•Foliage clumping at the branch scale is measured for different heights above ground.•Clumping increased with height at humid sites, and was largely absent at dry sites.•At humid sites, competition ...for light appears to drive leaf clumping at canopy tops.•clumping at dry sites arises at larger spatial scales, which suggest water is a main driver.
Foliage clumping is a forest canopy structural feature influencing light interception, the inversion of light interception for estimating leaf area index, and photosynthesis rates. In this study we estimated clumping factor values at the branch level from different heights in broadleaf tree species at four sites in two climatic zones (two sites in dry climate and two sites in humid climate) using laser light. We found that branch level foliage distribution tends to be random at dry sites where water is the main limiting resource, and that foliage aggregation at the top of emergent canopy tops increases with branch height at humid site where competition for light is high. Our results suggest that high competition for light leads to the production of large amounts of leaves grouped together in high irradiance areas of the canopy, even though this significantly reduces the light interception efficiency on a per leaf area basis. A comparison of these results with foliage clumping factor values derived at larger spatial scales for two sites (one dry and one humid) revealed a crossover in the scales at which clumping arises: leaves are essentially clumped at the branch level but much less so at the plot level for the humid site, and leaves are clumped at the plot level but not at the branch level for the dry site. This suggests that the occurrence of foliage clumping may be related to the intensity of resource limitation (light or water in this case) and its availability in 3D space. Additional research is required to confirm the role of environmental conditions in determining foliage clumping across various forest biomes. Further confirmation and understanding of causal relations between abiotic stresses and canopy foliage clumping may lead to significant refinement in forest canopy radiative transfer modelling schemes.
Airborne laser scanning (LiDAR) is used in forest inventories to quantify stand structure with three dimensional point clouds. However, the structure of point clouds depends not only on stand ...structure, but also on the LiDAR instrument, its settings, and the pattern of flight. The resulting variation between and within datasets (particularly variation in pulse density and footprint size) can induce spurious variation in LiDAR metrics such as maximum height (hmax) and mean height of the canopy surface model (Cmean). In this study, we first compare two LiDAR datasets acquired with different parameters, and observe that hmax and Cmean are 56cm and 1.0m higher, respectively, when calculated using the high-density dataset with a small footprint. Then, we present a model that explains the observed bias using probability theory, and allows us to recompute the metrics as if the density of pulses were infinite and the size of the two footprints were equivalent. The model is our first step in developing methods for correcting various LiDAR metrics that are used for area-based prediction of stand structure. Such methods may be particularly useful for monitoring forest growth over time, given that acquisition parameters often change between inventories.
•The structure of a LiDAR point cloud depends on the device and its settings.•So do metrics that describe the extremes of the vertical distribution.•Using probability theory, we model how two such metrics vary with pulse density.•We demonstrate how the bias in estimating maximum height varies with pulse density.•We also quantify the effect of footprint size and plot size on this bias.
With an increasing pressure on forested landscapes, conservation areas may fail to maintain biodiversity if they are not supported by the surrounding managed forest matrix. Worldwide, forests are ...managed by one of two broad approaches—even‐ and uneven‐aged silviculture. In recent decades, there has been rising public pressure against the systematic use of even‐aged silviculture (especially clear‐cutting) because of its perceived negative esthetic and ecological impacts. This led to an increased interest for uneven‐aged silviculture. However, to date, there has been no worldwide ecological comparison of the two approaches, based on multiple indicators. Overall, for the 99 combinations of properties or processes verified (one study may have evaluated more than one property or process), we found nineteen (23) combinations that clearly showed uneven‐aged silviculture improved the evaluated metrics compared to even‐aged silviculture, eleven (16) combinations that showed the opposite, and 60 combinations that were equivocal. Furthermore, many studies were based on a limited study design without either a timescale (44 of the 76) or spatial (54 of the 76) scale consideration. Current views that uneven‐aged silviculture is better suited than even‐aged silviculture for maintaining ecological diversity and processes are not substantiated by our analyses. Our review, by studying a large range of indicators and many different taxonomic groups, also clearly demonstrates that no single approach can be relied on and that both approaches are needed to ensure a greater number of positive impacts. Moreover, the review clearly highlights the importance of maintaining protected areas as some taxonomic groups were found to be negatively affected no matter the management approach used. Finally, our review points to a lack of knowledge for determining the use of even‐ or uneven‐aged silviculture in terms of both their respective proportion in the landscape and their spatial agency.
We reviewed 76 papers worldwide that compared the two approaches regarding their effects on species/structural diversity and ecological processes. Current views that uneven‐aged silviculture is better suited than even‐aged silviculture for maintaining ecological diversity and processes are not substantiated by our analyses.
Terrestrial lidar data are known to be useful for estimating the three‐dimensional (3D) distribution of leaf area in forests. This type of product holds great potential for modelling canopy ...reflectance and light interception to study the links between structure and function. However, little is currently known about its potential and limits in dense forests. Higher leaf area density implies that more laser pulses emitted by the ground‐based instrument are intercepted in lower canopy levels, and the implications of such occlusion effects on radiative transfer simulations are unknown. Occlusion effects can be minimized by increasing the number of locations lidar data is acquired from; how many locations are required for a forest with a given structure? This paper aims to address these knowledge gaps.
We acquired terrestrial lidar data using a very high density of scanning positions (5 m between positions) over four dense forest 60 m × 60 m plots along a structural gradient. Occlusion effects were quantified, and the 3D distribution of leaf area density was mapped using voxels (cubic volumes) for four different scan densities (one original and three downsampled). The voxel arrays were then input into a radiative transfer model to simulate bidirectional reflectance factors and vertical fraction of absorbed radiation.
We found that the summation of leaf area estimates for all voxels within the plot provided leaf area index (LAI) values close to LAI values estimated using traditional methods at each site. Occluded areas occurred mostly at the top of bottom heavy canopies. Radiative transfer simulations suggest that modelling small scale (<1 m) bidirectional reflectance factors (BRF) and light interception requires the highest scan position density used (5 m between scan positions), particularly at bottom heavy sites, and that 10 m between scan positions can be used for plot scale BRF simulations in forests with foliage density and vertical profiles similar to those tested here.
This work establishes some initial guidelines for establishing terrestrial lidar survey protocols for mapping leaf area density in forests. The leaf area density voxel arrays derived are among the most accurate plot‐level 3D characterizations of foliage arrangement produced to date.
•Terrestrial LiDAR data is used to estimate the distribution of tree leaf area density.•A method for improving wood-leaf separation in the LiDAR data is proposed.•The effect of the sampling volume ...size on leaf area density estimation is explored.•Leaf size and branch architecture are shown to influence optimal sampling sizes.•Occlusion within the LiDAR data also influences the optimal sampling size.
Terrestrial LiDAR scanners have been shown to hold great potential for estimating and mapping three dimensional (3-D) leaf area distribution in forested environments. This is made possible by the capacity of LiDAR scanners to record the 3-D position of every laser pulse intercepted by plant material. The laser pulses emitted by a LiDAR scanner can be regarded as light probes whose transmission and interception may be used to derive leaf area density at different spatial scales using the Beer–Lambert law or Warren Wilson's contact frequency method among others. Segmenting the canopy into cubic volumes –or voxels- provides a convenient means to compute light transmission statistics and describe the spatial distribution of foliage area in tree crowns. In this paper, we investigate the optimal voxel dimensions for estimating the spatial distribution of within crown leaf area density. We analyzed LiDAR measurements from two field sites, located in Mali and in California, with trees having different leaf sizes during periods with and without leaves.
We found that there is a range of voxel sizes, which satisfy three important conditions. The first condition is related to clumping and requires voxels small enough to exclude large gaps between crowns and branches. The second condition requires a voxel size large enough for the conditions postulated by the Poisson law to be valid, i.e., a turbid medium with randomly positioned leaves. And, the third condition relates to the appropriate voxel size to pinpoint the location of those volumes within the canopy which were insufficiently sampled by the LiDAR instrument to derive reliable statistics (occlusion effects). Here, we show that these requirements are a function of leaf size, branching structure, and the predominance of occlusion effects. The results presented provide guiding principles for using voxel volumes in the retrieval of leaf area distributions from terrestrial LiDAR measurements.
Malaysia and Indonesia have been affected by deforestation caused in great part by the proliferation of oil palm plantations. To survey this loss of forest, several studies have monitored these ...southeast Asian nations with satellite remote sensing alert systems. The methods used have shown potential for this approach, but they are limited by imagery with coarse spatial resolution, low revisit times, and cloud cover. The objective of this research is to improve near real-time operational deforestation detection by combining three sensors: Sentinel-1, Sentinel-2 and Landsat-8. We used Change Vector Analysis to detect changes between non-affected forest and images under analysis. The results were validated using 166 plots of undisturbed forest and confirmed deforestation events throughout Sabah Malaysian State, and from 70 points from drone pictures in Sumatra, Indonesia. Sentinel-2 and Landsat-8 yielded sufficient results in terms of accuracy (less than 11% of commission and omission error). Sentinel-1 had lower accuracy (14% of commission error and 28% of omission error), probably resulting from geometric distortions and speckle noise. During the high cloud-cover season optical sensors took about twice the time to detect deforestation compared to Sentinel-1 which was not affected by cloud cover. By combining the three sensors, we detected deforestations about 8 days after forest clearing events. Deforestations were only detectable during approximately the first 100 days, before bare soils were often coved by legume crop. Our results indicate that near real-time deforestation detection can reveal most events, but the number of false detections could be improved using a multiple event detection process.
Forest canopy structure has long been known to be a major driver of the processes regulating the exchange of CO2 and water vapour between terrestrial ecosystems and the atmosphere. It is also an ...important driver of terrestrial vegetation dynamics. Information about fine-scale ecosystem structure is needed to better understand and predict how terrestrial ecosystems respond to and affect environmental change. LiDAR remote sensing from ground-based instruments is a promising technology for providing such information, and physically-based models are ideally suited to process the data and derive reliable products. While complex ray tracing algorithms have been developed to help in the interpretation of LiDAR data, none of these tools are currently widely available. In this paper we present the VoxLAD model; a parametric model using computational geometry that allows to compute estimates of leaf area density at the voxel scale on the basis of terrestrial LiDAR data. This modelling framework removes the need to compute the exact point of entry and exit into and out of the voxels for all individual laser pulses, and thus allows for easier usage. The model requires that each point in the LiDAR point cloud should be classified as wood, foliage, or noise. Here we provide the algorithmic details of the model, and demonstrate that the output of the model closely fits the output of a model using more complex ray tracing techniques.
•A parametric model (VoxLAD) for estimating leaf area density (LAD) from ground-based LiDAR is proposed.•The LAD estimates agree well with those obtained from a complex ray-tracing model.•The physically based VoxLAD model will improve accessibility of LiDAR processing methods.
The monitoring of forest ecosystems is significantly affected by the lack of consistent historical data of low-severity (forest partially disturbed) or gradual disturbance (e.g. eastern spruce ...budworm epidemic). The goal of this paper is to explore the use of a subset of Landsat time series and deep learning models to identify both the type and the year of disturbances, including low-severity and gradual disturbances, in the boreal forest of eastern Canada at the pixel level. Remote sensing data such as the spectral information from Landsat time series are the best available option for large scale observations of disturbances that go back decades. Traditional modeling approaches, like LandTrendr, require substantial handcrafted pre-processing to remove noise and to extract temporal features from the image sequences before using them as input to a classical machine-learning model. Deep-learning models can autonomously discern which features are relevant within the coarse temporal and spectral information from the Landsat annual dense time series.
We evaluated the performance of TempCNN and Transformer model in detecting and classifying the type and the year of the forest disturbance using Landsat time series subsequences. Our findings resulted in the generation of four disturbance maps outlining the forest history from 1986 to 2021 within the eastern Canadian boreal forest. Our experimental outcomes demonstrate several significant benefits of employing deep learning models. Firstly, using noisy Landsat time series they achieve comparable accuracy for classifying fire and total harvesting than existing publicly available disturbance maps. Secondly, the use of shorter time series subsequence with deep learning models enables to map adequately different overlapping disturbances occurring in the complete time series. Finally, they increase the number of distinguishable disturbance classes by adding partial harvesting, gradual disturbances, and forest recovery from older events, making them useful approaches for obtaining the first remote sensing-based map for areas affected by the eastern spruce budworm.
•TempCNN, Transformer tested on Landsat time series for disturbance classification.•The type and the year of disturbance events were classified with good accuracy.•Deep learning effective for identifying partial and progressive disturbances.•Landsat subsequences extract overlapping disturbances.