of desirable vegetation than in other types of reclamation (e.g., Robichaud et al. 2000, Berquist et al. 2007).
Koeleria macrantha. USDA-NRCS PLANTS Database. Hitchcock A.S. 1950. Manual of the Grassess of the United States. Washington, DC: USDA Miscellaneous Publication No. 200.
Acknowledgements
This project would not have taken place without the support of Jim Raley and WPX Energy. Patient field assistants were Krystal Leidholm, Andrew Carrlson and Trevor Besosa. Maureen O’Mara prepared the soil samples for analysis. Brenda Schladweiler (BKS Environmental) generated the concept of cover crop cocktails for reclamation. We thank the three affiliated tribes (Mandan, Hidatsa, and Arikara) for accommodating the project. The manuscript was improved by comments from John Gaskin, Myla Aronson, and two anonymous reviewers. Funding supplied by USDA appropriated project #5436-22000-017-00. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the US Department of Agriculture.
