•Relationships between crop reflectance in the visible and the near infrared wavelengths are closely correlated with the amount of photosynthetically active tissue in the crop. Reflectance measurements were used to quantify genotypic differences in light interception, dry matter (DM) conversion efficiency and senescence pattern within the genus Miscanthus. The aim was to verify this method as a selection tool in plant breeding programmes.
•Spectral reflectance of nine genotypes was measured weekly throughout their second and third growing seasons in a field experiment conducted in Denmark. Leaf greenness was assessed by visual scoring.
•Significant differences between genotypes in the calculated fraction of PAR intercepted in green tissue (f ipar ) occurred mainly early and late in the growing season. The f ipar values correlated well with visual estimates of leaf greenness. Within genotypes accumulated intercepted PAR ranged from 632 to 737 MJ m −2 in the third year, while the DM : radiation quotient, ɛ, ranged from 1.06 to 2.53 g MJ −1 .
•Yield variation between genotypes was mainly caused by differences in ɛ. Measuring spectral reflectance was less time consuming than visual leaf scoring. The significant physiological variation within the genus Miscanthus gives good prospects for future breeding.
The perennial C4-grass miscanthus originating from south-east Asia has been extensively investigated for its biomass production-potential in Europe in the 1980s and 1990s (Lewandowski et al., 2000; Jones & Walsh, 2001). Compared with other C4-genera, miscanthus is more tolerant to the cool climate of north-west Europe (Beale & Long, 1995). Once established, miscanthus is harvested annually and in Denmark needs a rotation of minimum 10–12 yr in order to depreciate establishment costs (Parsby, 1996). The European investigations during the first decade were almost exclusively conducted with one genotype, the sterile, triploid hybrid M. × giganteus Greef & Deuter ex Hodkinson & Renvoize (Hodkinson & Renvoize, 2001; similar to M. × ogiformis Honda ‘Giganteus’ (Linde-Laursen, 1993)). In northern Europe M. × giganteus was difficult to establish and had a rather poor combustion quality because it did not senesce, which delayed leaching of minerals from the crop during winter (Jørgensen, 1997; Venendaal et al., 1997). Therefore, the genetic base of miscanthus has been broadened in Europe by collecting and screening existing genotypes (Jørgensen, 1997; Eppel-Hotz et al., 1998; Jones & Walsh, 2001) and by developing breeding methods for miscanthus (Deuter & Abraham, 1998).
During the ‘European Miscanthus Improvement’ (EMI) project, 15 genotypes were grown at five sites from Portugal to Sweden, and their biomass production and chemical composition were measured (Clifton-Brown et al., 2001). Spectral reflectance of the genotypes established at the Danish site was measured in 1998 and 1999. The aim was to quantify genotypic differences in light interception and senescence pattern. The ultimate goal was to use the method as a selection tool in plant breeding. Spectral reflectance measurements of miscanthus genotypes have not been reported before.
Relationships between reflectance in the visible and in the near infrared spectral regions (Vegetation Indices (VI)) are closely correlated with the amount of photosynthetically active tissue in plant canopies (Wiegand & Richardson, 1990; Myneni & Williams, 1994). The fraction of photosyntethically active radiation intercepted (fipar), which is an important parameter for crop growth modelling, may be directly derived from VI (Christensen, 1992; Christensen & Goudriaan, 1993). We have confirmed the relationship between VI and fipar in miscanthus and have shown that, compared with the traditional use of line quantum sensors, spectral reflectance measurements provided a better estimate of PAR-interception in green leaves during the late part of the growing season (Vargas et al., 2002).
Hand-held equipment for remote sensing of spectral reflectance has been developed and the VI of one experimental unit can be measured and logged in less than 1 min. Remote sensing of canopy spectral reflectance is a rapid, accurate and nondestructive tool for screening genotypes for their development and production capacity (Christensen, 1992). Furthermore, spectral reflectance may be used to describe differences in the course of senescence, which influences biomass quality for combustion (Jørgensen, 1997). In this article we present miscanthus seasonal fipar, accumulated intercepted PAR, biomass yield, DM : radiation quotient and crop greenness of nine miscanthus genotypes.
Materials and Methods
Design and management of field experiments
The study was conducted on a sandy loam soil (typic Fragiudalf (USDA soil taxonomy)), at the Danish Institute of Agricultural Sciences, Research Centre Foulum, Jutland, Denmark (56°30′ N, 9°35′ E). Fifteen genotypes were planted in 1997 in a randomised complete block design in three replications. The plants were evenly distributed in 25 m2 plots at a density of two plants m−2. Weeds were managed manually/mechanically in 1997 and 1998. In 1999 the area was treated with glyphosate before emergence of the miscanthus shoots. Later, miscanthus was sprayed twice with 1 l ha−1 of ‘Flux Extra’ herbicide (80 g l−1 Fluroxypyr, 40 g l−1 Clopyralid, 100 g l−1 MCPA and 50 g l−1 Dicamba). Miscanthus was harvested on 19 November, 1998 and 18 October, 1999 for determination of above-ground DM. Further details regarding soil composition and harvest procedure are provided in Clifton-Brown et al. (2001). Climatic data were collected at a meteorological station 1 km from the research site. The mean air temperature during the 1998 growing season was slightly below the long-term average, while it was 0.7°C above the long-term average in 1999 (Table 1). Precipitation was close to normal in 1998 and above normal in 1999.
Table 1. Climatic data for the growing seasons (May–September) 1998 and 1999 compared with the long-term mean (1961–90)
Air temp. (°C)
Global radiation (MJ m−2)
The experiment included four acquisitions of M. × giganteus (genotype 1–4), which did not survive the first winter in Denmark (Jørgensen & Schwarz, 2000). Genotypes 5 and 9 poorly survived the first winter, which resulted in 33 and 40% of the original plant stands, respectively. Therefore, the measurements of spectral reflectance in 1998 and 1999 only included nine genotypes (Table 2).
Table 2. Genotypes planted at Foulum, Denmark, 1997 and measured for spectral reflectance in 1998 and 1999
M. sinensis triploid hybrid from cross pollination in a miscanthus population
Hybrid between two M. sinensis
Hybrid between M. sacchariflorus and M. sinensis
Hybrid between M. sacchariflorus and M. sinensis
M. sinensis genotype collected in central Honshu, Japan (88–110 * )
M. sinensis genotype collected in central Honshu, Japan (88–111 * )
M. sinensis genotype collected in central Honshu, Japan (90–5 * )
M. sinensis genotype collected in central Honshu, Japan (90–6 * )
M. sinensis genotype collected in central Hokkaido, Japan
‘Crop greenness’ was determined as a measure of crop senescence by measuring the leaves on the tallest shoot of three randomly selected plants within a plot. Leaves were scored as green if they had more than 60% green leaf area. ‘Greenness’ was estimated as the relation between number of ‘green leaves’ and the total number of leaves on the shoot.
Canopy reflectance was measured weekly, beginning at first leaf development and stopping at first severe frost (< −2°C at 2 m height). Measurements were obtained by using a SDL 1800 two-band sensor from Skye Instruments Ltd, Llandrindod Wells, Wales, UK and logged on a Hewlett Packard HP95LX computer. The SDL 1800 was fitted with two sensors each consisting of two silicon photodiodes equipped with specific interference filters for the wavelength intervals 640–660 nm (red) and 790–810 nm (infrared). The sensors were mounted on a portable retractile mast. One sensor was cosine-corrected and used to measure the incoming radiation. The other sensor without diffuser had a restricted view angle of 15°. The sensor was kept inverted and was used for measurements of reflected radiation. The sensors were placed 4.5 m above ground to be able to cover a 1.1-m2 ground-level area. In each plot, four repeated measurements were taken centred above each of two plants in the plot centre to reduce the effect of leaf flutter. For each plot, only the mean of the eight measurements was used in the forthcoming calculations of statistical differences. Reflectance of radiation from the soil in 1999 was measured in a plot kept free of vegetation by repetitive soil rotary cultivation. Canopy reflectance was subsequently corrected for reflectance from the soil.
The Ratio Vegetation Index (RVI) was calculated as the ratio between the reflection of incoming infrared radiation (ρi) and the reflection of incoming red (ρr) radiation. The fraction of PAR intercepted (fipar) was subsequently calculated by iteration from the theoretically based model derived by Christensen & Goudriaan (1993):
( Eqn 1)
and ρi,s and ρr,s are the reflectance from bare soil.
ρi,∞ and ρr,∞ are the reflectance from the crop at maximum leaf area.
The incoming PAR (Sp) was estimated as 50% of the measured global radiation. The daily intercepted PAR (IPAR) during the growing season was calculated for each genotype by multiplication of Sp with estimates of fipar obtained by linear interpolation between measurement dates. Accumulated IPAR (AIPAR) was summed over the growing season for each genotype:
AIPAR =ΣSpfipar(Eqn 4)
The DM : radiation quotient (ɛ, sometimes called radiation use efficiency) was calculated for each genotype from the above-ground DM yield harvested in autumn:
ɛ= DM/AIPAR(Eqn 5)
Genotypic differences were evaluated on the basis of measurement standard deviation and of Tukey's minimum significant difference calculated by the GLM procedure (SAS Institute, ver. 6.12).
Fraction of PAR intercepted in green leaves
To calculate fipar from eqn 1, the reflectance of bare soil and of crop at maximum leaf area was estimated (Table 3). The bare soil values were means of measurements in an unvegetated plot throughout 1999, and the maximum crop values were means of genotype 13 on 23 July and 12 August 1999.
Table 3. Parameter values for Eqn 1 estimated from reflectance measurements above bare soil and above miscanthus at maximum leaf area in 1999
Reflectance from bare soil
ρi,s = 0.174
ρr,s = 0.111
Reflectance at max. leaf area
ρi,∞ = 0.469
ρr,∞ = 0.0262
All miscanthus genotypes were still in the establishment phase in 1998 and matured in 1999. Accordingly, mean canopy development was c. 30 d earlier in the third than in the second growth period thus resulting in a higher maximum PAR interception (Fig. 1). The initial decline of fipar in 1998 was due to an early weed cover. Weeds were removed by row cultivation on day 153 and by hand-weeding on day 159.
The fipar in green leaves of genotype 10 was higher than that of other genotypes in 1998. Genotype 10 differed significantly from the other genotypes from the third measurement date (19 June) until 12 August (Fig. 2). Genotypes 10 and 15 started senescing before other genotypes after the measurement on day 267 (24 September). On day 286 (13 October) fipar of genotype 15 was significantly lower than that of other genotypes. On day 303 (30 October), just before the first frost, fipar of genotypes 10 and 15 was significantly lower than that of all other genotypes.
In 1999, crop development was again earliest in genotype 10, but also genotype 8 had significantly higher fipar than the remaining genotypes in May and early June (Fig. 2). From mid-July to the end of August, genotype 13 had the significantly highest fipar reaching 0.89 on Day 224 (12 August). A slight reduction in fipar occurred for all genotypes in late August followed by a more significant reduction in October. Genotype 15 senesced first and had significantly lower fipar than the remaining genotypes on day 291–298 (18–25 October), while on day 306 (2 November) genotypes 8 and 10 reached a similarly low level as genotype 15. First frost occurred on day 314 (10 November).
Accumulated intercepted PAR (AIPAR)
There was a nearly two-fold increase in AIPAR, calculated according to eqn 4, from 1998 to 1999 (Table 4). In 1998, genotype 10 from the beginning of the season differed from all other genotypes by intercepting more PAR, and at the end of the season it had accumulated more than 100 MJ m−2 more than any other genotype (Table 4). In 1999, the highest accumulation of intercepted PAR occurred in genotypes 8, 13 and 10.
Table 4. Dry matter yield at harvest in autumn (DM), accumulated intercepted PAR (AIPAR) and DM:radiation quotient (ɛ) of nine genotypes in 1998 and 1999. The Tukey minimum significant difference ( P = 0.05) is given for each parameter
DM (g m−2)
AIPAR (MJ m−2)
ɛ (g MJ−1)
Yield and dry matter : radiation quotient (ɛ)
DM yield in autumn increased in most genotypes with a factor of c. 3 from 1998 to 1999 (Table 4). The yield was highest in genotype 10 in both years. However, in 1999 genotypes 6, 8 and 11 also yielded 15 t DM ha−1 or more. These genotypes were amongst those that intercepted the most PAR. However, genotype 13 intercepted the second highest amount of PAR in 1999, but was only ranked sixth with respect to yield. Accordingly, when calculating the DM : radiation quotient (eqn 5) genotype 13 had a significantly lower ɛ of 1.53 g MJ−1 compared with the value of 2.53 g MJ−1 in genotype 10 (Table 4).
The ɛ-value increased from 1998 to 1999, but the ranking of the genotypes was similar regardless of year. Genotypes 14 and 15 had the lowest ɛ-values.
The visually scored ‘greenness’ values are plotted next to the fipar-values in Fig. 3. Some graphs show mean greenness of two to three genotypes, which were very similar. The shape of the two curves in each graph is similar until day 300, when the ‘greenness’ decreases at a higher rate.
The fraction of intercepted PAR in green tissue was successfully estimated from the measured spectral reflectance of the miscanthus genotypes. Vargas et al. (2002) showed that the use of VI in M. sinensis‘Goliath’ provided a better description of green leaf area development than did the use of solarimeters, especially late in the growing period when leaves senesced. Solarimeters inserted above and below the canopy or a ‘Sunfleck Ceptometer’ have hitherto been the methods used for fipar estimation in miscanthus (van der Werf et al., 1993; Beale & Long, 1995; Greef, 1996; Bullard et al., 1997; Vleeshouwers, 1998; Clifton-Brown et al., 2000). By using these methods, however, radiation intercepted by nonphotosynthetically active tissues (senescent leaves, stems, and flowers) is included. Greef (1996) kept the lower solarimeter above the senesced leaves at the bottom of the crop so that they did not influence the measurement. Beale & Long (1995) in addition made a visual estimation of percentage senesced shoots and flowers in the upper canopy to correct the measured interception. Neither of these methods accounted for dynamic changes in the canopy of leaves, shoots and inflorescences. Like Gallo et al. (1993), we recommend spectral reflectance measurements for the estimation of intercepted PAR in photosynthetically active tissue.
The fipar-values calculated from the measured spectral reflectance reached a maximum of c. 0.8, which was considerably lower than the values obtained by the solarimeter/Sunfleck Ceptometer in other studies (van der Werf et al., 1993; Beale & Long, 1995; Greef, 1996; Bullard et al., 1997; Vleeshouwers, 1998; Clifton-Brown et al., 2000), where fipar values above 0.95 were reached in the late part of the growing season. Senescence of leaves generally starts in mid- to late summer in M.×giganteus and M. sinensis (Greef, 1996; Jørgensen, 1997), which is probably the main reason for the lower fipar calculated from spectral reflectance compared with the solarimeter-based values (Vargas et al., 2002). Also, development of flowers may reduce the PAR interception in green tissue significantly, such as in rape (Andersen et al., 1996). However, in miscanthus fipar-values decreased only slightly after initial flowering around day 240 in 1999 (Fig. 2).
The changes in leaf greenness followed changes in fipar very closely (Fig. 3). Making visual estimates took six to seven times longer to determine than those of spectral reflectance. Furthermore, spectral reflectance is better suited to integrate uneven patterns of senescence. Physiological senescence before the crop is killed by frost is essential to mediate removal of the elements K and Cl by leaching of the plant tissue (Jørgensen, 1997). These elements are harmful to biomass combustion processes (Sander, 1997). Most genotypes senesced before the first frost, but genotype 7 still had c. 40% green leaves at the time of frost. This stay-green trait of genotype 7 was also observed at the other European sites (Clifton-Brown et al., 2001).
In 1999, the DM : radiation quotient (ɛ) calculated from above-ground harvested biomass in the autumn was higher than that in 1998. The main reason was probably that miscanthus allocates a large proportion of DM into roots and rhizomes below ground in the establishment phase (Greef, 1996). It is likely that the ceiling yield level of a fully established crop was almost reached in 1999. Therefore, ɛ was not much affected by augmented below-ground translocation. The average yield of October and February harvest of the nine genotypes in 1999–2000 was 11.0 t DM ha−1, in January 2001 it was 11.4 t DM ha−1, and in January 2002 it was 12.8 t DM ha−1 (U. Jørgensen, unpublished). Another reason for the increased ɛ from 1998 to 1999 might have been the higher temperatures in 1999 (Table 1). Photosynthesis in miscanthus is very temperature sensitive in the range from 7 to 25°C (Sawada & Iwaki, 1978; van der Werf et al., 1993; Beale et al., 1996). Finally, the contribution from weeds to AIPAR early in 1998 (Fig. 1) might also have contributed to the difference in ɛ-value between years.
The calculated ɛ-values ranged from 1.06 g MJ−1 in genotype 15 to 2.53 g MJ−1 in genotype 10 in 1999. This is a very high intraspecies variation compared with the results of 12 cultivars of spring barley (Hordeum vulgare L.) grown in Denmark (Christensen & Goudriaan, 1993). Significant differences in the above-ground growth rate of cultivars during the season were calculated, but no significant difference in ɛ over the whole season was detected (Christensen, 1992). The miscanthus genotypes differed in time of onset of senescence. We measured biomass yield at one point in time during crop senescence (18 October 1999), and the genotypes might have translocated different proportions of reserve material to rhizomes at that time. It is likely that a 1-month difference in the onset of DM translocation to rhizomes could result in a difference of up to 20% in the measured above-ground DM (based on Bullard et al., 1995; Greef, 1996; Jørgensen, 1997). However, even with a 20% change in DM yield-difference between genotype 10 and 15, ɛ would still be c. two times higher in genotype 10.
The genotypes 14 and 15 were the first ones to start flowering in July (Clifton-Brown et al., 2001) and also had the lowest ɛ, which might indicate that flowering uncoupled photosynthesis in still green leaves similar to the effect of drought in certain miscanthus genotypes (Clifton-Brown et al., 2002). However, genotype 11 flowering in August had a higher ɛ than genotypes 7 (did not flower) and 8 (flowering in October). The genotypes 6 and 10, flowering in September had the highest ɛ. Whether there is a causal relation between flowering time and the size of ɛ therefore needs further investigation.
It is difficult to compare ɛ from different studies because of differences in the methods for measuring PAR interception, different periods of biomass accumulation (Gallo et al., 1993), and the impacts of other limiting factors such as water, temperature and nutrients (Demetriades-Shah et al., 1992). However, interspecies variation in ɛ of C4-grasses grown under similar conditions and measured by similar methods has been demonstrated by Beale & Long (1995), who estimated ɛ of Spartina cynosuroides to 2.1 g MJ−1 and of M. ×giganteus to 3.3 g MJ−1. Kiniry et al. (1999) measured ɛ of four grasses and reported three to four times higher values in switchgrass (Panicum virgatum) compared with sideoats grama (Bouteloua curtipendula).
In a detailed study of M. sinensis‘Goliath’ at Research Centre Foulum in 1996, Vargas et al. (2002) estimated an ɛ-value of 1.90 g MJ−1 from VI-calculated fipar. This value was similar to the average value of the genotypes in 1999 (Table 4). Other investigations in temperate, but warmer climates than the Danish, have reported higher ɛ-values of M. × giganteus even when PAR-interception was measured by solarimeters or ‘Sunfleck Ceptometer’, which reflects the high temperature response of the photosynthetic production of a C4-crop in a cool climate: 2.6 g MJ−1 in the Netherlands (van der Werf et al., 1993), 3.3 g MJ−1 in south-eastern England (Beale & Long, 1995), 2.4 g MJ−1 in northern Germany (Greef, 1996), 3.3 g MJ−1 in the Netherlands (Vleeshouwers, 1998) and 2.4 g MJ−1 in Ireland (Clifton-Brown et al., 2000).
Genotype 10, which was a hybrid between M. sinensis and M. sacchariflorus showed some remarkable advantages compared with the other genotypes. Firstly, genotype 10 survived well and provided a complete plant cover. Secondly, it had an early vigorous growth in the second growing season and the highest shoot density (U. Jørgensen, unpublished), which resulted in significantly higher accumulation of intercepted PAR. Under practical agronomic conditions, early leaf development is important for the crop to compete successfully with weeds and to reduce the need of weed management. Thirdly, genotype 10 had the highest calculated ɛ-values in both years. The last important observation was that even though genotype 10 remained green throughout the autumn, the green leaf area decreased rapidly just before the first frost.
Miscanthus is a wild grass that only recently has been cultivated for agricultural purposes (Jørgensen & Schwarz, 2000). ‘The developmental stage of the dedicated energy crops is at stone-age level compared with conventional agricultural crops, which have been bred and investigated for centuries’, said the late professor David Hall, Kings College, London, at the first European Energy Crops Conference in 1996. Our results showed that even within the current European gene pool there is a high physiological variation, which provides good prospects for future breeding of improved miscanthus genotypes adapted to different climatic regions in Europe. When comparing the 1999 AIPAR and ɛ of the lowest yielding genotype 15 with genotype 10 it appeared that AIPAR was only 12% higher in the high yielding genotype, while ɛ was 139% higher (Table 4) and thus clearly more important to the 168% yield increase. Similarly, Tollenaar & Aguilera (1992) attributed 80% of the yield increase of a new maize hybrid to improved ɛ and only 20% to improved PAR absorption.
The use of spectral reflectance seemingly is an efficient selection tool for miscanthus genotypes with optimal traits as the spectral technique offers the possibility to distinguish between PAR interception and conversion efficiency. Furthermore, a description of green canopy development during the growing season may reveal and quantify important traits, such as early leaf area development and course of senescence.
This study was funded by the EU contract no. FAIR3 CT-96–1392. We thank Kaj Eskesen and Jens Bonderup Kjeldsen for their measurements in the miscanthus field.