WETLANDS, Vol. 13, No. 2, Special Issue, June 1993, pp. 115-121 ? 1993, The Society of Wetland Scientists T H E R E L A T I O N S H I P B E T W E E N V A R I A B L E HYDROPERIOD, P R O D U C T I O N A L L O C A T I O N , A N D B E L O W G R O U N D O R G A N I C T U R N O V E R I N F O R E S T E D W E T L A N D S Frank P. Day, Jr. Department of Biological Sciences OM Dominion University Norfolk, VA 23529 J. Patrick Megonigal Department of Botany Duke University Durham, NC 27706 Abstract: Belowground processes m forested wetland ecosystems are exceptionally important, yet most attention seems to focus on surface flooding regimes and other aboveground features of these systems. Field studies in the Dismal Swamp and several manipulative experiments examined belowground dynamics in relation to a flood intensity gradient. Generally, more extensive flooding results in less production allocation belowground. Erroneous conclusions regarding wetland production are reached if aboveground parameters alone are considered. Root decomposition rates are slowest where the duration of soil saturation is the longest. Organic accumulation rates in wetlands are determined by the amount of production of particular biomass types (eg., leaves vs. roots) and the rate at which they decompose. Biomass allocation patterns seem to change in response to a flooding gradient. This represents a major implication for wetland ecosystem functions, as carbon allocation patterns determine the array of litter types that affect decomposition rates and thus nutrient availability. The hydroperiod data from the Dismal Swamp demonstrate the highly variable nature of flooding in forested wetlands, especially during the growing season. The data suggest that it is unwise to rely on hydroperiod as a direct criterion for identifying a jurisdictional wetland. Key Words: allocation, belowground, forested wetland, hydrology, hydroperiod, production INTRODUCTION Hydrology is the dominant controlling influence on ecosystem dynamics in forested wetlands (Conner et al. 1981, Brinson et al. 1981, Brown 1981, Wharton et al. 1982, Brinson et al. 1984, Day et al. 1988). How- ever, surface flooding seems to receive an inordinate amount of attention in spite of the knowledge that the hydroperiod in most forested wetlands is extremely variable and the hydrodynamics of these systems pri- mariIy occur below the soil surface (Day et al. 1988). The overemphasis on aboveground features of forested wetlands also extends to estimates of primary produc- tion and organic turnover, even though belowground contributions often differ dramatically from above- ground contributions (McClaugherty et al. 1982, Vogt et al. 1986, Megonigal and Day 1988, Powell and Day 1991, Megonigal and Day 1992). For instance, roots can represent a highly significant contribution to sys- tem productivity. In the Dismal Swamp, roots con- 115 tribute as much as 60% of the annual increment to soil organic matter (Megonigal and Day 1988). There is a lot of activity below the soil surface in terms of bio- logical processes and hydrological control; consequent- ly, ecosystem-level patterns and relationships cannot be inferred from aboveground data alone. A stress-subsidy gradient has been proposed, which suggests that too little or too much water reduces pri- mary productivity and organic turnover rates (Mitsch and Ewel 1979) (Figure 1 ). Our primary objective in this paper was to review some of our previous work in relation to such a stress-subsidy gradient and par- ticularly to emphasize the significance ofbelowground data in the evaluation of these relationships. We have used results from field studies in the Great Dismal Swamp in Virginia, a greenhouse experiment on red maple (Acer rubrum L.) seedings, and a mesocosm- scale experiment with bald cypress(Taxodium disti- chum (L.) Richard). 116 WETLANDS, Volume 13, No. 2, Special Issue, 1993 13_ 13. Z N o Flooding I Periodic Flooding Continuous Flooding Flooding Figure 1. Hypothetical stress-subsidy relationship that sug- gests too little or too much water inhibits net primary pro- duction in forested wetlands. METHODS Field Studies Study Area. The Great Dismal Swamp is located on the coastal plain in southeastern Virginia and north- eastern North Carolina. The soils are highly acidic (pH 3.2-5.6), relatively low in nutrients, and high in or- ganics. Flooding occurs during the winter and spring months and following periods of abundant rain. Flood depth and frequency vary throughout the swamp. The research was conducted on four sites that have been extensively studied (Dabel and Day 1977, Day 1982, Megonigal and Day 1988). The cypress site is dominated by bald cypress, red maple, and black gum (Nyssa sylvatica var. biflora(Walter) Sargent). The ma- ple-gum site is characterized by water gum (Nyssa aquatica L.), red maple, and black gum. Dominant tree species on the cedar site include Atlantic white cedar (Chamaecyparis thyoides(L.) BSP.), black gum, and red maple. The mixed hardwood community is dominated by laurel oak (Quercus laurifotia Michaux), white oak (Quercus alba L.), and sweet gum (Liquidambar styr- aciflua L.). More detailed site descriptions can be found in Dabel and Day (1977) and Day (1982). Hydrology. Water levels were continuously recorded by Stevens model F water-level recorders installed on a shallow ground-water well on each site (Day et al. 1988). The depth of the wells was 1.3 m. Water levels were recorded from December, 1984 to July, 1986, and the data were digitized daily. A non-recording deep well was also installed to a depth of 3.3 m, and water levels were determined with a chalked tape. Biomass and Production. Aboveground biomass and production were estimated from diameter-mass re- gressions (Day and Dabel 1978) and biomass incre- ments based on diameter growth (Megonigal and Day 1988). On each site, the diameter of all trees in ten 10 x 10 m plots was recorded during the summer of 1975. Leaf and wood biomass were estimated from regres- sion equations of the form log~0 dry mass = A + B log10 diameter. Separate equations were used for leaves, boles, and branches and for hardwoods, cedar, and cypress. Aboveground woody production was deter- mined by increasing the original diameters by annual increments measured with vernier tree bands (Day 1985) and then estimating the increase in biomass. Leaf production was assumed to equal leaf biomass for deciduous species and leaf litter production for ev- ergreen species. Belowground biomass and production were deter- mined by the sequential coring technique (Powell and Day 1991). Ten soil cores were extracted each month from each of the four sites from March 1985 through February 1986. Cores were obtained with a 7-cm di- ameter bucket auger to a depth of 40 cm. The roots were washed in nested sieves, sorted by size category and dead or alive, oven dried at 70 C for 48 hr, and weighed. An estimate of belowground production was obtained by summing significant increases in biomass throughout the year. Even though the aboveground and belowground data were not determined in the same year, we feet the patterns are valid. Decomposition. Root decompos i t ion rates were quantified by a modified litter bag method that in- volved recover3, o f 40-cm-long mesh bags of pre- weighed roots inserted vertically into the soil (Tupacz and Day 1990). Roots used in the bags were collected from the maple-gum site. Most of the roots were 2 to 5 mm in diameter and about 10-20% by volume were less than 2 mm in diameter. All bags were implanted during the first 2 weeks of January 1985. The first nine sets of samples were taken approximately every 4 weeks. The remaining three sets were spaced approximately 7,6, and 8 weeks apart. Five replicates were removed from each site on each sample date. After retrieval, litter bags were washed, opened, and ingrown roots removed. The samples were oven dried and weighed. Rates of decomposition for the entire study period were calculated with a simple linear model (y=mx+b) where the slope (m) represents the instantaneous decay rate in mg g- ' day- ' . Greenhouse Studies Conditions of the experiment included potted first year red maple seedlings exposed to three different flooding regimes in a greenhouse (no flooding, periodic Day & Megonigal, PRODUCTION IN FORESTED WETLANDS 117 flooding, and continuous flooding). Seedlings were transplanted into 15-cm clay pots filled with soil from the maple-gum site. At the start of the 1985 growing season, the potted seedlings were placed into' 33 c m x 33 cm ? 20.5 cm deep, water-tight, plastic boxes (4 pots per box). Flooded plants were inundated to 5 cm above the soil surface throughout the 7 month exper- iment. The periodically flooded plants were alternately drained and reflooded every two weeks. The unflooded plants were watered two to three times a week with constant volumes of water. The experiment was completed over one growing season, at which time the plants were beginning to show a root-bound condition in the pots. After 7 months, 20 plants from each treatment were harvested, and aboveground and belowground biomass were de- termined (Day 1987). Preweighed unconfined bundles of maple-gum roots were also inserted into the pots, and 20 bundles were retrieved at the end of the study to determine mass loss rates (Day et al. 1989). Mesocosm Studies Twenty-five cypress seedlings were planted at regular spacing into two large (8.0 m 2 ? 1.5 m deep) wetland rhizotron cells in the Savannah River Ecology Labo- ratory rhizotron. One cell was continuously flooded, and a seasonal flooding regime was simulated in the other (Day et al. 1989, Megonigal and Day 1992). The reconstructed soil profiles consisted of 100 cm of a sapric Histosol underlain with 10 cm of sand and 10 cm of gravel. A pipe in the bottom of each cell allowed soil-water to drain at a controlled rate; the soil-water pool turned over about every 9 too. Creek water was pumped from a polyethylene storage tank and sprin- kled onto the soil surface to replace drained water. A system of wells, floats, solenoid switches, and pumps prevented the water table from falling below a pre- scribed depth. The prescribed depth was + 20 cm above the soil surface in the continuously flooded cell and varied in the periodically flooded cell according to the following schedule: - 2 0 cm from March to May and October to December, - 5 0 cm from June to Septem- ber, and + 20 cm from January to February. The experiment was conducted over a three-year period (May 1986 to October 1988), and thus, some of the deficiencies of potted plant studies were reme- died. A sample of plants (N = 8 in 1986 and 1988; N = 9 in 1987) was harvested each October prior to leaf abscission, and aboveground biomass was determined. Plants were harvested in a pattern designed to maxi- mize interplant distances among remaining plants. Root biomass was quantified by pit excavations centered on each plant (30 ? 30 cm to 30-cm depth) and, in the final year, complete excavation of the root systems. Root decomposition rates were measured by the same vertical litter bag technique used in the Dismal Swamp (Day et al. 1989). Five bags of cypress roots were re- covered from each treatment after 17 months and per- cent mass loss was determined. Vertical patterns of redox potential and oxygen levels were simultaneously determined with three fused platinum/copper redox probes and two oxygen chambers buried at 20-cm depth intervals. RESULTS AND DISCUSSION Hydrology Large fluctuations in water levels o c c u r in the Dismal Swamp (Figure 2). Very little surface flooding occurs; only the cypress site showed surface flooding during the growing season in this study. On all sites, the du- ration of soil saturation in the root zone is relatively short during the growing season. All but the mixed hardwood site experienced extensive saturation in the root zone throughout the winter months. The water level at the cypress site dropped precipitously in the summer (3.25 m below the soil surface). The ranking of sites based on mean annual water levels in 1985 was cedar> mapIe-gum> cypress> mixed hardwood. Our data demonstrate the highly variable nature of flooding in forested wetlands, especially during the growing sea- son. Biomass and Production Allocation Our data suggest that attempts to fit forested wet- lands to a stress-subsidy curve (Figure l) may be in- accurate if belowground production is not included. Biomass data show no trend aboveground (Figure 3), although there were no continuously flooded sites rep- resenting the wet extreme of the stress-subsidy gradi- ent. The flooded sites in the Dismal Swamp had sig- nificantly less belowground biomass than the unflooded mixed hardwood site (Figure 3), but the lack of pattern aboveground is not altered by adding the belowground data. The production data present a different story. The flooded sites had significantly greater aboveground production, but less belowground, than the unflooded site (Figure 4). Production allocation patterns are dif- ferent under different flooding regimes. Aboveground production data suggest that the unflooded site is the least productive, but the addition of the belowground data demonstrates that the site is actually the most productive. The possible interpretations of above- ground values can be considerably different with the addition of belowground data. In the mesocosm study, the results from the first two growing seasons were similar to the findings of many 118 WETLANDS, Volume 13, No. 2, Special Issue, 1993 20 10 O i ~ F , [ -, --113 - - 2 8 --3C ~11 , ~ %/ -~o , / : -~? I , / - - 8 8 j - , 8 l! - " i ! - 9 o I" / - - 1 0 8 I --110 --121) --130 --11.0 1BEC81. 1985 GROWING SEASON ,,,, l"J t,. ,,,, ~L , i ~ I,' ',~,-'"~. '4 ~',~' V ',__ ~ _ ~ _ b-' i I 1APR85 1ALJG85 1BECS5 %/, ,~ -x , j'\ j " " 1986 GROWING SEASON~ L I 1APR86 3OJULY - - CEDAR SITE . . . . . . HARDWOOD SITE - - - - CYPRESS SITE . . . . MAPLE--GUM SITE Figure 2. Water levels generated from weekly means. Top horizontal line indicates soil surface; lower horizontal line indicates approximate bottom of root zone. Vertical lines approximate limits of growing season. Surface flooding dur- ing growing season is shaded black. The deepest levels re- corded represent the bottom of the wells. (modified from Day et al. 1988) 30 co 'o 20 x c~ 'E c~ 10 co o'J E .? CCI 1.887 Cedar Maple- Cypress Mixed Gum Hardwoods Figure 3. Aboveground and belowground biomass alloca- tions on each of the Dismal Swamp study sites. Sites with different letters are significantly different (ANOVA and Dun- can's Range Test, P-<0.05). N= 10 for aboveground data and N=l l0 for belowground data (106 for mixed hardwood site).Note that the aboveground and belowground scales dif- fer. ,~, ~, .5 g Z .5 1.176 1.o97 1050 v .989 Cedar Maple- Cypress Mixed Gum Hardwoods Figure 4. Aboveground and belowground net primary pro- duction allocations on each of the Dismal Swamp study sites. N = 10 for aboveground data. Sites with different letters are significantly different (ANOVA and Duncan's Range Test, P-< 0.01). The method of estimating root production did not allow statistical testing since those values are the sum of differences between sample data means. However, other techniques used on these same sites demonstrate the same statistically significant trends (Powell and Day 1991). l-year studies of potted plants (Keeley 1979, Sena Gomes and Kozlowski 1980, Norby and Kozlowski 1983, Kozlowski 1984, Peterson and Bazzaz 1984, Donovan et al. 1988). There was significantly less aboveground and belowground biomass in the contin- uously flooded treatment compared to the periodically flooded treatment (Table 1). However, by the end of the third growing season, the abovegrou.nd biomass of the continuously flooded plants was not statistically different from the periodically flooded plants, although belowground production was still significantly greater in the periodically flooded treatment. Total production was about equal in the two treatments by the third growing season. The first year response was possibly due to inhibition by the flooding in the continuously flooded seedlings, and acclimation subsequently took place (Megonigal and Day 1992). A greater proportion of the stems of the first year plants were under water. A conclusion of reduced production in more exten- sively flooded conditions is reached if the study du- ration is short, but different conclusions are reached if the study is of longer duration. Allocation ratios of aboveground and belowground production vary in re- Day & Megonigal, P R O D U C T I O N IN F O R E S T E D W E T L A N D S 119 Table 1. Annual biomass increment (g plant-~ yr ~) for red maple seedlings in the greenhouse and bald cypress seedlings in the mesocosms. Treatments with different letters are significantly different (ANOVA and Duncan's Range Test, P --< 0.05). N = 20 for greenhouse study; N = 8 in mesocosms in 1986 and 1987 and N = 9 in 1988. Aboveground Belowground Total Study Year Treatment Production Production Production Greenhouse 1985 no flooding 8.5 A 7.2 A 15.7 A periodic flooding 6.4 B 3.5 B 9.9 B continuous flooding 4.5 C 2.1 C 6.6 C Mesocosm 1986 periodic flooding 64.7 A 31.7 A 96.4 A continuous flooding 22.0 B 11.4 B 33.4 B 1987 periodic flooding 420.9 A 179.0 A 599.9 A continuous flooding 308.4 B 129.6 B 438.0 B 1988 periodic flooding 718.7 A 739.8 A 1,458.5 A continuous flooding 1,026.8 A 448.8 B 1,475.6 A sponse to hydroper iod and durat ion of exposure to a given hydroperiod. In the greenhouse study, there was a significant gra- dient o f decreasing product ion (aboveground and be- lowground) f rom the unflooded t reatment to the con- t inuously flooded t rea tment (Table 1). The more extensively flooded plants also showed a lower percent allocation of product ion belowground. A study of lon- ger durat ion may have produced different results. The continuously flooded maple seedlings showed signs of compensatory activity and acclimation by producing extensive adventi t ious water roots (Day 1987). Decomposi t ion Root decomposi t ion rates were consistently lower on the sites with the longest durat ion o f soil saturation in the Dismal Swamp and in the more extensively flooded experimental t reatments in the greenhouse and mesocosm studies (Table 2). The flooding extremes in all three studies were significantly different. Inhibi t ion of root decomposi t ion rates was at least partly due to more anaerobic condit ions in the more extensively flooded treatments (Day et al. 1989). Thus, sites or condi t ions that favor the greatest belowground pro- duct ion also show the highest organic tu rnover rates. However , sites or condit ions that result in a higher propor t ion of product ion allocated belowground may be putting more into more recalcitrant tissues (roots). C O N C L U S I O N S Product ion allocation patterns (ratios) seem to vary in response to a flooding gradient. The major impli- cation with regard to ecosystem function and process rates is that carbon allocation patterns de termine the array o f litter types (aboveground versus belowground or leaf versus root) produced, which in turn affects decomposi t ion rates and thus nutrient availability (Abet et al. 1985, Powell and Day 1991). Because o f the Table 2. Decomposition rates of maple-gum roots in the field and greenhouse and cypress roots in the mesocosms. All data are from the top 20 cm of soil. Treatments or sites with different letters are significantly different (ANOVA and Duncan's Range Test, P _< 0.05). N for each study was field = 58-60, greenhouse = 20, mesocosm = 5. Decomposition Period Rate Percent Mass Study (months) Site Treatment (rag g- ~ day- ~) Remaining Dismal Swamp 12 cedar -0 .92 B (field) maple-gum -0 .83 B cypress - 1.15 AB mixed hardwood -1 .36 A Greenhouse 7 no flooding periodic flooding continuous flooding Mesocosm 17 periodic flooding continuous flooding 54A 68 B 70 B 67A 77 B t 20 WETLANDS, Volume 13, No. 2, Special Issue, 1993 variability in production allocation patterns, studies that neglect to quantify the belowground portion of the system are missing a substantial component of the sys- tem's response. Many ecological evaluations of for- ested wetland ecosystems with regard to flooding gra- dients (Figure 1) or other parameters should be viewed with caution if belowground data are absent. Above- ground data alone can generate inaccurate conclusions. The results of the mesocosm study suggest that the stress-subsidy curve for flooding influence on produc- tion in wetlands may need some revision. Total pro- duction levels were about the same after three years in continuously flooded and periodically flooded treat- ments. The effects of the conditions of an experiment should be carefully considered before generalizing the results. Short-term studies of flooding effects on potted plants may yield results that do not represent plant responses to longer durations of flooding. One growing season in a pot represents a limited view of the effects of flooding on plant growth and production. Finally, the hydroperiod data from the Dismal Swamp provide a basis for evaluation of the regulatory significance of hydrology. The highly variable nature of flooding and soil saturation during the year, from year to year, and from one location to another suggests that it is unjustified and impractical to use hydrology as a criterion for identifying jurisdictional wetlands. Delineation methods for jurisdictional wetlands re- quiting use of three criteria (hydrology, soils, vegeta- tion), rather than two of these three criteria, are thus not practical. Soils and vegetation are products of their hydrological regimes and are persistent features of the system. The absence of flooded or saturated soils dur- ing a specific period does not indicate the area is a non- wetland. ACKNOWLEDGMENTS We thank Susan Powell, Jerry Tupacz, and Lyndon Lee for contributing portions of the work upon which this paper is based. Funding for the research has come from a variety of sources but principally the NERP program of the Savannah River Ecology Laboratory and the Division of Wetlands (contract DE-AC09- 76SR00-819 between the U.S. Department of Energy and the University of Georgia), the Oak Ridge Asso- ciated Universities Faculty Research Participation program (contract S-3221), and the National Science Foundation (DEB-7708609, DEB-7708609-A01, and BSR-8405222). LITERATURE CITED Aber, J.D., J.M. Melillo, K.J. Nadelhoffer, C.A. McClaugherty, and J. Pastor. 1985. Fine root turnover in forest ecosystems in re- lation to quantity and form of nitrogen availability: a comparison of two methods. Oecologia 66:317-321. Brinson, M.M., A.E. Lugo, and S. Brown. 1981. Primary produc- tivity, decomposition and consumer activity in freshwater wet- lands. Annual Review of Ecology and Systematics 12:123-161. Brinson, M.M., H.D. Bradshaw, and E.S. Kane. 1984. Nutrient assimilative capacity of an alluvial floodplain swamp. Journal of Apphed Ecology 21: 1041-t057. Brown, S. 1981. A comparison of the structure, primary produc- tivity, and transpiration of cypress ecosystems in Florida. Eco- logical Monographs 51: 403-427. Conner, W.H., J.G. Gosselink, and R.T. Parrondo. 1981. Com- parison of the vegetation of three Louisiana swamp sites with different flooding regimes. American Journal of Botany 68: 320- 331. Dabel, C.V. and F.P. Day. 1977. Structural comparisons of four plant communities in the Great Dismal Swamp, Virginia. Bulletin of the Torrey Botanical Club 104: 352-360. Day, F.P. 1982. Litter decomposition rates in the seasonally flood- ed Great Dismal Swamp. Ecology 63: 670-678. Day, F.P. 1985. Tree growth rates in the periodically flooded Great Dismal Swamp. Castanea 50: 89-95. Day, F.P. 1987. Effects of flooding and nutrient enrichment on biomass allocation in Acer rubrum seedlings. American Journal of Botany 74: 1541-1554. Day, F.P. and C.V. Dabel. 1978. Phytomass budgets for the Great Dismal Swamp ecosystem. Virgmia Journal of Science 29: 220- 224. Day, F.P., J.P. Megonigal, and L.C. Lee, 1989. Cypress root de- composition in experimental wetland mesocosms. Wetlands 9: 263-282. Day, F.P., S.K. West, and E.G. Tupacz. 1988. The influence of ground-water dynamics in a periodically flooded ecosystem, the Great Dismal Swamp. Wetlands 8: l - I 3. Donovan, L.A., K.W. McLeod, K.C. Sherrod, and N.J. Stumpff. 1988. Response of woody swamp seedlings to flooding and in- creased water temperatures. I. Growth, biomass, and survivorship. American Journal of Botany 75:1181-1190. Keeley, J.E. 1979. Population differentiation along a flood fre- quency gradient: physiological adaptations to flooding in Nyssa sylvatica. Ecological Monographs 49: 89-108. Kozlowski, T.T. 1984. Responses of woody plants to flooding, p. 129-162. In T.T. Kozlowski (ed.) Flooding and Plant Growth. Academic Press, New York, NY, USA. McClaugherty, C.A., J.D. Aber, and J.M. Melillo. 1982. The role of fine roots in the organic matter and nitrogen budgets of two forested ecosystems. Ecology 63: 1481-1490. Megonigal, J.P. and F.P. Day. 1988. Organic matter dynamics in four seasonally flooded forest communities of the Dismal Swamp. American Journal of Botany 75:1334-1343. Megonigal, J.P. and F.P. Day. 1992. Effects of flooding on root and shoot production of bald cypress in large experimental enclo- sures. Ecology 73:1182-1193. Mitsch, W.J. and K.C. Ewel. 1979. Comparative biomass and growth of cypress in Florida wetlands. The American Midland Naturalist 101: 417--426. Norby, R.J. and T.T. Kozlowsk~. 1983. Flooding and SO2 stress interaction in Betula papyrtfera and B. mgra seedlings. Forest Science 29: 73%750. Peterson, D.L. and F.A. Bazzaz. 1984. Photosynthetic and growth responses of silver maple (Acer saccharmum L.) to flooding. The American Midland Naturalist 112: 262-272, Powell. S.W. and F.P. Day. 1991. Root production in four com- munities m the Great Dismal Swamp. American Journal of Botany 78: 288-297. Sena Gomes, A.R. and T.T. Kozlowski. 1980. Growth responses and adaptation of Fraxinus pennsylvanica seedlings to flooding. Plant Physiology 66:267-271. Tupacz, E.G. and F.P. Day. 1990. Decomposition of roots in a seasonally flooded swamp ecosystem. Aquatic Botany 37: 199- 214. Vogt, K.A., C.C. Grier, and D.J. Vogt. 1986. Production, turnover, Day & Megoniga l , P R O D U C T I O N IN F O R E S T E D W E T L A N D S 121 and nutrient dynamics of aboveground and belowground detritus of world forests. Advances in Ecological Research 15: 303-377. Wharton, C.H., W.M. Kitchens, E.C. Pendleton, and T.W. Sipe. 1982. The ecology ofbottomland hardwood swamps ofthe south- east: a community profile. U.S. Fish and Wildlife Service, Bio- logical Services Program, Washington, DC, USA. FWS/OBS-81/ 37. Manuscript received 27 February 1992; revision received 10 July 1992; accepted 25 September 1992.