Index | Search | Home | Table of Contents

Biomass Yield Stability of Switchgrass Cultivars

Roger G. Fuentes and Charles M. Taliaferro

INTRODUCTION

The development of viable bio-based energy systems offers many potential benefits relative to energy availability, national security, a cleaner environment, and associated economic rewards (DOE 2000). Large-scale bioenergy use will require the deployment of environmentally acceptable energy crops and cropping systems for producing large quantities of low-cost, high-quality biomass feedstocks (DOE 1999). Switchgrass (Panicum virgatum L., Poaceae), an indigenous perennial herbaceous species distributed over much of the contiguous US, was chosen by the Department of Energy (DOE) as the model herbaceous species for development as a bioenergy feedstock crop. It was chosen on the basis of its wide adaptation, high production potential on marginal soils, and tolerance to biotic and abiotic stress agents (McLaughlin 1993). Based on morphology and habitat preference, switchgrass has been classified into upland and lowland ecotypes (Porter 1966). Lowland ecotypes are adapted to flood plains and are generally taller, larger in tiller diameter, and more robust than their upland counterparts (Anon. 1954; Porter 1966). The much higher dry matter (DM) yield capability of the robust lowland cultivars compared to the smaller, less robust, upland ecotype cultivars in the southern US is well documented (Anon. 1954; Porter 1966; Sladden et al. 1991). What is less well documented is the capability of cultivars for sustained high DM production, particularly lowland cultivars grown on non-alluvial soils or marginal soils, or both. In addition, very little information is available on cultivar by environment (CE) interactions for such studies, which are of major importance in selecting and developing improved switchgrass cultivars. CE interaction causes difficulty in identifying those cultivars that perform best over the range of environmental conditions to which they will be exposed (Eberhart and Russell 1966). Therefore, this study was initiated to evaluate long-term yield performance of selected commercial upland and lowland switchgrass cultivars and cultivar blends and to estimate and characterize the magnitude of CE interaction.

MATERIALS AND METHODS

The switchgrass cultivars were ‘Alamo’, ‘Kanlow’, ‘PMT 279’, (lowland ecotypes) and ‘Blackwell’, ‘Caddo’, ‘Cave-in-Rock’, ‘Late Synthetic High Yield’, ‘Shelter’, and ‘Summer’ (upland ecotypes). The cultivar blends (equal amounts of pure live seed) were ‘Alamo’ + ‘Summer’, ‘Alamo’ + ‘Kanlow’, and ‘Kanlow’ + ‘Blackwell’. Cultivars and blends will be referred to simply as cultivars. In 1993, seeded (10 kg PLS ha^-1) sward plots (3 × 6 m) were established on research stations near Chickasha (McLain silt loam soil) and Haskell (Taloka silt loam soil), Oklahoma. Table 1 summarizes site descriptors for each of the two locations. The experimental design at each location was a randomized complete block with three replications. Plots were fertilized each spring with 78 and 90 kg N ha^-1 at Chickasha and Haskell, respectively. Beginning in 1994, plots were harvested one time annually, near the end of the growing season. A 6 m² area (1 × 6 m) from each plot was harvested using a mechanical plot harvester. Total biomass fresh weight per plot was recorded and biomass moisture content for each plot was determined to obtain total biomass dry matter (DM) per plot, which was converted to tonnes (t) DM ha^-1.

Table 1. Site description information for Chickasha and Haskell, Oklahoma.

Analyses of variance were conducted on data arranged as split-plot in time as described by Steel and Torrie (1980). Means were separated using Fisher’s protected least significant difference procedure. The yield stability of cultivars across the 14 environments (7 yrs × 2 locations) was assessed by: (1) analysis of variance to obtain the effects of cultivars, environments (14), and the CE interaction, (2) partitioning of the environmental sum of squares into linear regression and residual and the CE interaction sum of squares into heterogeneity of regressions and residual according to Freeman and Perkins (1971), and (3) estimating five genotypic stability parameters each cultivar. The parameters were:

(1) Wricke’s ecovalence (1962);

(2) Shukla’s stability variances and (1972);

(3) Finlay and Wilkinson’s regression coefficient (1963);

(4) Eberhart and Russell’s deviation from regression parameter (1966).

Wricke’s evaluates stability based on the contribution of each cultivar to the total CE interaction sum of squares. Shukla’s and use the variance of a genotype across environments as its measure of stability. Finlay and Wilkinson’s considers a cultivar stable if its response to environments is parallel to the mean response of all cultivars in the trial. Eberhart and Russell’s considers a cultivar stable if the residual mean square from Finlay and Wilkinson’s regression model is not significant (Lin et al. 1986). Pearson’s correlation coefficients and Spearman’s rank correlation coefficients were determined between all pair combinations included for the five stability parameters and the mean biomass yield. For the rank analyses, mean DM yield rankings were ordered in a descending manner (rank 1 to highest yield) and stability parameters were ordered in an ascending manner (rank 1 to lowest values for each of the parameters). The level of significance of the respective correlations was tested using Student’s t-test. All analyses were conducted using SAS version 8.1.

RESULTS AND DISCUSSION

Satisfactory plot stands were maintained for all cultivars except ‘Summer’ and ‘Shelter’ at Chickasha. Therefore, ‘Summer’ and ‘Shelter’ were excluded from the analyses of Chickasha data and combined data for the two locations. Results from ANOVA for Chickasha and Haskell and for the combined data are given in Tables 2 and 3, respectively. Cultivars, locations, years and their 1st and 2nd order interactions generally represented significant (P<0.05) sources of variation.

Table 2. Analysis of variance for each location, including 10 cultivars at Chickasha and 12 at Haskell, Oklahoma.

^zSignificant at the 0.01 probability level.

Table 3. Analysis of variance across years and locations 10 cultivars common to Chickasha and Haskell, Oklahoma.

^zSignificant at the 0.01 probability level.

Cultivar DM yield differences were significant (P<0.05) for all environments except at Chickasha in 1996 and 1999 (Tables 4, 5, and 6). When averaged over cultivars, DM yields ranged from 6.7 (1998) to 18.6 (1995) t ha^-1 at Chickasha and 9.7 (1997) to 19.0 (1994) t ha^-1 at Haskell. The two locations differed in DM yield all years except 1995. The overall mean DM yield at Haskell (14.6 t ha^-1) was higher than at Chickasha (11.4 t ha^-1), likely reflecting the higher mean annual precipitation received at Haskell. Mean yield variations over years were closely associated with amount and distribution of precipitation during the growing season. Both locations had near or above normal precipitation for most of the 7 yrs. Coefficients from simple regression of cultivar DM yields on annual precipitation (data not shown) were significant for all cultivars. R-square values ranged from 0.68 for the ‘Blackwell’ + ‘Kanlow’ blend to 0.92 for ‘Cave-In-Rock’.

Table 4. Biomass dry matter yield (Mg ha^-1) for cultivars grown at Chickasha, Oklahoma.

Table 5. Biomass dry matter yield (Mg ha^-1) for cultivars grown at Haskell, Oklahoma.

Table 6. Biomass dry matter yield (t ha^-1) for 10 cultivars grown at Chickasha and Haskell, Oklahoma.

‘Alamo’, ‘Kanlow’, and the blends that they were in produced the highest DM yields at both locations. PMT-279 had the lowest DM yields among lowland ecotypes. ‘Shelter’ and ‘Summer’ were the lowest yielding cultivars at Haskell. The DM yields of ‘Cave-In-Rock’, ‘Caddo’, ‘Late Synthetic High Yield’, and ‘Blackwell’ were of similar magnitude at both locations. Blending of cultivars did not result in definitive performance enhancement relative to the best cultivars grown in monoculture.

The mean DM yield of lowland cultivars was higher than the mean of upland cultivars every year at both locations (Figs. 1 and 2).

Fig. 1. Yearly comparisons of biomass yields from lowlands and upland switchgrasses at Haskell, Oklahoma.

Fig. 2. Yearly comparisons of biomass yields from lowland and upland switchgrasses at Chickasha, Oklahoma.

Cultivars, environments, and their interaction represented significant sources of variation (Table 7). Partitioning of the environment sum of squares revealed that the linear regression of DM yield on the environmental index was significant and accounted for most of the environment variation (Table 7). The residual (deviation from regression) was not significant. Partitioning of the cultivar by environment interaction sum of squares revealed that the variability due to heterogeneity among the slopes of the different regression lines was a significant source of variability and revealed differences in the slopes of the regression lines among cultivars. Most of the variability from the CE interaction was accounted for by the residuals (Table 7).

Table 7. Analysis of variance of 10 cultivars over 14 environments, including partitioning of the environment and of the entry × environment interaction sum of squares.

Values for each of the five stability parameters for each cultivar are summarized in Table 8. Wricke’s ecovalence ( ) values ranged from 12.05 for ‘Caddo’ to 49.23 for ‘Kanlow’. Five of the ten cultivars had significant values when tested using the procedure described by Kang and Miller (1986). Shukla’s stability variance ( ) values ranged from 0.82 for ‘Caddo’ to 4.39 for ‘Kanlow’. Five of the ten cultivars had values for significantly different from zero. The significant and values are considered as indicators of low stability for DM yield. None of Shukla’s values, ranging from 0.25 for ‘Caddo’ to 1.53 for ‘Kanlow’ and ‘Cave-In-Rock’, were significant. Values for are obtained after the effect of the covariate has been removed from the CE interaction sum of squares as heterogeneity of regression and they are part of the residual variance of the CE interaction. The discrepancy between and as indicators of cultivar stability is due to the linear effect of the covariate. Use of covariate analysis was effective in removing this effect. Based on , all of the ten switchgrass cultivars evaluated for stability had stable biomass production across the range of environmental conditions tested. Analysis of stability using Finlay and Wilkinson’s () regression coefficient revealed that only one cultivar, the blend ‘Alamo’ and ‘Kanlow’, had a regression coefficient significantly higher than 1.0 ( = 1.32). The rest of the cultivars had values ranging from 0.95 for Late Synthetic High Yield to 1.27 for the ‘Alamo’ and ‘Summer’ blend. Eberhart and Russell’s deviation from regression ( ) values for all cultivars were not different from zero, except for PMT-279 ( = 3.07). Correlation coefficients (both Pearson’s and Spearman’s ranks) were significant for biomass yield and the regression coefficient () and for the pair combinations between Wricke’s , Shukla’s , and Shukla’s (Table 9). Figure 3 illustrates the relationship between biomass yields and regression coefficients (). Correlations between any other pairs of combinations were insignificant. In general, the results indicated relatively high biomass yield stability for the 10 cultivars evaluated for stability in the study.

Table 8. Summary of mean biomass yield and four stability parameters for each of the cultivars evaluated.

Entry	Mean yield t/ha	Wricke's	Shukla's	Shukla's	Finlay & Wilkinson's	Eberhart & Russell's
AlSummer	16.2	19.88	1.57	0.39	1.27	1.89
KanAlamo	15.5	23.96	1.96	0.46	1.32*	1.66
Kanlow	15.4	49.23*z	4.39*	1.53	1.21	2.18
Alamo	14.9	27.55	2.31	0.80	1.16	2.35
PMT-279	14.3	32.96*	2.82*	1.02	1.00	3.07*
BlKanlow	13.8	31.67*	2.70*	0.99	1.05	3.69
Blackwell	11.6	45.99*	4.09*	1.16	0.79	1.26
Caddo	11.4	12.05	0.82	0.25	0.96	0.78
CIR	11.2	48.38*	4.31*	1.53	0.96	0.92
LateSyn	9.9	29.95	2.53	0.84	0.95	1.72

*= significant at a=0.05

Table 9. Correlation coefficients between yield and stability parameters (Pearson's above diagonal and Spearman's below diagonal).

Variable	Yield
Yield	1.0000	-0.1476	-0.1478	-0.1636	0.8468*	0.4753
	0.2121	1.0000	0.9999*	0.9681*	-0.3068	0.0295
	0.2121	1.0000*	1.0000	0.9674*	-0.3077	0.0273
	0.2492	0.9969*	0.9969*	1.0000	-0.3086	0.1368
	-0.8997*	-0.2736	-0.2736	-0.3018	1.0000	0.2416
	-0.4061	0.1758	0.1758	0.1459	0.3891	1.0000

*= significant at a=0.05

Fig. 3. Relationship between regression coefficients and mean biomass yields.

The results are of practical significance because they demonstrate the ability of adapted switchgrass cultivars to maintain good stands and high biomass production potential over a long period of time. The high mean DM yields and relatively good stability of ‘Alamo’ (=14.9 t DM ha^-1, =1.16, =2.35, = 0.80) and ‘Kanlow’ ( = 15.4 t DM ha^-1, =1.21, =2.18, = 1.53) make them choice candidates for use as bioenergy feedstock crops under the conditions tested. ‘Alamo’ and ‘Kanlow’ maintained relatively good DM yields during the years of lowest mean DM production at Chickasha (1998’s =6.7 t DM ha^-1, ‘Alamo’s’ =6.7 t DM ha^-1, ‘Kanlow’s’ =7.5 t DM ha^-1) and Haskell (1997’s =9.7 t DM ha^-1, ‘Alamo’s’ =13.0 t DM ha^-1, ‘Kanlow’s’ =13.4 t DM ha^-1). As new switchgrass breeding lines and cultivars are performance tested in different environments, the use of stability parameters will enhance the effectiveness of identifying the most stable cultivars.

REFERENCES

• Anon. 1954. Indian grass (Sorghastrum nutans L. Nash) and switchgrass (Panicum virgatum L.). Oklahoma Agr. Res. Sta. Forage Crops Leaflet 17. Stillwater, OK.
• Department of Energy/Biopower Program. 1999. Review report of the regional biomass energy program technical projects. Washington, DC.
• Department of Energy/Biopower Program. 2000. A strategy for the future. Washington, DC.
• Eberhart, S.A. and W.A. Russell. 1966. Stability parameters for comparing varieties. Crop Sci. 6:36–40.
• Finlay, K.W. and G.N. Wilkinson. 1963. The analysis of adaptation in a plant breeding programme. Austral. J. Agr. Res. 14:742–754.
• Freeman, G.H. and J.M. Perkins. 1971. Environmental and genotype-environmental components of variability. VIII Relation between genotypes grown at different environments and measures of these environments. Heredity 27:15–23.
• Kang, M.S. and J.D. Miller. 1986. Genotype × environment interaction for cane and sugar yield and their implications in sugarcane breeding. Crop Sci. 24:435–440.
• Lin, C.S., M.R. Binns, and L.P. Lefkovitch. 1986. Stability analysis: Where do we stand? Crop Sci. 26:894–900.
• McLaughlin, S.B. 1993. New switchgrass biofuels research program for the southeast. p. 111–115. In: Proc. 1992 Annual Automotive Technol. Dev. Contractors Coordinating Meeting. 2–5 Nov. 1992. Dearborn, MI.
• Porter, C.L. 1966. An analysis of variation between upland and lowland switchgrass, Panicum virgatum L., in central Oklahoma. Ecology 47:980–992.
• SAS Institute Inc. 2000. SAS/STAT User’s Guide. Release 8.1. Cary, NC.
• Shukla, G.K. 1972. Some statistical aspects of partitioning genotype-environment components of variability. Heredity 29:237–245.
• Sladden, S.E., D.I. Bransby, and G.E. Aiken. 1991. Biomass yield, composition, and production costs for eight switchgrass varieties in Alabama. Biomass Bioenergy 1:119–122.
• Steel, R.F.D. and J.H. Torrie. 1980. Principles and procedures of statistics: A biometrical approach. 2nd ed. McGraw-Hill, NY.
• Wricke, G. 1962. Uber enie methode zur erfasssung der okologischen streubreite in feldversuchen. Z. Pflanzen. 47:92–96.