Does a C3-C4 forage mix simultaneously improve forage production and carbon sequestration?

Final Report for GNC07-077

Project Type: Graduate Student
Funds awarded in 2007: $9,911.00
Projected End Date: 12/31/2009
Grant Recipient: University of Wisconsin-Madison
Region: North Central
State: Wisconsin
Graduate Student:
Faculty Advisor:
Randall Jackson
University of Wisconsin-Madison
Expand All

Project Information


The tallgrass prairie ecosystem is one of the most endangered ecosystems in North America. Interest in restoring prairie is growing because of the potential for increasing soil organic carbon (SOC) sequestration when degraded soils and ecosystems are restored. Also, finding long-term C storage in terrestrial ecosystems is being promoted as a key part to climate stabilization as greenhouse gases (GHG) continue to accumulate in the atmosphere. Re-introducing warm-season grasses (C4) to cool-season (C3) pastures has the potential not only to increase forage production, but also improve wildlife habitat, soil organic matter, and resilience to drought. While much is known about C4 prairie grasses and C3 pastures, there is a paucity of information about mixed C4-C3 grasslands in the upper Midwest.

We conducted an experiment in two restored prairies in southern Wisconsin to assess their carbon sequestration potential under a gradient of C4-C3 grass ratios. We identified areas of naturally occurring C4-C3 gradients in two restorations (10- and 17-year-old) that have been recolonized by C3 grasses and quantified above- and below-ground net primary productivity (ANPP and BNPP) and soil respiration (Rs) for two years. Across both sites and years, we found a weak but positive relationship between net ecosystem production (NEP) and C4 grass cover. However, this relationship was driven by the abundance of a particular species, Andropogon gerardii. Also, the relationships between the major components of the C cycle (ANPP, BNPP, and Rs) and C4 grass cover and A. gerardii cover were modified primarily by site, indicating that aboveground dynamics were driving C balance at the high-productivity site and belowground processes controlled the site with lower productivity. These results have important implications for land managers and policymakers seeking to promote C sequestration.


The tallgrass prairie, which is dominated by warm-season (C4 photosynthesis) grasses such as Andropogon gerardii Vitman (big bluestem), Panicum virgatum L. (switchgrass), and Sorghastrum nutans (L.) Nash (indiangrass) has been transformed mainly into grassland/agricultural mosaics if not totally converted to annual crops (Rhemtulla et al., 2007). Interest in restoring the native prairie has increased at the same time that the potential for increasing soil organic carbon (SOC) sequestration when degraded soils and ecosystems are restored have been touted (Lal, 2003). Finding long-term C storage in terrestrial ecosystems is being promoted as a key part to climate stabilization as greenhouse gases (GHG) continue to accumulate in the atmosphere (Bruce et al., 1999; Smith, 2004). Much emphasis has been placed on the adoption of best management practices and restoration to perennial vegetation in agricultural systems (Smith et al., 2008; Johnson et al., 2005) because soil carbon increases the most after a carbon enhancing land-management change is adopted (Smith, 2004; McLauchlan et al., 2006; Matamala et al., 2008). Indeed, studies have reported that the restoration of tallgrass prairie has the potential to accumulate soil organic carbon (SOC) on the order of 40 to 60 g C m-2 yr-1 (Baer et al., 1992; McLauchlan et al., 2006; Mahaney et al., 2008). A successful example of unification of food production and conservation is the conservation reserve program (CRP), whose primary goal is to protect erodible land with the establishment of perennial grasslands, but it also has been adopted because of the reported potential to increase SOC levels decreasing atmospheric CO2 (Gebhart et al., 1994; Reeder et al., 1998; Follett et al., 2001; Baer et al., 2002; Post et al., 2004).

Working grasslands, such as sown pastures, tend to be dominated by European species that are highly productive in the spring and fall since they are mostly cool-season (C3 photosynthesis) grasses (Paine et al., 1999). These pastures sustain the profitable dairy (Taylor and Foltz, 2006) and beef industry (CIAS, 2008) in Wisconsin.

Combining the challenges of food production and the need for environmental conservation on such working lands have the potential to positively impact a large part of the landscape since in the U.S. alone over 50% of US is cropped or grazed (Robertson and Swinton, 2005). The reintroduction of C4 prairie grasses into working lands offers a compromise between the complete restoration and the complete eradication of the native prairie (Woodis, 2008; Nelson and Burns, 2006) and simultaneously promotes the multifunctional use of the rural landscape (Buttel, 2003; Western, 2001; Hilderbrand et al., 2005).

While much is known about C4 prairies and C3 dominated pastures ecosystems, there is a lack of information about mixed C3-C4 grasslands in the upper Midwest. Adding functional diversity to the C3 dominated pastures have the potential to increase ecosystem properties through positive interaction among functional groups by complementarity and facilitation (Spehn et al., 2000; Hooper et al., 2005). For instance, using CENTURY model to evaluate how plant communities (100% C3, 100% C4, and 50% C3 50% C4 mix of grasses) could affect NPP revealed that plant production was lowest for C3 grasses and highest for the mixed C3-C4 community (Seastedt et al., 1994). However, most of the recent studies evaluating the effects of photosynthetic pathway on ecosystem services focus on the effects of totally converting agricultural lands to prairie or to CRP (Camill et al., 2004; McLauchlan et al., 2006). Only a few studies have recognized the need to understand how ecosystems are affected by the reintroduction of C4 native grasses into C3 grassland (sensu Hooper et al., 2005); but even then, the comparisons are made between C4 dominated and C3 dominated communities (Mahaney et al., 2008), not mixed C3-C4 grasslands per se. Little is known about how the C3 dominated grasslands may change as C4 prairie grasses are reintroduced to working lands—this information will assist land managers in decision-making on issues such as: How much prairie grass is needed to boost productivity? Is the ecosystem storing carbon? How does the seasonality of production change with various rations of C3 and C4 grasses?

To address the above mentioned questions, we compared how net ecosystem production (NEP) and the major components of this metric, net primary production (NPP) and soil respiration (Rs), were influenced by the relative abundance of C3 and C4 grasses in restored tallgrass prairie of southern Wisconsin over two years. We expected that NEP would increase as C4 grass abundance increased because C3 and C4 grasses differ in important functional traits such as quantity and quality of below- and above-ground biomass that can directly and indirectly alter soil processes (Dijkstar et al., 2006).

Project Objectives:

The project seeks to improve understanding of ecosystem support, provisioning, and regulating services provided by pasture ecosystems in the Upper Midwest. This work intend to provide much needed understanding about C3 pastures and the benefits of re-storing native grasses to working lands.

Products include peer-reviewed publications, an Agroecology MS thesis, and presentations at field-days, conferences, and to interested conservation-research groups.


Click linked name(s) to expand/collapse or show everyone's info
  • Randall D. Jackson


Materials and methods:

This study was conducted at two Wisconsin prairie restoration sites, Bison Ridge Ranch (BR Ranch) in Marquette County (89° 27' W, 43° 44' N) and the Wisconsin Integrated Cropping Systems Trial (WICST) at the University of Wisconsin–Madison’s Arlington Agricultural Research Station in Columbia County (89° 19' W, 43° 18' N).

The soils of BR Ranch were classified as Gotham loamy fine sand and Metea fine sandy loam on 2 to 6 % slope (NRCS). The grass community is comprised of Elymus repens (L.) Gould (quackgrass), Andropogon gerardii Vitman (big bluestem), Bouteloua curtipendula (Michx.) Torr. (side oats grama), Bromus tectorum L. (downy brome), Digitaria cognata (Schult.) Pilg. (fall witchgrass), Panicum virgatum L.(switchgrass), Poa pratensis L. (Kentucky bluegrass), Schizachyrium scoparium (Michx.) Nash (little bluestem) and Sorghastrum nutans (L.) Nash (indiangrass).

Soils at WICST were classified as Plano silt loam on 0 to 2% slopes (NRCS). Today the grass community is mainly comprised of Elymus repens (L.) Gould (quackgrass), Andropogon gerardii Vitman (big bluestem), Bromus inermis Leyss. (smooth brome), Dactylis glomerata L.(orchardgrass), Elymus canadensis L. (Canada wildrye), Phleum pratens L. (timothy), Poa pratensis L. (Kentucky bluegrass), and Sorghastrum nutans (L.) Nash. (indiangrass).

At each site, we chose thirty ~100-m2 plots (15 in 2007 and 15 in 2008) for their respective C3:C4 grass ratio. At each plot, a centralized 1-m² quadrat was permanently marked to monitor soils, environmental conditions, nutrient content, and respiration over time. Plant species cover was estimated with the line-point method performed on permanently marked soil monitoring stations. The quadrat area to be sampled was divided by 5 horizontal and vertical lines forming 25 intersections where the first intercept of a sharpened rod with any part of herbaceous vegetation at each intersection was recorded (Heady, 1959).

For each sampling event, total species cover was calculated as total species hits divided by the total possible hits for each quadrat. Plants were identified to the species level and grouped into functional groups. We mirrored the classification method used in previous prairies studies (Kindscher and Wells, 1995; Tilman et al., 1997).

Plant cover data per species and functional groups were taken three times during the season and aggregated to calculate an annual average of cover for each experimental unit.

The two main dominant C4 grass species were the same for both sites, e.g. Andropogon gerardii Vitman and Sorghastrum nutans (L.) Nash; however, there were differences in the plant communities worth noting. For instance, BR Ranch had other species of C4 grasses present such as Panicum virgatum L., Schizachyrium scoparium (Michx.) Nash and Bouteloua curtipendula (Michx.) Torr. The dominant C3 grass species in both sites were also quite similar. Yet, only BR Ranch had the winter annual Bromus tectorum L. and only WICST had Elymus canadensis L. and Phleum pratense L. present.

Net ecosystem production
Annually, SOC storage or loss is the net balance between inputs and outputs of C (Six et al., 2002). Carbon sequestration is calculated as the difference between C inputs as NPP and C outputs as heterotrophic respiration (Rh), harvest, and fire. Net primary production is the sum of aboveground net primary production (ANPP) and belowground net primary production (BNPP). Soil respiration (Rs) is a combination of autotrophic (i.e., root respiration, Ra) and heterotrophic respiration (i.e., microbial respiration, Rh). Because separating these forms of respiration is quite difficult (Kuzyakov and Larionova, 2006), we used literature estimates of Ra and calculated Rh from Rs measurements. We estimated potential carbon sequestration in grams of carbon per unit area by measuring the major fluxes of C such as ANPP, BNPP, and Rs of each experimental unit (Chapin et al., 2002). Because our intention was to compare the potential in C sequestration of different areas along the C3:C4 gradient for each year, we did not estimate C loss from burning, which occurred intermittently (i.e., every 3 years or so).

Net primary production
Within each delineated 100-m2 experimental unit, we clipped available biomass within a single 50 × 50-cm randomly placed quadrat to ~3 cm residual stubble height. Biomass was bagged, dried to constant weight at 60°C, and weighed. Prior to each clipping event, we estimated leaf area index (LAI) on the 50 × 50-cm quadrat, which was calculated from intercepted photosynthetically active radiation readings (Accupar LP-80, Decagon, Inc., Pullman, WA) made above and below the leaf canopy at four points along each quadrat. Because the focus of our study was to capture differences along the C3:C4 grass gradient, we estimated ANPP not by simply harvesting the biomass at peak standing biomass, as is normally done to estimate ANPP in prairie, but by summing the biomass increments between clipping events throughout the season to account for the simultaneous and overlapping production and mortality of both plant functional groups (Vogt et al., 1986). In each experimental unit, annual ANPP was calculated as the sum of monthly aboveground production values for a growing season (approximately May through October). The aboveground biomass monthly production was calculated as the difference in biomass measured at one month from biomass at an earlier month. Since we did not measure the biomass production after the harvest in August 2007 at the BR Ranch, we estimated the fall biomass with an allometric equation. The allometric equation (r² =0.69) was developed from the relationship between LAI and biomass weight measurements taken from both sites and both years. In the same way, we estimated the amount of biomass left on the pasture after harvesting events.

Belowground NPP was estimated using root ingrowth cores in each experimental unit (4 subsamples per experimental unit in 2007 and six subsamples per experimental unit in 2008). The root biomass of each core was bagged, refrigerated until washing, dried to constant weight at 60°C, and weighed. We did not attempt to separate live to dead roots assuming that only a very small fraction was dead (Fahey and Hughes, 1994). For each experimental unit, we averaged the two root biomass weights to estimate root biomass per unit area for each harvest. We calculated the belowground biomass production as the difference in root biomass harvested at one time from the root biomass of an earlier harvest. The root net production was calculated as the sum of positive production values. Since the root ingrowth cores do not account for the amount of roots that are produced and die during the period when the cores are deployed (i.e., fine root turnover (FRT)), we corrected the net root production using the FRT rates using equation below from the literature (Gill and Jackson, 2000; Fahey et al.,1999). FRT = 0.2884 e (0.046 * (Mean Annual Temperature))

We also estimated the belowground root stock of C4 and C3 grasses at the end of April, August, and November of 2008 using soil coring (Fahey et al., 1999). Soil cores were 5 cm in diameter and taken to a 60-cm depth immediately above six C4, and six C3 individual grasses totaling 12 randomly chosen plants in each site. The 60-cm soil cores were divided into four 15-cm sections, sieved in 2-mm mesh, washed out of debris, rocks and soil, dried to constant weight at 60°C, and weighed. No attempt was made to separate living from dead roots. For each functional group of grasses, the six root weight values per segment were averaged to access the vertical distribution of root stock biomass. We used the percent average of biomass found in the first 15 cm of the soil profile to later extrapolate BNPP to 60-cm depth.

Above- and below-ground production was calculated by multiplying shoot and root biomass estimates by a constant C concentration of the prairie plant tissues as 50% (studies in restored and remnant prairies found 40 to 50% C in shoot and root tissues, i.e. Brye et al., 2002 and Matamala et al., 2008).

Soil Respiration
Soil respiration was measured once monthly between 0900 to 1500 hours with a soil respiration chamber linked to an infrared gas analyzer (LiCor 6400-09, Li-Cor Biosciences, Lincoln, NE) (Norman et al., 1992). The efflux chamber was used in conjunction with polyvinyl chloride (PVC) thin-walled collars that were inserted 2 cm into the soil surface at least 30 min prior to conducting measurements. There were three collars per 1-m² quadrat, which were repositioned in a random manner within each experimental unit every month. We estimated Rs, the soil CO2 efflux (µmol CO2 m-² s-¹), in each experimental unit by averaging the three measurements taken in each of the three collars. Because shoots were clipped prior to measuring Rs, the efflux of soil CO2 measured excludes shoot respiration. Annual Rs, the mass of C per unit area respired annually was estimated for each experimental unit by summation of daily interpolations of the monthly measurements (Brye et al., 2002; Chou et al., 2008).

In the fall of 2007 and spring of 2009, we used the hammer-type core sampler when the soil was not too wet or dry to minimize compaction and extracted one soil core (5 × 15-cm) per experimental unit. Soils were sent to The University of Wisconsin Soil & Plant Analysis Lab to determine soil total carbon (TC) and total nitrogen (TN), potassium (K), phosphorus (P), and organic matter (OM). Percent of TC and TN were determined by dry combustion using a Leco CNS-2000 analyzer (Organic Carbon Dry Combustion method, Leco CN-2000, FP 2000, or CNS-2000). Plant available K and P were estimated using the Bray P method. Soil organic matter was estimated by the loss of weight in a sample heated at a temperature high enough to burn organic matter but not so high as to decompose carbonates (Weight Loss-on-Ignition - LOI 360o). Soil pH was measured in water using a 1:1 soil: solution ratio and in a buffer solution with a 1:1:1 of soil: water: buffer ratio.

Research results and discussion:

We used linear regression to address potential relationships between net ecosystem production (NEP) and vegetation cover by functional group, by plant species cover, and then by plant species richness. Cook’s distance test, which measures the influence of individual observations on the regression coefficients, was assessed to identify potential outliers. Furthermore, the fit of a quadratic function was compared to a linear fit using the generalized least squares algorithm in S-plus (S-Plus 8.0, Insightful Corporation Seattle, WA). Models were compared with likelihood ratio tests. When significant differences were determined (p < 0.05) the model with the lowest AIC values were chosen, otherwise the simpler model was determined to be the better fit.

Net ecosystem production and C4 grass cover (%) were better described by a quadratic than a linear function (p < 0.01). Net ecosystem production was not related to C3 grass cover, species richness, or cover of any of the individual species with the exception of Andropogon gerardii. A quadratic fit best represented the relationship between NEP and Andropogon gerardii cover (p = 0.0001). Total NPP was not significantly correlated to C4 grass cover. While ANPP increased significantly with increasing C4 grass cover, no relationship was found between C4 grass cover and BNPP. Warm-season grass cover explained 17% of the annual variability in soil CO2 efflux.

While we did not find a correlation between BNPP and plant cover, we did find significant differences in root stock biomass found immediately below C4 and C3 grasses at both sites.

When NEP data were analyzed separately by sites using the regression tree, we found that at WICST, the best predictor of NEP was ANPP—calculated as peak standing biomass—and that at BR Ranch, the best predictor of NEP was BNPP. At WICST, 73% of the deviance was reduced by the split of the database by ANPP estimated as peak standing biomass and at BR Ranch, 76% of the deviance was reduced when the database was split by BNPP (trees not shown). Clearly, the aboveground parameters were most important at the more productive site which was reflected in NEP, ANPP, and BNPP estimates. Net ecosystem production increased as C4 grass abundance increased, but the results indicated that species composition was more important than functional groupings for C dynamics. The relationships between the estimated major components of the C cycle (ANPP, BNPP, and Rs) to functional group and plant species cover followed the predicted directions but were modified primarily by site productivity.

Participation Summary

Educational & Outreach Activities

Participation Summary:

Education/outreach description:

The first year's field data collection successfully ended in October 2007. Since then some of the results were presented at two winter field days, one in Lancaster Research Station and another at the WICST annual meeting, both well attended by farmers, researchers and professionals.

At the Lancaster meeting two other SARE granted students and I talked about the "Tradeoffs in ecosystem services provided by warm-season grasses in pastures" while at the WICST meeting I talked about "How do ecosystem support services vary along a C3:C4 grass gradient?" In addition, abstracts were sent to two coming conferences.

The second year's field data collection successfully ended in November 2008. During the year, we presented two posters at national meetings divulgating the work and results of the project. One of the posters (ECOSYSTEM SERVICES AS A FRAMEWORK FOR AGROECOLOGICAL RESEARCH) was made as a collaborative work amongst all members of the Grassland Ecology Jackson Lab and presented at the 93rd Annual Meeting of the Ecological Society of America (ESA) in August 3-8, 2008, Milwaukee, Wisconsin. The poster was written and created by our team members: Lawrence G. Oates, Emma L. Bouressa, David S. Duncan, Ellen E. Hamingson, Julie E. Woodis (Doll), Susan K. Chamberlain, Andrew R. Jakubowski, Randall D. Jackson and I.

The other poster (TRADE-OFFS IN PASTURE PRODUCTION WITH INCREASING C4 GRASS ABUNDANCE) was written by Julie Doll and I and presented first at the Wisconsin Ecology Group at the Fall Symposium in Madison, Wisconsin and later at the Farming with Grass meeting sponsored by The Soil and Water Conservation Society (SWCS) in October 20-22, 2008, Oklahoma city, Oklahoma. The conference intended to address the factors driving change in mixed agricultural systems and was well attended by farmers, researchers and professionals.

In addition, the Agroecology MS thesis was defended and approved by the University of Wisconsin in the summer of 2009.

Project Outcomes

Project outcomes:

There are several conclusions in this project that can guide land managers and policy makers regarding mixed C4-C3 grasslands.

At the management level, the July C4 grass cover can guide the land managers in assessing the productivity of the grassland before the end of the season when the grasses are very tall and the heat strong. It seems also that low-diversity grasslands that are mainly dominated by productive warm-season grasses, such as the CRP lands, may prove to be a great alternative to improve SOC. At the same token, a low cover of cool-season grasses in restored prairies may not threaten the potential of the restoration to rebuilt SOC if the C4 grasses still the dominating grass.

Our results also show that the higher C sequestration in consequence of a higher cover of warm-season grasses is not general across sites and nutrient-poor permanent grasslands may need to be moderately intensified either by increasing organic carbon input or moderate fertilization (Soussana et al., 2004). In a positive note, many forage species such as Panicum virgatum L. have been studied as good sources of bioenergy, and according to our study, more attention could also be given to the potential of Andropogon gerardii. This project shows that research at the farm level, across various soil types and management still necessary if the C sequestered by mixed C4-C3 grasslands is to be considered as an effective strategy to long-term C storage and a key part to climate stabilization.


Areas needing additional study

The relationships between C parameters (NEP, NPP, and Rs) and C4 cover were modified by site, where C dynamics were driven by BNPP at the low-resource site and ANPP at the high-resource site, but more replication along a gradient of resources availability is needed to determine whether this effect is general.

Despite differences in root stock biomass found under C4 and C3 grasses, we did not find a correlation between BNPP and plant cover. This result runs counter to our prediction and it is inconsistent with other studies that report positive changes in BNPP when community composition shifts to C4 dominance in restored prairies (Baer et al., 2002; Camill et al., 2004; Mahaney et al., 2008) and CRP lands (Kucharik et al., 2001). This result may not be surprising considering that aboveground functional cover may not translate directly into belowground functional cover. One possible explanation is that spatial variability under bunch and sod forming grasses is high enough that estimates of aboveground cover do not match well belowground composition. Another point to consider is that most of the studies mentioned above compared cool-season pastures or agricultural lands that were restored into grasslands highly dominated by C4 grasses (60-80%). An efficient and productive technique to separate roots by functional groups is necessary to access the belowground responses along a C3:C4 grass gradient.

Because BNPP provides most of the carbon to the soil in grassland systems (Jobbagy and Jackson, 2000; Gill et al., 2002), we need a more precise estimate of BNPP along the C3:C4 gradient to evaluate the potential carbon sequestration in either working lands that intend to reintroduce warm-season grasses to pastures or in restored prairies recolonized by introduced C3 grasses.

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture or SARE.