Consequences of environmental heterogeneity for the photosynthetic light environment of a tropical forest

• Whole rainforest canopy absorbs 92% of incident light with little seasonality. • Canopy transmittance, reflectance and absorbance were found to be primarily influenced by solar elevation and cloudiness. • Seasonal variation was found in a NDVI surrogate, but not in LAI or in the PPFD budgets. • Canopy reflectance varies in several ways, owing to gaps and the rugose outer canopy. • Interior light environments depend on inhomogeneities of both atmosphere and canopy. We studied the interplay of atmospheric and canopy structure factors driving the canopy light environment (Photosynthetic Photon Flux Density, PPFD) in primary moist tropical forest in the Tapajós National Forest, Brazil. We quantified the temporal and spatial length scales that characterize intact rain forest inhomogeneities, asking: Are seasonal changes in the canopy radiation balance evident at these scales? We sought to describe the components of intensity, duration and spatial variation of within-canopy PPFD in light of these inhomogeneities. Do fluctuating atmospheric conditions, especially the presence of clouds and precipitation, affect the radiative inputs at both the canopy top and the forest floor? We examined the characteristic scales of heterogeneity in the vertical and the horizontal using a two-part approach. For radiation we combined long-term continuous high-frequency measurements of down- and up-welling short-wave and visible wavebands above the canopy with similarly frequent observations from a dense sensor network at forest floor. Vertical variations in canopy structure, obtained with a ground-based LIDAR, similarly combined intensive observations at the sensor network with occasional large-scale transects along the forest floor. Close similarities in both radiation and canopy structure at both scales support the representativeness of local observations of the wider area. A composite broadband measure similar to the Normalized Difference Vegetation Index (denoted cbNDVI) was constructed from above-canopy observations to compare with reports of remotely sensed canopy reflectivity at this site. We estimated the canopy Leaf Area Index (LAI) by combining observations of the variation in understory transmittance by solar elevation with a commonly-used algorithm. We obtained the conventional whole-canopy extinction coefficient, by applying the Beer-Lambert law. Over the course of a year this forest receives 11,795 mol m −2 , only 62% of potential clear-day PPFD – atmospheric transmissivity is reduced by clouds, precipitation, smoke and haze. Very little PAR (≈ 2%) is reflected from the canopy and only 5.7% penetrates to 1 m above the forest floor - overall 92% of PAR is absorbed by the canopy. All radiation balance components closely tracked the dynamics of incoming light, showing little seasonal variation. Understory light observations across the 7.5–28.5 m spanning the understory array sensors showed essentially constant correlation between sensor pairs over time. There was a high degree of local persistence in the understory spatial pattern that varied slowly and directionally with the changing geometry of the sun and canopy structure. Over larger distances (to 1000 m), the patterns of spatial autocorrelation of understory PPFD and outer canopy structure were remarkably similar in shape, both declining rapidly to a more-or-less constant level around 15–20 m, a scale consistent with the dimensions of outer canopy crowns. The vertical pattern of transmission and absorption was estimated by combining understory transmittance with the distribution of canopy surface area obtained from the ground-based LIDAR system. It showed the maximum absorption relatively low in the canopy (8–19 m above ground). Although rather tall (canopy height is at 41.5 m), the extremely elaborate outer surface of the forest suggests the layers highest above ground are of little consequence to the PAR absorption budget. The cbNDVI measure exhibited seasonal variation consistent with other reports (somewhat higher in the wet season) but in contrast to a recent argument that the forest ‘greens up’ during the early dry season. However, except for the variations caused by the angle and intensity of incoming light modulated by atmospheric effects, there was little seasonality in the forest light environment at km67, including: canopy reflectances in several wavebands, all PPFD radiation budget components, the estimated LAI and mid-day PAR extinction coefficient and the length scales of understory PPFD. Canopy transmission was somewhat greater (6.4%) under diffuse skies compared to the least diffuse conditions (6.0%), and slightly more PPFD was absorbed under diffuse (92.7%) versus the least diffuse conditions (92.3%), but was not a significant contribution to the budget. The large diurnal variations in the extinction coefficient seriously affects its utility as a descriptor of canopy radiative properties. We propose an alternative approach for transmitted light: the canopy behaves as: 1.) a constant fraction filter under diffuse conditions, combined with 2.) a variable filter depending on solar elevation for sunfleck conditions. These regimes may be described with simple parameters each having mechanistic relations to canopy structure. In summary, we demonstrate that obtaining the radiation signal at forest floor at high data rate for long periods exploits seasonal sun angle changes to probe canopy structure. Combined with occasional long transect information, temporal and vertical sampling spatially allow improved definition of characteristic scales to describe both the understory light environment and canopy structure.