INTRODUCTION
Community-level patterns in plant species diversity, composition and functional-trait variation are often linked to differences in resource availability along environmental gradients (McGill et al. Reference MCGILL, ENQUIST, WEIHER and WESTOBY2006), which are often defined with respect to the resources required for plant growth and survival (soil mineral nutrients, water and light). Gradients defined with respect to one resource may nonetheless also vary with respect to other resources. For example, forests growing on soils of different texture or along elevation gradients often have different disturbance regimes, creating covariation between these environmental factors and light (Matson & Boone Reference MATSON and BOONE1984, Myster & Fernandez Reference MYSTER and FERNANDEZ1995, Ohkubo Reference OHKUBO2007a, Reference OHKUBO2007b; Richards Reference RICHARDS1996, Wells et al. Reference WELLS, STEWART and DUNCAN1998).
Feedbacks involving differential plant growth provide a mechanism by which resources can vary in disparate ways along gradients (Chapin et al. Reference CHAPIN, BLOOM, FIELD and WARING1987, Pearson et al. Reference PEARSON, BURSLEM, GOERIZ and DALLING2003). Denser forest canopies have been observed to grow on more fertile, moister soils, casting greater shade in the understorey (Ashton Reference ASHTON1964, Ashton & Hall Reference ASHTON and HALL1992, Coomes & Grubb Reference COOMES and GRUBB1996). The degree of shading results from several interacting characteristics of the overstorey, including the biomass of leaves and their sizes, inclinations and longevities and the branching structure of tree crowns (Coomes & Grubb Reference COOMES and GRUBB2000). Variation in light environments between canopy gaps of different sizes and the forest understorey is well-documented (Chazdon & Fetcher Reference CHAZDON and FETCHER1984, Dalling et al. Reference DALLING, HUBBELL and SILVERA1998, Denslow et al. Reference DENSLOW, ELLISON and SANFORD1998, Hartshorn Reference HARTSHORN1980, Whitmore Reference WHITMORE, Tomlinson and Zimmerman1978), and there is also widespread evidence of finer-scale spatial variation in understorey light among microhabitats within closed-canopy forest (Canham et al. Reference CANHAM, FINZI, PACALA and BURBANK1994, Coomes et al. Reference COOMES, ALLEN, BENTLEY, BURROWS, CANHAM, FAGAN, FORSYTH, GAXIOLA-ALCANTAR, PARFITT, RUSCOE, WARDLE, WILSON and WRIGHT2005, Montgomery & Chazdon Reference MONTGOMERY and CHAZDON2001, Poorter & Arets Reference POORTER and ARETS2003, Whitmore & Wong Reference WHITMORE and WONG1959, Yoda Reference YODA1974). What remains poorly understood is how understorey light regimes vary between neighbouring closed-canopy forest types that experience the same rainfall, but grow on different soil types.
Quantifying soil-related differences in understorey irradiance is important for understanding the growth and competitive interactions of juvenile trees because even small differences in photosynthetic photon flux density (PPFD) and the frequency and intensity of sunflecks, short bursts of direct photosynthetically active radiation (PAR), affect carbon gain (Chazdon Reference CHAZDON1986, Koizumi & Oshima Reference KOIZUMI and OSHIMA1993, Leakey et al. Reference LEAKEY, PRESS and SCHOLES2003, Reference LEAKEY, SCHOLES and PRESS2005; Zipperlen & Press Reference ZIPPERLEN and PRESS1997). Furthermore, light-defined niches of tree species can be finely partitioned (Barker et al. Reference BARKER, PRESS and BROWN1997), even among shade-tolerant tree species (Kobe Reference KOBE1999, Montgomery & Chazdon Reference MONTGOMERY and CHAZDON2002, Poorter & Arets Reference POORTER and ARETS2003). Recent studies have found a greater diversity of light-defined niches in forests on more-fertile, moist soil, compared with adjacent soil types that were more nutrient-limited and well-drained (Coomes et al. Reference COOMES, KUNSTLER, CANHAM and WRIGHT2009).
We used an edaphic gradient underlying mixed dipterocarp forest in Malaysian Borneo to investigate covariation between the availabilities of understorey light and soil resources. This region is well-known for having a high turnover of tree species between soil-defined forest types (Ashton Reference ASHTON1964, Baillie et al. Reference BAILLIE, ASHTON, COURT, ANDERSON, FITZPATRICK and TINSLEY1987, Brunig Reference BRUNIG1974) due to specialization of species on different soils (Davies et al. Reference DAVIES, TAN, LAFRANKIE, POTTS, Roubik, Sakai and Hamid2005, Paoli et al. Reference PAOLI, CURRAN and ZAK2006, Potts et al. Reference POTTS, ASHTON, KAUFMAN and PLOTKIN2002). We tested the hypothesis that regimes of PAR differ significantly in the understorey of rain forest growing on four soil types along an edaphic gradient in Borneo. Soils along this gradient range from coarse loams that are sandstone-derived, nutrient-depleted and well-drained, to clays that are shale-derived, less nutrient-depleted and less well-drained (Baillie et al. Reference BAILLIE, ASHTON, CHIN, DAVIES, PALMIOTTO, RUSSO and TAN2006, Russo et al. Reference RUSSO, CANNON, ELOWSKY, TAN and DAVIES2010, Tan et al. Reference TAN, YAMAKURA, TANI, PALMIOTTO, MAMIT, PIN, DAVIES, ASHTON and BAILLIE2009), but the forests share the same rainfall regime, as they are located at one site. Variation in light regimes between and within forest types was characterized in terms of total daily PPFD and the frequency, duration and intensity of sunflecks and in terms of their seasonal variation between the non-monsoon season and the generally cloudier, early monsoon season.
METHODS
Study site
This research was conducted in Lambir Hills National Park, Sarawak, Malaysia (4°11′N, 114°01′E). The Park encompasses 6800 ha of lowland mixed dipterocarp forest with the highest tree species richness recorded in the Palaeotropics (Ashton & Hall Reference ASHTON and HALL1992, Lee et al. Reference LEE, DAVIES, LAFRANKIE, TAN, YAMAKURA, ITOH, OHKUBO and ASHTON2002). Rainfall is c. 3000 mm y−1, with all months averaging > 100 mm (Watson Reference WATSON1985). Despite the lack of a well-defined dry season, there is seasonal variation in rainfall (Palmiotto Reference PALMIOTTO1998): the monsoon season, which brings greater monthly average rainfall, occurs approximately from November to February, with other times of year receiving less rainfall. In 1991, a 52-ha plot (hereafter, Lambir) was established in the Park following standardized methods (Condit Reference CONDIT1998) and pegged with permanent markers at every 20-m grid corner (Figure 1). Floristic composition, stand structure, soils, soil nutrient and water contents, and geomorphology of Lambir are described elsewhere (Baillie et al. Reference BAILLIE, ASHTON, CHIN, DAVIES, PALMIOTTO, RUSSO and TAN2006, Lee et al. Reference LEE, DAVIES, LAFRANKIE, TAN, YAMAKURA, ITOH, OHKUBO and ASHTON2002, Russo et al. Reference RUSSO, CANNON, ELOWSKY, TAN and DAVIES2010, Tan et al. Reference TAN, YAMAKURA, TANI, PALMIOTTO, MAMIT, PIN, DAVIES, ASHTON and BAILLIE2009).
Figure 1. Map of the 52- ha forest dynamics plot at Lambir Hills National Park, Malaysian Borneo, showing soil types, elevation contours (14 m) and sampling locations (filled circles).
The high species richness of this region is at least partly attributable to high beta-diversity, arising from substantial turnover of tree species composition between soil-defined habitats (Ashton Reference ASHTON1964, Baillie et al. Reference BAILLIE, ASHTON, COURT, ANDERSON, FITZPATRICK and TINSLEY1987, Potts et al. Reference POTTS, ASHTON, KAUFMAN and PLOTKIN2002). Soils range from coarse loams that are sandstone-derived, leached, nutrient-depleted and well-drained, with substantial raw humus, to clays that are shale-derived, less nutrient-depleted and less well-drained, with little raw humus. The distributions of most tree species in Lambir are strongly biased with respect to four soil habitats defined based on variation in nutrients (total C, N and P, and exchangeable K, Ca and Mg) and elevation at a 20 × 20-m scale (Davies et al. Reference DAVIES, TAN, LAFRANKIE, POTTS, Roubik, Sakai and Hamid2005). Ranked in ascending order of nutrient content and descending order of particle size, they are: sandy loam, loam, fine loam and clay (Table 1). In addition to variation in mineral nutrients, soil volumetric water content at 0–10 cm depth is significantly higher on the clay, relative to sandy loam, at all times of year (Russo et al. Reference RUSSO, CANNON, ELOWSKY, TAN and DAVIES2010). Furthermore, elevation and slope vary significantly among soil types: the sandy loam and clay soils occur at the highest and lowest elevations, respectively, and the loam and fine loam soils occur at intermediate elevations in Lambir (Baillie et al. Reference BAILLIE, ASHTON, CHIN, DAVIES, PALMIOTTO, RUSSO and TAN2006). The loam and fine loam soils are located on steep slopes, where there are more frequent landslips (Ohkubo Reference OHKUBO2007a, Reference OHKUBO2007b).
Table 1. Mean total nitrogen (%), total phosphorus (mg kg−1), pH, exchangeable magnesium (cmolc kg−1) and calcium (cmolc kg−1), and elevation (m) for the four soil types in the Lambir plot (Davies et al. Reference DAVIES, TAN, LAFRANKIE, POTTS, Roubik, Sakai and Hamid2005). Significant differences among soil types are indicated by different lowercase letters, with standard errors following the means.
Measurement of understorey light and slope
Photosynthetic photon flux density (PPFD) was measured in the understorey of forest on each of the four soil types using photosynthetically active radiation (PAR) sensors (LI-190SA quantum sensor, LiCor Biosciences, Lincoln, Nebraska, USA) from 8 November–7 December 2007 and 28 May–21 June 2008. The first sampling period represents the early monsoon season, whereas the second sampling period (non-monsoon season) encompasses a period of relatively less rainfall. Sensors were arrayed in groups of three at randomly selected 20 × 20-m grid points (locations) on each soil type. Each sensor was placed 5 m from the grid point, in one of the four directions parallel to the plot axes. Sensors were mounted on tripods and levelled at a height of 1 m. Sensors were kept at each location for 72 h then moved to a new location. In 2007, seven locations each were sampled on the loam and fine loam soils, and 14 locations each were sampled on the sandy loam and clay soils, for a total of 189 sensor-days; in 2008, six and 12 locations, respectively, were sampled, for a total of 162 sensor-days. This placement of sensors was designed to sample spatial variation in light within soil types at both coarse scales (variation among locations) and fine scales (variation among sensors within a location).
PPFD from each sensor was recorded using battery-powered data-loggers (CR-800 and CR-1000; Campbell Scientific, Logan, Utah, USA). In 2007, data were scanned every 5 s, with average, maximum and total PPFD recorded every 15 min and 24 h. In 2008, data were scanned and recorded every second, with average, maximum and total PPFD recorded every 15 min and 24 h. Two PAR sensors were mounted in a tower < 1 km from the plot, over the forest canopy and completely exposed to the sky, measuring the maximum possible PPFD given cloud-cover. Data from the canopy and understorey sensors were recorded at the same frequency.
The slope in degrees was calculated using standard methods (Condit Reference CONDIT1998) for each location, based on elevation data from the Lambir plot. Four slopes were calculated from the four corners of each 20 × 20-m quadrat surrounding each location, and the average was used.
Statistical analysis
Total daily PPFD was calculated in the understorey and over the canopy for each sensor-day based on the 15-min totalized data from 06h00–18h00 in 2007 and 2008. Relative total daily PPFD (hereafter, relative PPFD) in the understorey was calculated as a percentage of the above-canopy PPFD (the average of data from the two sensors) on that day ((understorey PPFD/over-canopy PPFD) × 100). Relative PPFD was analysed using a linear, mixed-effects model with normally distributed errors using the Mixed Procedure in SAS software version 9.2 (SAS Institute Inc., Cary, NC, USA). Slope, soil type and their interaction were treated as fixed effects, and location, sampling date (date) and sensor number (sensor) were treated as nested random effects: location (habitat), date (location) and sensor (location), where the parentheses indicate the grouping variable within which effects were nested. For purposes of variance partitioning, all effects were treated as fixed effects in order to calculate coefficients of determination for each. Relative PPFD was logarithmically transformed prior to analysis to improve normality of residuals. Over-canopy PPFD was used to test for seasonal differences in light availability using a general linear model with normally distributed errors and season (early monsoon or non-monsoon) as a fixed effect.
The frequency of sunflecks in 2007 and 2008 was modelled as a binomial process, mj ~ Binomial(Nj, pj), where Nj is the total number of 15-min sampling intervals in a day from 06h00–18h00, mj is the number of those 15-min intervals in which the maximum PPFD exceeded 20, 50 or 100 μmol m−2 s−1, and pj is the expected probability of a sunfleck occurring in an interval. Thus, a 15-min interval was categorized as having experienced a sunfleck of low, medium or high intensity based on whether the maximum PPFD exceeded 20, 50 or 100 μmol m−2 s−1 in that interval. These thresholds of light intensity were selected based on previous definitions of sunflecks (Chazdon & Fetcher Reference CHAZDON and FETCHER1984, Koizumi & Oshima Reference KOIZUMI and OSHIMA1993). Variation in the expected probability of a sunfleck occurring was analysed using generalized linear mixed-effect models employing a logit link function (McCullach & Nelder Reference MCCULLACH and NELDER1989) using the Glimmix Procedure in SAS software version 9.2 (SAS Institute Inc., Cary, NC, USA). Slope, soil type and their interaction were treated as fixed effects, and location, date and sensor were treated as nested random effects, as in the analyses of relative PPFD. Data from 2007 and 2008 were analysed separately for both relative PPFD and sunfleck probability, because one set of locations that was sampled in 2007 was not sampled in 2008. To identify seasonal differences, statistical tests were also performed for the identical subset of locations that were sampled in both seasons. The probability distribution of the maximum PPFD within a 15-min interval, across all sampling sensors, locations and days within a soil type and season were tested for differences (1) between forests on different soil types in the same season and (2) between seasons in forests on the same soil type. The latter comparisons used the identical set of locations for both seasons. These differences were identified using a series of two-sample Komolgorov–Smirnov tests, which test whether two empirical distributions come from the same probability distribution.
To assess patterns in dynamic radiation at finer temporal scales, PPFD was also scanned and recorded at 1-s intervals in 2008. The number of seconds (duration) during which PPFD fell within eight intensity bins was counted at each sensor location from 06h00–18h00 over 18 d. The independence of sunfleck duration and intensity across soil types was tested using log-linear models on the binned data (Agresti Reference AGRESTI1990). For each sensor location, the per cent of total PPFD contributed by sunflecks of different intensities was also calculated. Differences due to soil type and slope were tested using multivariate analysis of variance (MANOVA). Data were square-root transformed to improve normality. Model selection using Akaike's Information Criterion (AIC) (Burnham & Anderson Reference BURNHAM and ANDERSON2002) was used to select the best-supported model for all analyses.
RESULTS
Seasonal variation in total daily photosynthetic photon flux density
Above the forest canopy, total daily PPFD was significantly higher in the non-monsoon than the monsoon season (F (1,65) = 5.31, P = 0.024). Total daily PPFD averaged (± SE) 35.6 ± 0.16 mol m−2 in the non-monsoon season and 31.2 ± 0.15 mol m−2 in the early monsoon season.
Soil-related variation in total daily photosynthetic photon flux density
The slope of the locations at which PPFD was sampled differed significantly among soil types (F (3,38) = 9.53, P < 0.001). Loam was steeper than all other soil types (all P < 0.005), but no other comparisons differed significantly. For both seasons, slope correlated strongly with total daily PPFD only in the understorey of forest on loam soil (Table 2). However, the model with only the main effect of forest type had the lowest AIC for all data sets, and so the effect of slope and its interaction with forest type were removed (ΔAIC relative to the full model: early monsoon season, all locations, 25.3; early monsoon season, subset of locations sampled in both years, 24.1; non-monsoon season, 17.5). Variation between the forests on the four different soil types in total daily PPFD at 1-m height explained 13% and 11% of the total variation in PPFD in the early monsoon and non-monsoon season, respectively. The PPFD varied significantly among these forests in both the early monsoon season (Figure 2a; all locations: F (3,38) = 3.63, P = 0.021; subset of locations sampled in both seasons: F (3,32) = 3.29, P = 0.033) and non-monsoon seasons (Figure 2b; F (3,32) = 3.10, P = 0.040). Overall, the dominant pattern in both seasons was that forest on the clay soil had the darkest understorey and that forest on loam and fine loam had the brightest understorey, but which comparisons showed statistically significant differences were not always consistent. In the early monsoon season, the forest understorey on clay was significantly darker than on both loam and fine loam (based on all locations sampled) and than on loam only (based on locations sampled in both seasons). In the non-monsoon season, the forest understorey on clay was significantly darker than on both sandy loam and loam. In the early monsoon season, the forest understorey on loam also had significantly greater PPFD than did the sandy loam, based on locations sampled in both seasons. The disparate results between datasets based on all locations sampled were largely due to one location on fine loam near a large gap from a landslip that was sampled in the early monsoon, but not the non-monsoon, season.
Table 2. Correlations between slope and total daily photosynthetic photon flux density in the understorey of four forest types in Borneo underlain by sandy loam, loam, fine loam and clay in the early monsoon and non-monsoon seasons.
Figure 2. Average total daily photosynthetic photon flux density (PPFD) in the understorey of Bornean rain forest underlain by four soil types. The inset shows comparisons between forest types in the early monsoon season (November 2007) based on all sampled locations (189 sensor-days at 7 (loam and fine loam) and 14 (sandy loam and clay) locations per forest type), and the main chart shows comparisons based on the subset of those locations that were also sampled in the monsoon season (a). The main chart shows comparisons between forest types in the non-monsoon season (May 2008; 162 sensor-days at 6 (loam and fine loam) and 12 (sandy loam and clay) locations per forest type) (b). Error bars are ± 1 SE. Different letters over bars indicate significant differences between soil types, based on relative PPFD.
Within forest types, course- and fine-scale spatial effects accounted for more of the total variation in PPFD than did temporal effects. Variation between locations within a forest type (coarse scale) and between sensors at a location (fine scale) accounted for 45% and 37% in the early monsoon season (all locations) and for 41% and 41% in the non-monsoon season, respectively, of the total variation in PPFD. In contrast, variation in PPFD among days at a location accounted for only 3% and 4% of the total variation in PPFD in the early monsoon (all locations) and non-monsoon seasons, respectively.
Soil-related variation in frequency and intensity of sunflecks
Forests types differed strongly in the frequency of understorey sunflecks of different intensities in the early monsoon (Figure 3a) and non-monsoon (Figure 3b) seasons, based on data recorded every 15 min. In all within-season pairwise comparisons of forests on different soils, the probability distributions of maximum PPFD for each soil type were significantly different from each other (all P < 0.001). Forests on the same soil type also differed in their probability distributions of maximum PPFD across seasons (all P < 0. 001).
Figure 3. Distribution of the intensity of sunflecks in the understorey of Bornean rain forest underlain by four soil types. The per cent of daytime 15-min intervals in which the maximum photosynthetic photon flux density (PPFD) fell into each light intensity class is shown for the understories of sandy loam, clay, loam and fine loam soils, across all sampling dates, locations and sensors within soil type and season. The inset shows comparisons between forest types in the early monsoon season (November 2007) based on all sampled locations (189 sensor-days at 7 (loam and fine loam) and 14 (sandy loam and clay) locations per forest type), and the main chart shows comparisons based on the subset of those locations that were also sampled in the monsoon season (a). The main chart shows comparisons between forest types in the non-monsoon season (May 2008; 162 sensor-days at 6 (loam and fine loam) and 12 (sandy loam and clay) locations per forest type) (b).
The model with only the main effect of forest type had the lowest AIC for all data sets, and so the effect of slope and its interaction with forest type were removed from models. The frequency of sunflecks with low and medium, but not high, intensities varied significantly among forest types in both the early monsoon and non-monsoon seasons, but differences among forest types in sunfleck frequency and intensity shifted between seasons (Table 3). In the early monsoon season, based on all sampled locations, the forest on clay had the lowest frequency of low- and medium-intensity sunflecks, whereas the forest on sandy loam was intermediate in frequency of low- and medium-intensity sunflecks. The forest on loam had the highest frequency of low-intensity sunflecks, and the forest on fine loam had the highest frequency of medium-intensity sunflecks. Based only on locations sampled in both seasons, however, there were no significant differences among the forests in the early monsoon season. In the non-monsoon season, the forest on clay again had the lowest frequency of low- and medium-intensity sunflecks, but, in contrast to the wetter season, the forest on sandy loam had the highest frequency of low- and medium-intensity sunflecks. The forests on loam and fine loam were intermediate in the frequency of low- and medium-intensity sunflecks.
Table 3. Significant differences within the understorey of Bornean rain forest growing on four soil types in the frequency of sunflecks of low, medium and high intensity (> 20, > 50 and > 100 μmol m−2 s−1, respectively) in two seasons. Sunflecks were defined based on the maximum photosynthetic photon flux density observed in 15-min intervals. Test statistics are shown for the main effect of forest type, with numerator and denominator degrees of freedom (ndf and ddf, respectively). Statistically significant post hoc comparisons between forest types, tested after finding a statistically significant main effect of forest type, were assessed at P < 0.05. Abbreviations for forest types: SL = forest on sandy loam; L = forest on loam; FL = forest on fine loam; C = forest on clay.
Based on PPFD recorded every second during the monsoon season, the duration and intensity of sunflecks differed significantly across forest types. The log-linear model with all effects and interactions had the lowest AIC (448), and a likelihood ratio test strongly favoured inclusion of the three-way interaction (P < 0.001), indicating that sunfleck intensities and durations were not independent across soil types. Both long-duration and high-intensity sunflecks were rare on all soil types. The understorey of forest on sandy loam and loam had higher frequencies of moderate to high-intensity sunflecks, compared with the fine loam and clay soils, where PPFD was more concentrated into short-duration sunflecks of low-intensity. These patterns in sunfleck intensity were confirmed in analyses of the per cent of total PPFD received in 3 d at locations on each soil type (Figure 4). The effect of forest type on the multivariate variance in PPFD was significant (Pillai's P = 0.420, df = 3, P = 0.006). Univariate tests within each sunfleck PPFD category revealed that more of the total PPFD in the understorey of fine loam and clay was contributed by the lowest intensity sunflecks, compared with sandy loam and loam. In contrast, the per cent of the total PPFD contributed by moderate- to high-intensity sunflecks was higher in the sandy loam and loam understorey, compared with fine loam and clay. However, forest types differed in the relationship between slope and the per cent of total PPFD contributed by sunflecks of differing intensities, and this effect was strongest for the forest type on loam, which had the steepest slopes (forest type–slope interaction: Pillai's P = 0.411, df = 3, P = 0.008; Figure 5). Univariate tests within each sunfleck category demonstrated that for the understorey on sandy loam, clay, and fine loam, slope had no effect on the per cent of total PPFD contributed by sunflecks of any intensity, except that on fine loam, steeper locations had a lower per cent contributed by high intensity sunflecks (100–500 μmol m−2 s−1; Figure 5d). In contrast, on loam, locations with steeper slopes had a greater per cent of total PPFD contributed by high-intensity sunflecks (25–50 and 50–100 μmol m−2 s−1; Figure 5b and 5c), but had a lower per cent contributed by the lowest intensity sunflecks (0–5 μmol m−2 s−1; Figure 5a).
Figure 4. The per cent of total photosynthetic photon flux density (PPFD) over 3 d contributed by sunflecks of different PPFD in the understorey of Bornean rain forest underlain by four soil types in the non-monsoon season (May 2008). Points are means across sample locations, with standard error bars, grouped by sunfleck PPFD category. Different letters within each category indicate significant differences among soil types within that category, based on univariate tests after a significant overall multivariate analysis of variance. For categories lacking letters, there were no significant differences among soil types. Symbols: circle, sandy loam; diamond, loam; square, fine loam; triangle, clay.
Figure 5. Variation between slope and the per cent of total photosynthetic photon flux density (PPFD) over 3 d contributed by sunflecks of different PPFD in the understorey of Bornean rain forest underlain by four soil types in the non-monsoon season (May 2008). Only sunfleck PPFD categories with a significant interaction between forest type and slope are shown: 0–5 umol m−2 s−1 (a), 25–50 umol m−2 s−1 (b), 50–100 umol m−2 s−1 (c), and 100–500 umol m−2 s−1 (d). Points represent sample locations, with the least-squares regression lines for forest types having a non-zero slope of the relationship between per cent of total PPFD and slope.
DISCUSSION
The light regimes that juvenile trees experience in the understorey of closed-canopy forest have a profound influence on above-ground competitive interactions, life histories and allometries of trees as they vie for a position in the canopy (Pickett & White Reference PICKETT and WHITE1987), with consequences for the maintenance of species diversity (Ricklefs Reference RICKLEFS1977). Variation in light regimes is, however, both a cause and consequence of above-ground vegetation, the structure of which is influenced by other resources, such as soil nutrients and water availability, and disturbance. We found evidence of such feedbacks involving above- and below-ground resources: there was significant variation in both total daily PPFD and sunfleck frequency and intensity in the understorey of forest growing on four soil types along an edaphic gradient in Borneo. Although the number of sample locations on each forest type was limited, the results we obtained using PAR sensors were consistent with a study of two of these forest types using 98 hemispherical photographs, which found saplings in the understorey on sandy loam to experience 6% greater canopy openness compared with those on clay soil (Russo et al. Reference RUSSO, CANNON, ELOWSKY, TAN and DAVIES2010). Rank differences in understorey light regimes between soil types were consistent between seasons, although above-canopy PAR was lower in the monsoon, relative to the non-monsoon, season. In addition to this soil-related variation, we also found understorey light regimes to vary strongly at finer spatial scales within a forest type and to vary with the slope of the land surface for two forest types.
Causes and consequences of soil-related variation in light regimes
In our study, forests on the different soil types varied in several edaphic factors, including soil texture, soil nutrient and water availability, elevation, slope, and aspect. The independent effects of all of these factors cannot be distinguished here, but we nevertheless found significant soil-related effects on understorey insolation among these forest types. Our results are consistent with studies from tropical and temperate forests that also found understorey light availability to decrease with increasing availability of soil resources (Coomes & Grubb Reference COOMES and GRUBB2000).We did not, however, find a simple pattern of monotonic decline, as forests on the intermediate soil types had greater light availability than those at either end of the gradient. This was always true for the forest on the loam soil, but only true for the fine loam soil in the early monsoon season, largely due to one location on fine loam near a large gap from a landslip that was sampled in the early monsoon, but not the non-monsoon, season.
The within and between forest-type variation in light regimes at Lambir may have several origins. First, the four forest types examined here occur within a 52-ha area, and so do not result from variation in rainfall. Nevertheless, due to differences in soil texture, elevation and slope, the same rainfall produces significant variation in water availability in each forest type (Russo et al. Reference RUSSO, CANNON, ELOWSKY, TAN and DAVIES2010). Despite constant rainfall, our results comparing irradiance in forests on sandy loam and clay are still consistent with studies along rainfall gradients, which generally have found leaf area index to decrease in parallel with water availability, accompanied by an increase in light penetration (Coomes & Grubb Reference COOMES and GRUBB2000). Second, the variation in light transmission among the forests along this edaphic gradient is likely to be related to the divergent leaf and crown traits of the tree species comprising the different assemblages associated with each soil type. Average leaf size of canopy trees is larger on the clay and fine loam than on the sandy loam or loam soil types at Lambir (Ashton Reference ASHTON1964, Ashton & Hall Reference ASHTON and HALL1992), which would reduce PAR transmission to the understorey. Variation among the different forest types at Lambir in the sizes of canopy leaves, which are better lit, is most likely a consequence of differences in soil resource supply, since leaves with larger specific leaf area are generally found on soils with greater nutrient and water availability (Wright et al. Reference WRIGHT, WESTOBY and REICH2002; S. E. Russo, unpubl. data). Furthermore, the decreased penetration of light on clay occurs in opposition to the generally shorter leaf life spans of the clay specialists (S. E. Russo, unpubl. data), in parallel with forests on soil gradients in the temperate zone (Coomes & Grubb Reference COOMES and GRUBB2000). Third, the four forest types vary not only in below-ground resources, but also in slope. The loam soil and parts of the fine loam soil comprise steep slopes with greater occurrence and larger sizes of landslips than do the clay, low-lying parts of fine loam, and sandy loam soils (Ohkubo Reference OHKUBO2007a, Reference OHKUBO2007b; Yamakura et al. Reference YAMAKURA, KANZAKI, ITOH, OHKUBO, OGINO, CHAI, LEE and ASHTON1996). Steep slopes would result in larger, more frequent canopy openings and could explain why the decline in understorey irradiance was not monotonic with increases in soil resource availability. In addition, forest types differed in the relative impact of slope versus other edaphic factors, such as resource availability, on understorey insolation. Because they differed the least in slope, comparisons between forests on clay, fine loam and sandy loam are likely to better reflect the effect of soil texture differences than of slope on understorey irradiance. In contrast, the steep slopes of loam likely explain why the understorey on this forest type had the highest insolation, since only in this forest type did slope correlate strongly with total daily PPFD and sunfleck frequencies and intensities.
Although we found significant soil-related differences in understorey irradiance among forest types, there was also substantial spatial variation within forest types. Tree species composition shifts dramatically among forests on these soil types (Davies et al. Reference DAVIES, TAN, LAFRANKIE, POTTS, Roubik, Sakai and Hamid2005), but the spatial distribution of species, even for soil specialists within forest types, also varies considerably. Spatial turnover in species composition within forest types likely contributes to finer-scale spatial variation in light transmission due to interspecific differences in leaf and crown traits (Canham et al. Reference CANHAM, FINZI, PACALA and BURBANK1994, Turner et al. Reference TURNER, LUCAS, BECKER, WONG, YONG, CHOONG and TYREE2000). Furthermore, although most tree species distributions are strongly associated with categorically defined soil habitats (Davies et al. Reference DAVIES, TAN, LAFRANKIE, POTTS, Roubik, Sakai and Hamid2005), the within-soil variability in understorey irradiance suggests that finer-scale continuous variation in soil resources may also influence the species composition and leaf and crown traits that ultimately determine light transmission.
The different light regimes among the forest types along this edaphic gradient may have several consequences for forest structure and dynamics. It is well-established that even small differences in photosynthetic photon flux density (PPFD) and the frequency and intensity of sunflecks affect the growth of seedlings and saplings in the understorey (Chazdon Reference CHAZDON1986, Koizumi & Oshima Reference KOIZUMI and OSHIMA1993, Leakey et al. Reference LEAKEY, PRESS and SCHOLES2003, Reference LEAKEY, SCHOLES and PRESS2005; Zipperlen & Press Reference ZIPPERLEN and PRESS1997). We found that statistically similar total daily PPFD among these Bornean forest types nonetheless resulted from different distributions of sunfleck frequency and intensity, as did Chazdon (Reference CHAZDON1986). Assuming that trees have the physiological capacity to respond to dynamic irradiance, we would expect these differences to translate into different photosynthetic responses, consequences for carbon gain, and growth rates (Chazdon & Pearcy Reference CHAZDON and PEARCY1991, Leakey et al. Reference LEAKEY, SCHOLES and PRESS2005). At Lambir, maximum rates of carbon assimilation are faster on average for tree species that specialize on the sandy loam soil type, compared with clay, which is contrary to expectations based only on soil resource availability (Russo et al. Reference RUSSO, CANNON, ELOWSKY, TAN and DAVIES2010). However, the reduced soil resources on sandy loam soil may ultimately limit carbon gain or shift carbon allocation below-ground or to making longer-lived tissues, since juvenile trees on this soil have the slowest diameter growth rates (Russo et al. Reference RUSSO, DAVIES, KING and TAN2005). The fastest diameter growth rates are achieved by trees on the loam and fine loam soils, followed by those on clay (Russo et al. Reference RUSSO, DAVIES, KING and TAN2005), paralleling the differences in PPFD and dynamic irradiance for these three soil types. These patterns suggest that competitive dynamics among juvenile trees may be dominated more by soil resource availability at the resource-poor end of this edaphic gradient and by understorey irradiance at the other end, as would be predicted by Tilman (Reference TILMAN1988). Faster predicted tree height growth, based on empirically parameterized allometric growth models, of tree species specializing on the clay soil, relative to sandy loam specialists, is consistent with this hypothesis (Heineman et al. Reference HEINEMAN, JENSEN, SHAPLAND, BOGENRIEF, TAN, REBARBER and RUSSO2011).
Differences among forest types in the frequency of sunflecks of varying intensities may also lead to soil-related differences in the range of light-regeneration niches available, resulting in diversification of tree crown and height allometries and allometric growth rates (Coomes et al. Reference COOMES, KUNSTLER, CANHAM and WRIGHT2009, Heineman et al. Reference HEINEMAN, JENSEN, SHAPLAND, BOGENRIEF, TAN, REBARBER and RUSSO2011, Poorter & Arets Reference POORTER and ARETS2003). Ultimately, these shifts in photosynthetic carbon gain and allometric growth responses are expected to sort tree species among soil habitats, resulting in differences in composition among forest types. Consistent with this idea, the frequency of light-demanding, fast-growing tree species is greater on the loam and fine loam soil types (Russo et al. Reference RUSSO, DAVIES, KING and TAN2005), which would further reinforce soil-related patterns of variation in light regimes due to increased light transmission through the canopies of early-successional species (Canham et al. Reference CANHAM, FINZI, PACALA and BURBANK1994, Coomes & Grubb Reference COOMES and GRUBB2000).
Conclusions
Beyond the importance of understanding the causes and consequences of the plant–soil feedbacks generating soil-related variation in understorey light regimes, our study has implications for the methods used to infer factors important in the assembly of and competitive interactions in forest communities. First, our results bring renewed attention to the need to consider multiple resource dimensions when testing hypotheses about environmental factors driving plant community structure and dynamics along gradients (Chapin et al. Reference CHAPIN, BLOOM, FIELD and WARING1987, Pearson et al. Reference PEARSON, BURSLEM, GOERIZ and DALLING2003). The availability of different resources along gradients can vary in opposing directions, generating opposing environmental selection pressures and hampering inferences concerning the relative importance of different abiotic factors in the assembly and functional trait variation of plant communities.
Second, basal area is often used as a metric of neighbourhood competition in forests, with greater basal area implying a shadier environment (Canham et al. Reference CANHAM, PAPAIK, URIARTE, MCWILLIAMS, JENKINS and TWERY2006). At Lambir the forest on sandy loam has greater basal area and stem density compared with the forest on clay (Davies et al. Reference DAVIES, TAN, LAFRANKIE, POTTS, Roubik, Sakai and Hamid2005, Lee et al. Reference LEE, DAVIES, LAFRANKIE, TAN, YAMAKURA, ITOH, OHKUBO and ASHTON2002), yet understorey light availability varies in the opposite direction. These findings suggest that the relationship between standing basal area and light transmission to the understorey is not the same between these two forest types. It is therefore important to account for variation due to soil types when using standing basal area as a measure of competition for light between forest types in which floristic composition (Davies et al. Reference DAVIES, TAN, LAFRANKIE, POTTS, Roubik, Sakai and Hamid2005), leaf functional traits (S. E. Russo, unpubl. data) and canopy leaf structure (Ashton Reference ASHTON1964, Ashton & Hall Reference ASHTON and HALL1992) differ as greatly as they do between the forests on the different soil types at Lambir.
ACKNOWLEDGEMENTS
The authors thank the Sarawak Forest Department and Forest Research Corporation for their kind permission to conduct research in Lambir Hills National Park. The 52-ha Long-Term Ecological Research Project is a collaborative project of the Forest Department of Sarawak, Malaysia, Harvard University, USA (under NSF awards DEB-9107247 and DEB-9629601 to P. S. Ashton), and Osaka City University, Japan (under Monbusho grant 06041094 to T. Yamakura, 08NP0901 to S. Tamura and 09NP0901 to S. Sasaki). Financial support was also provided by the University of Nebraska – Lincoln through a Faculty Seed Grant and a Jane Robertson Layman Award. Thanks also to David Billesbach and Tanvir Shah for assistance with sensors and programming data loggers and to Steve Portnoy for advice concerning statistical analysis.
