- Email: [email protected]
An evaluation of a rapid visual technique for estimating seagrass biomass
Aquatic Botany, 42 ( 1991 ) 67-73
67
Elsevier Science Publishers B.V., Amsterdam
Technical Communication An evaluatio of a rapid visual technique for estimating seagrass biomass Jane E. Mellors Northern Fisheries Centre, Box 5396, Cairns Mail Centre, Cairns, QId. 4871, Australia (Accepted 29 May 1991 )
ABSTRACT Mellors, J.E., 199 I. An evaluation of a rapid visual technique for estimating seagrass biomass..4quat. Bot., 42: 67-73. A visual census technique for estimating seagrass biomass has been adapted from a comparative pasture yield method. The above-ground biomass of seagrass within sampling quadrats was ranked with respect to a set of reference quadrats which were presdected to provide a scale of standing crop dry weights. At the end of each sampling period, sufficient quadrats were harvested to calibrate the scale. Using this method, monthly mean standing crops were estimated from May 1987 to April 1988 for a multispecifie seagrass bed on Green Island, North Queensland. Values obtained ranged between 61.52 and ! i 3.08 g dry weight m -2. The precision (SE/X') ofeach monthly estimate ranged from 0.05 to 0.13, a satisfactory level for field programs. This method is more precise and time efficient, and is less destructive than some traditional harvesting methods.
INTRODUCTION
Assessments of biomass, production, nutrient cycling and community dynamics all rely, to a degree, on estimates of macrophyte standing stock, which in turn are necessary for the assessment of habitat resources (Downing and Anderson, 1985 ). Given the costs of fieldwork, it is advantageous to be able to take measurements quickly and accurately. Measurement of ~.he aboveground biomass of seagrass, which is usually estimated by quantitative harvesting of seagrass contained in randomly placed samplers (Downing and Anderson, 1985 ), is not only costly, but destructive. Quantitative harvesting measures the biomass of each sample accurately (exactly). However, the limitation remains that each measurement represents only one sample from a very variable seagrass bed. The main problem lies in the variability in the set of measurements, rather than the accuracy with which an individual sample is measured. Consequently, many samples 0304-3770/91/$03.50 © 1991 Elsevier Science Publishers B.V. All rights reserved.
68
J.E. MEI.L()RS
estimated with an acceptably lower accuracy are better than a few samples measured exactly, provided there is no systematic error. In this paper, a visual technique for estimating the above-ground biomass of seagrass is described and evaluated, and compared with more traditional methods. The technique is an adaptation of a comparative pasture yield method (Haydock and Shaw, 1975; Tothill et al., 1978) and was developed as part of a study on the intra-annual variations of seagrass habitat and its effect on the associated crustacean fauna at Green Island, North Queensland, Australia. MATERIALS AND METHOD
Study site Green Island (16°46'S; 145°58'E) is a vegetated sand cay on the inner edge of the Great Barrier Reef, approximately 27 km offshore from Cairns, North Queensland, Australia. It is a marine park and flora or fauna cannot be removed without permits. An extensive multi-specific se~£rass meadow interspersed with coral bommies surrounds most of the cay. The meadow is subtidal and densest on the sheltered north northwest side of the island. Here, ltalodtde uninervis (Forsk.) Aschers. (broad-leaf variety) predominates, followed by Cymodocea serrulata (R. Br.) Aschers. and Magnus, and Cymodocea rotundata Ehrenb. et Hempr. ex Aschers. Halophila minor (Zoll.) den Hartog is found scattered throughout the meadow. On the northeast and southern sides of the island, the bed is less dense and intertidal, and Thalassia hemprichii (Ehrenb.) Aschers. and C. rotundata are the main seagrasses present. Sampling was restricted to the subtidal, dense northwestern bed, between May 1987 and April 1988. During this time, underwater visibility varied from 2 to 12 m.
Description of vL~ualtechique The technique At the start of each sampling period, tive reference quadrats were selected to represent the scale against which the above-ground seagrass biomass in each sample quadrat was co,npared. These ~eference quadrats encompassed the range of seagrass biomasses likely to be encountered and were agreed on by those undertaking the sampling. The five reference q,adrats were ranked 1 (least) to 5 (most) on a linear scale. The reference quadrats for Ranks 1 and 5 were selected first, followed by the Rank 3 quadrat (estimated to have a dry weight (dw) yield of above-ground biomass half-way between that of Ranks 1 a,~d 5 ) and finally the Rank 2 and 4 quadrats halfway between Ranks 1 and 3, and 3 and 5, respectively. The reference quadrats were left in place
E S T I M A T I N ( I SEA(iR AF,S BII }M.'~SS IIY V ISI t.'t.I_ ('ENS( IS
69
until sampling was completed to provide observers with the opportunity to reacquaint themselves with the scale.
Sampling Observers swam along four transects, each 200 m long and 50 m apart. using a compass for direction. Quadrats (0.25 m z, the largest size that could be handled conveniently under:eater) were placed at intervals of 25 swim beats (equivalent to 20 m) along each transect. An observer ranked the seagrass standing crop in each quadrat (correc' to one decimal place) according to the pre-delermined scale. If quadrats occ,~rrpd on coral clumps or other .4~rl non-seagrass areas, data were trc,,,,,u as missing in the analysis; bare sand patches were considered a potential seagrass habitat and were ranked as zero.
C~'osscalibration of oh~ervers At the completion of each sampling period, the observers involved crosscalibrated their rankings by individually ranking ten quadrats to cover the range of biomasses encountered. These quadrats were then harvested by digging up the total biomass of the seagrass. In the laboratory, each excavated quadrat sample was washed and the aboveground biomass separated from the below-ground biomass. Samples were not acid treated as there was little contamination with epiphytes and sediment. The samples were then oven-dried to a constant weight (80°C for 48 h) and weighed to 0.01 g. For each sampling period, calibration curves for each observer were established by regressing above-ground dry weights against the corresponding rank for the ten calibration quadrats. In most cases, the relationship was close to linear. Zero values were retained in the analysis as a measure of the patchiness within the seagrass bed. The regresgion line was not forced through the origin, as this would have unduly influenced the entire relationship. Occasionally, low ranks predicted negative values due to large intercepts calculated by the linear regression. When this occurred, the smallest positive value determined by the accuracy of the balance (0.01 g) was inserted for that rank.
Intertidal environments The technique was also tried on a sparse intertidal seagrass bed with four observers. They paced rather than swam the required distance along the transects. Other procedures remained the same.
Comparison of methodologyfrom otherstudies For each month of sampling, the precision (SE/£; Andrew and Mapstone, 1987) of the visual technique for estimating the above-ground biomass (g dw
70
J.E. MELLORS
0.25 m - : ) ofseagrass was compared with that obtained by the following (more tradition:'.l ) harvesting methods. ( 1 ) Four of the ten calibration qua,lrats were selected randomly to simulate the random harvest method (Aioi. 1980; Orth and Moore, 1983; Downing and Anderson, 1985; Boon, 1986). (2) The four calibration quadrats with biomass closest to the modal rank were selected to simulate the partially randomized sampling technique of Coles et al. (1987). This technique (R. Coles, personal communication, 1990) involves the harvesting of four quadrats of similar biomass that have been chusen as being 'typical' of the bed. The actual data (g dw 0.25 m -2 ) were used in all comparisons ofsampling methodologies per se, other results are presented in the more usual g dw m--'. RESULTS
lqsual technique The linear regressions relating ranks to above-ground biomass were significant for both operators for each monthly set of calibration quadrats. Coefficients of determination (r-') rag.ged from 0.65 to 0.96 (mean=0.88) for the subtidal trials and from 0.85 to 0.96 (mean=0.91) for the intertidal trial. The average monthly bion,.ass ranged between 61.52 and 113.08 g dw m -2 Precisions ranged from 0.05 to 0.13. Using this teclmique, 40 quadrats could be ranked i,~ 50 min, while observers swam over an area of 40000 m 2ofseagrass. To excavate the ten caiibration quadrats took another 84 min and laboratory processing of these quadrats was, on average, 60 rain for each excavated sample. The total processing time was 734 min per sampling period or i 8.35 min per quadrat.
Comparison of techniques The monthly mean biomass calculated for the random harvest method ranged fi'om 31.56 to 179.44 g dw m --~, while the biomasses of the partially randomized method were between 103.08 and 216.64 g dw m--'. Estimates of mean biomass per sampling period were significantly different between sampling methods (two-way ANOVA; F = 13.26, 2 × 22 d.f., P < 0.001 ) as were biomass estimates between sampling periods (two-way ANOVA; F=4.6!, I 1 X 22 d.f., P < 0.001 ). The least significant difference pairwise comparisons (LSD) between sampling methods recognized two homogeneous groups (Fig. I a): partially randomized sampling was significantly different from the other two methods. Precisions for the random harvest method were the most variable (0.I 50.60) while those calculated for the partially random harvest method were
|!STIM.~.TIN(i SE3,(iRASS BIOMASSBY 'ISUALI.'ENSt:5;
7]
3,0.00
~E 2 8 . 0 0 tt]
¢5
L,S.D.
26.00 'O o
~
24,00
_o en 2 2 . 0 0 Z ILl
:E 20.00 I
•
b -0.20 Z o
-0.40 I L,S.D.
w e¢-
O -O.60 O ,,J
z < tu
-0.80
-1.00
-'i . 2 0
I
-.
VISUAL
I
RANDOM HARVEST
I
PARTIALLY RANDOMIZED
SAMPLING METHOD Fig. 1. Comparisonsof (a) biomassand (b) precisionfor each samplingmethod. Leaslsignificant differences(LSD) are shown. between 0.12 and 0.29. Results of the two-way ANOVA showed that precision (log transformed to equalize variances) was significantly different between sampling methods ( F = 66.45, 2 X 22 d.f., P < 0.001 ), but not significantly different between sampling periods ( F = 1.74, 11 X 22 d.f., P > 0.05 ). The LSD pairwise comparison of precision for sampling methods recognized three homogeneous groups (Fig. I b). The time taken to visually estimate 40X0.25 m 2 quadrats was 2.7 times longer than that taken to harvest four quadrats, but was only 27% of the time required to harvest 40 quadrats.
72
J.E. MELLORS
DISCUSSION
The major goal in designing a sampling program is to achieve an accurate measurement with high precision for the least effort (Downing and Anderson, 1985 ). The visual technique does this by increasing the number of samples that can be measured, ensuring that many areas are assessed within a single bed. The regression relationship between rank and standing crop was generally linear, although on occasions it can be quadratic due to the difficulty of visually estimating low biomass. In this circumstance, a quadratic function may be used provided it does not unduly bias the predicted values at the extremes, particularly the lower end of the scale. Coefficients &determination for the observed relationships were high with the exception of 0.65 (November). The second assessor on this occasion was untrained and apparently used percentage cover rather than above-ground biomass per se as the basis for ranking. Not surprisingly, the mean biomass estimale of the partially random harvest method was significantly different from those derived by the other two methods (Fig. l a). The partially random method overestimated biomass (Fig, I a) as this technique chooses quadrats of similar biomass that are considered to be representative of the bed being sampled. Consequently, this method tends not to include samples of low biomass. The visual and the random harvest techniques returned similar biomass estimates (Fig. l a), suggesting that the visual technique is accurate. A prime objective of any census technique is to provide consistent results. Of the three methods examined, only the visual technique approached the 0.1-0.2 level of precision aimed for in field programs (Thresher and Gunn, 1986). This precision is achieved due to the large number of replicates that can be taken. Differences in the precision of the two harvest methods (Fig. I b) reflected the variability in the quadrats chosen for harvesting.
Recommendationsfor using the visual technique Observers should undertake a training program before using this technique. During training, it should be emphasized that observers consider the area of bare ground between plants, plant height and the moisture content of each species to avoid confusing the dry weight biomass with percentage cover. To help observers become awe.re of the moisture content of different species of seagrass, they should weigh wet and dried samples of different species in the laboratory before field sampling. In seagrass beds of distinct species zonation, more than one set of calibration quadrats may be needed to take account of different moisture contents and growth habits of the species being estimated (e.g. ltalophila cf. Cymodocea).
ESTIMATING SEAGRASS BIOMASS BY VISLIAL CENSUS
73
The visual technique for estimating seagrass biomass returns adequate precision with greater time efficiency per quadrat than other commonly used sampling zechniques and is much less destructive than harvesting a large n u m b e r of quadratt. I consider that the use o f this technique along permanently marked transects offers substantial benefits in monitoring seagrass over large areas over time, or where the habitat is fragile or environmentally sensitive. The disadvantages of the technique are its inapplicability to subtidal seagrass beds in turbid waters and its inability to provide a precise estimate of below-ground biomass. ACKNOWLEDGEMENTS I gratefully acknowledge Joe Miller (Pasture Management Branch, Queensland Department of Primary Industries) for demonstrating the tecl"nique, and Anto Wilson, R o b Coles and Darren Dennis for their assistance in the field. I thank Helene Marsh and two a n o n y m o u s referees for their comments on this manuscript, and the Northern Fisheries Centre, Queensland Department of Primary Industries, Cairns for providing logistical support.
REFERENCES Aioi, K., 1980. Seasonal changes m the standing crop ofeelgrass (Zostera marina L. ) in Odawa Bay, Central Japan, Aquat. Bot., 8: 343-354. Andrew, N.L. and Mapstone, B.D., 1987. Sampling and the description of spatial pattern in marine ecology.Oceanogr. Mar. Biol. Annu. Rev., 25: 39-90. Boon, P.I., 1986. Nitrogen pools in seagrass beds ofCymodocea serrulata and Zostera capricorni of Moreton Bay, Australia. Aquat. Bol., 25: 1-19. Coles, R.G., Lee Long, W.J., Squire, B.A., Squire, L.C. and Bibby, J.M., 1987. Distribution of seagrasses and associated commercial penaeid prawns in north-eastern Queensland waters. Aust. J. Mar. Freshwater Res., 38: 103-120. Downing, J.A. and Anderson, M.R., 1985. Estimating the standing biomass of aquatic macrophyles. J. Fish Aquat. Sci., 42: 1860-1869. Haydock, K.P. and Shaw, N,H., 1975. The comparative yield method ofestimatin~ dry matter yield of pasture. Aust. J. Exp. Agric. Anim. Husb., 15: 663-670. Orth, R.J. and Moore, K.A., 1983. Seasonal and year-to-year variations in the growth of Zostera marina L. (celgrass) in the lower Chesapeake Bay. Aquat. Bot., 24: 335-341. Thresher, R.E. and Gunn, J.S., 1986. Comparative analysis ofvisual census techniques for highly mobile, reef-associated piscivo, zs (Carangidae). Environ. Biol. Fish., 17: 93-116. Tothill, J.C., Hargreaves, J.N.G. and Jones, R.M., 1978. Botonal - A comprehensive sampling and computing procedure for estimating pasture yield and composition. I. Field sampling. Tropical Agronomy Technical Memorandum Number 8.
67
Elsevier Science Publishers B.V., Amsterdam
Technical Communication An evaluatio of a rapid visual technique for estimating seagrass biomass Jane E. Mellors Northern Fisheries Centre, Box 5396, Cairns Mail Centre, Cairns, QId. 4871, Australia (Accepted 29 May 1991 )
ABSTRACT Mellors, J.E., 199 I. An evaluation of a rapid visual technique for estimating seagrass biomass..4quat. Bot., 42: 67-73. A visual census technique for estimating seagrass biomass has been adapted from a comparative pasture yield method. The above-ground biomass of seagrass within sampling quadrats was ranked with respect to a set of reference quadrats which were presdected to provide a scale of standing crop dry weights. At the end of each sampling period, sufficient quadrats were harvested to calibrate the scale. Using this method, monthly mean standing crops were estimated from May 1987 to April 1988 for a multispecifie seagrass bed on Green Island, North Queensland. Values obtained ranged between 61.52 and ! i 3.08 g dry weight m -2. The precision (SE/X') ofeach monthly estimate ranged from 0.05 to 0.13, a satisfactory level for field programs. This method is more precise and time efficient, and is less destructive than some traditional harvesting methods.
INTRODUCTION
Assessments of biomass, production, nutrient cycling and community dynamics all rely, to a degree, on estimates of macrophyte standing stock, which in turn are necessary for the assessment of habitat resources (Downing and Anderson, 1985 ). Given the costs of fieldwork, it is advantageous to be able to take measurements quickly and accurately. Measurement of ~.he aboveground biomass of seagrass, which is usually estimated by quantitative harvesting of seagrass contained in randomly placed samplers (Downing and Anderson, 1985 ), is not only costly, but destructive. Quantitative harvesting measures the biomass of each sample accurately (exactly). However, the limitation remains that each measurement represents only one sample from a very variable seagrass bed. The main problem lies in the variability in the set of measurements, rather than the accuracy with which an individual sample is measured. Consequently, many samples 0304-3770/91/$03.50 © 1991 Elsevier Science Publishers B.V. All rights reserved.
68
J.E. MEI.L()RS
estimated with an acceptably lower accuracy are better than a few samples measured exactly, provided there is no systematic error. In this paper, a visual technique for estimating the above-ground biomass of seagrass is described and evaluated, and compared with more traditional methods. The technique is an adaptation of a comparative pasture yield method (Haydock and Shaw, 1975; Tothill et al., 1978) and was developed as part of a study on the intra-annual variations of seagrass habitat and its effect on the associated crustacean fauna at Green Island, North Queensland, Australia. MATERIALS AND METHOD
Study site Green Island (16°46'S; 145°58'E) is a vegetated sand cay on the inner edge of the Great Barrier Reef, approximately 27 km offshore from Cairns, North Queensland, Australia. It is a marine park and flora or fauna cannot be removed without permits. An extensive multi-specific se~£rass meadow interspersed with coral bommies surrounds most of the cay. The meadow is subtidal and densest on the sheltered north northwest side of the island. Here, ltalodtde uninervis (Forsk.) Aschers. (broad-leaf variety) predominates, followed by Cymodocea serrulata (R. Br.) Aschers. and Magnus, and Cymodocea rotundata Ehrenb. et Hempr. ex Aschers. Halophila minor (Zoll.) den Hartog is found scattered throughout the meadow. On the northeast and southern sides of the island, the bed is less dense and intertidal, and Thalassia hemprichii (Ehrenb.) Aschers. and C. rotundata are the main seagrasses present. Sampling was restricted to the subtidal, dense northwestern bed, between May 1987 and April 1988. During this time, underwater visibility varied from 2 to 12 m.
Description of vL~ualtechique The technique At the start of each sampling period, tive reference quadrats were selected to represent the scale against which the above-ground seagrass biomass in each sample quadrat was co,npared. These ~eference quadrats encompassed the range of seagrass biomasses likely to be encountered and were agreed on by those undertaking the sampling. The five reference q,adrats were ranked 1 (least) to 5 (most) on a linear scale. The reference quadrats for Ranks 1 and 5 were selected first, followed by the Rank 3 quadrat (estimated to have a dry weight (dw) yield of above-ground biomass half-way between that of Ranks 1 a,~d 5 ) and finally the Rank 2 and 4 quadrats halfway between Ranks 1 and 3, and 3 and 5, respectively. The reference quadrats were left in place
E S T I M A T I N ( I SEA(iR AF,S BII }M.'~SS IIY V ISI t.'t.I_ ('ENS( IS
69
until sampling was completed to provide observers with the opportunity to reacquaint themselves with the scale.
Sampling Observers swam along four transects, each 200 m long and 50 m apart. using a compass for direction. Quadrats (0.25 m z, the largest size that could be handled conveniently under:eater) were placed at intervals of 25 swim beats (equivalent to 20 m) along each transect. An observer ranked the seagrass standing crop in each quadrat (correc' to one decimal place) according to the pre-delermined scale. If quadrats occ,~rrpd on coral clumps or other .4~rl non-seagrass areas, data were trc,,,,,u as missing in the analysis; bare sand patches were considered a potential seagrass habitat and were ranked as zero.
C~'osscalibration of oh~ervers At the completion of each sampling period, the observers involved crosscalibrated their rankings by individually ranking ten quadrats to cover the range of biomasses encountered. These quadrats were then harvested by digging up the total biomass of the seagrass. In the laboratory, each excavated quadrat sample was washed and the aboveground biomass separated from the below-ground biomass. Samples were not acid treated as there was little contamination with epiphytes and sediment. The samples were then oven-dried to a constant weight (80°C for 48 h) and weighed to 0.01 g. For each sampling period, calibration curves for each observer were established by regressing above-ground dry weights against the corresponding rank for the ten calibration quadrats. In most cases, the relationship was close to linear. Zero values were retained in the analysis as a measure of the patchiness within the seagrass bed. The regresgion line was not forced through the origin, as this would have unduly influenced the entire relationship. Occasionally, low ranks predicted negative values due to large intercepts calculated by the linear regression. When this occurred, the smallest positive value determined by the accuracy of the balance (0.01 g) was inserted for that rank.
Intertidal environments The technique was also tried on a sparse intertidal seagrass bed with four observers. They paced rather than swam the required distance along the transects. Other procedures remained the same.
Comparison of methodologyfrom otherstudies For each month of sampling, the precision (SE/£; Andrew and Mapstone, 1987) of the visual technique for estimating the above-ground biomass (g dw
70
J.E. MELLORS
0.25 m - : ) ofseagrass was compared with that obtained by the following (more tradition:'.l ) harvesting methods. ( 1 ) Four of the ten calibration qua,lrats were selected randomly to simulate the random harvest method (Aioi. 1980; Orth and Moore, 1983; Downing and Anderson, 1985; Boon, 1986). (2) The four calibration quadrats with biomass closest to the modal rank were selected to simulate the partially randomized sampling technique of Coles et al. (1987). This technique (R. Coles, personal communication, 1990) involves the harvesting of four quadrats of similar biomass that have been chusen as being 'typical' of the bed. The actual data (g dw 0.25 m -2 ) were used in all comparisons ofsampling methodologies per se, other results are presented in the more usual g dw m--'. RESULTS
lqsual technique The linear regressions relating ranks to above-ground biomass were significant for both operators for each monthly set of calibration quadrats. Coefficients of determination (r-') rag.ged from 0.65 to 0.96 (mean=0.88) for the subtidal trials and from 0.85 to 0.96 (mean=0.91) for the intertidal trial. The average monthly bion,.ass ranged between 61.52 and 113.08 g dw m -2 Precisions ranged from 0.05 to 0.13. Using this teclmique, 40 quadrats could be ranked i,~ 50 min, while observers swam over an area of 40000 m 2ofseagrass. To excavate the ten caiibration quadrats took another 84 min and laboratory processing of these quadrats was, on average, 60 rain for each excavated sample. The total processing time was 734 min per sampling period or i 8.35 min per quadrat.
Comparison of techniques The monthly mean biomass calculated for the random harvest method ranged fi'om 31.56 to 179.44 g dw m --~, while the biomasses of the partially randomized method were between 103.08 and 216.64 g dw m--'. Estimates of mean biomass per sampling period were significantly different between sampling methods (two-way ANOVA; F = 13.26, 2 × 22 d.f., P < 0.001 ) as were biomass estimates between sampling periods (two-way ANOVA; F=4.6!, I 1 X 22 d.f., P < 0.001 ). The least significant difference pairwise comparisons (LSD) between sampling methods recognized two homogeneous groups (Fig. I a): partially randomized sampling was significantly different from the other two methods. Precisions for the random harvest method were the most variable (0.I 50.60) while those calculated for the partially random harvest method were
|!STIM.~.TIN(i SE3,(iRASS BIOMASSBY 'ISUALI.'ENSt:5;
7]
3,0.00
~E 2 8 . 0 0 tt]
¢5
L,S.D.
26.00 'O o
~
24,00
_o en 2 2 . 0 0 Z ILl
:E 20.00 I
•
b -0.20 Z o
-0.40 I L,S.D.
w e¢-
O -O.60 O ,,J
z < tu
-0.80
-1.00
-'i . 2 0
I
-.
VISUAL
I
RANDOM HARVEST
I
PARTIALLY RANDOMIZED
SAMPLING METHOD Fig. 1. Comparisonsof (a) biomassand (b) precisionfor each samplingmethod. Leaslsignificant differences(LSD) are shown. between 0.12 and 0.29. Results of the two-way ANOVA showed that precision (log transformed to equalize variances) was significantly different between sampling methods ( F = 66.45, 2 X 22 d.f., P < 0.001 ), but not significantly different between sampling periods ( F = 1.74, 11 X 22 d.f., P > 0.05 ). The LSD pairwise comparison of precision for sampling methods recognized three homogeneous groups (Fig. I b). The time taken to visually estimate 40X0.25 m 2 quadrats was 2.7 times longer than that taken to harvest four quadrats, but was only 27% of the time required to harvest 40 quadrats.
72
J.E. MELLORS
DISCUSSION
The major goal in designing a sampling program is to achieve an accurate measurement with high precision for the least effort (Downing and Anderson, 1985 ). The visual technique does this by increasing the number of samples that can be measured, ensuring that many areas are assessed within a single bed. The regression relationship between rank and standing crop was generally linear, although on occasions it can be quadratic due to the difficulty of visually estimating low biomass. In this circumstance, a quadratic function may be used provided it does not unduly bias the predicted values at the extremes, particularly the lower end of the scale. Coefficients &determination for the observed relationships were high with the exception of 0.65 (November). The second assessor on this occasion was untrained and apparently used percentage cover rather than above-ground biomass per se as the basis for ranking. Not surprisingly, the mean biomass estimale of the partially random harvest method was significantly different from those derived by the other two methods (Fig. l a). The partially random method overestimated biomass (Fig, I a) as this technique chooses quadrats of similar biomass that are considered to be representative of the bed being sampled. Consequently, this method tends not to include samples of low biomass. The visual and the random harvest techniques returned similar biomass estimates (Fig. l a), suggesting that the visual technique is accurate. A prime objective of any census technique is to provide consistent results. Of the three methods examined, only the visual technique approached the 0.1-0.2 level of precision aimed for in field programs (Thresher and Gunn, 1986). This precision is achieved due to the large number of replicates that can be taken. Differences in the precision of the two harvest methods (Fig. I b) reflected the variability in the quadrats chosen for harvesting.
Recommendationsfor using the visual technique Observers should undertake a training program before using this technique. During training, it should be emphasized that observers consider the area of bare ground between plants, plant height and the moisture content of each species to avoid confusing the dry weight biomass with percentage cover. To help observers become awe.re of the moisture content of different species of seagrass, they should weigh wet and dried samples of different species in the laboratory before field sampling. In seagrass beds of distinct species zonation, more than one set of calibration quadrats may be needed to take account of different moisture contents and growth habits of the species being estimated (e.g. ltalophila cf. Cymodocea).
ESTIMATING SEAGRASS BIOMASS BY VISLIAL CENSUS
73
The visual technique for estimating seagrass biomass returns adequate precision with greater time efficiency per quadrat than other commonly used sampling zechniques and is much less destructive than harvesting a large n u m b e r of quadratt. I consider that the use o f this technique along permanently marked transects offers substantial benefits in monitoring seagrass over large areas over time, or where the habitat is fragile or environmentally sensitive. The disadvantages of the technique are its inapplicability to subtidal seagrass beds in turbid waters and its inability to provide a precise estimate of below-ground biomass. ACKNOWLEDGEMENTS I gratefully acknowledge Joe Miller (Pasture Management Branch, Queensland Department of Primary Industries) for demonstrating the tecl"nique, and Anto Wilson, R o b Coles and Darren Dennis for their assistance in the field. I thank Helene Marsh and two a n o n y m o u s referees for their comments on this manuscript, and the Northern Fisheries Centre, Queensland Department of Primary Industries, Cairns for providing logistical support.
REFERENCES Aioi, K., 1980. Seasonal changes m the standing crop ofeelgrass (Zostera marina L. ) in Odawa Bay, Central Japan, Aquat. Bot., 8: 343-354. Andrew, N.L. and Mapstone, B.D., 1987. Sampling and the description of spatial pattern in marine ecology.Oceanogr. Mar. Biol. Annu. Rev., 25: 39-90. Boon, P.I., 1986. Nitrogen pools in seagrass beds ofCymodocea serrulata and Zostera capricorni of Moreton Bay, Australia. Aquat. Bol., 25: 1-19. Coles, R.G., Lee Long, W.J., Squire, B.A., Squire, L.C. and Bibby, J.M., 1987. Distribution of seagrasses and associated commercial penaeid prawns in north-eastern Queensland waters. Aust. J. Mar. Freshwater Res., 38: 103-120. Downing, J.A. and Anderson, M.R., 1985. Estimating the standing biomass of aquatic macrophyles. J. Fish Aquat. Sci., 42: 1860-1869. Haydock, K.P. and Shaw, N,H., 1975. The comparative yield method ofestimatin~ dry matter yield of pasture. Aust. J. Exp. Agric. Anim. Husb., 15: 663-670. Orth, R.J. and Moore, K.A., 1983. Seasonal and year-to-year variations in the growth of Zostera marina L. (celgrass) in the lower Chesapeake Bay. Aquat. Bot., 24: 335-341. Thresher, R.E. and Gunn, J.S., 1986. Comparative analysis ofvisual census techniques for highly mobile, reef-associated piscivo, zs (Carangidae). Environ. Biol. Fish., 17: 93-116. Tothill, J.C., Hargreaves, J.N.G. and Jones, R.M., 1978. Botonal - A comprehensive sampling and computing procedure for estimating pasture yield and composition. I. Field sampling. Tropical Agronomy Technical Memorandum Number 8.