Estimates Of Forest Structure Research Articles

UAV‐derived estimates of forest structure to inform ponderosa pine forest restoration

AbstractRestoring forest ecosystems has become an increasingly high priority for land managers across the American West. Millions of hectares of forest are in need of drastic yet strategic reductions in density (e.g., basal area). Meeting the restoration and management goals requires quantifying metrics of vertical and horizontal forest structure, which has relied upon field‐based measurements, manned airborne or satellite remote sensing datasets. We used unmanned aerial vehicle (UAV) image‐derived Structure‐from‐Motion (SfM) models and high‐resolution multispectral orthoimagery in this study to quantify vertical and horizontal forest structure at both the fine‐ (<4 ha) and mid‐scales (4–400 ha) across a forest density gradient. We then used these forest structure estimates to assess specific objectives of a forest restoration treatment. At the fine‐scale, we found that estimates of individual tree height and canopy diameter were most accurate in low‐density conditions, with accuracies degrading significantly in high‐density conditions. Mid‐scale estimates of canopy cover and forest density followed a similar pattern across the density gradient, demonstrating the effectiveness of UAV image‐derived estimates in low‐ to medium‐density conditions as well as the challenges associated with high‐density conditions. We found that post‐treatment conditions met a majority of the prescription objectives and demonstrate the UAV image application in quantifying changes from a mechanical thinning treatment. We provide a novel approach to forest restoration monitoring using UAV‐derived data, one that considers varying density conditions and spatial scales. Future research should consider a more spatially extensive sampling design, including different restoration treatments, as well as experimenting with different combinations of equipment, flight parameters, and data processing workflows.

Effects of sampling approaches on quantifying urban forest structure

Researchers and practitioners obtain information on urban forest structure mainly through sampling a small percentage of the total population of trees. Nevertheless, people seldom examine the suitability of sampling designs for this task. In this study, we evaluated the performance of simple random sampling (SRS), stratified random sampling (STR), systematic sampling (SS), and generalized random-tessellation stratified sampling (GRTS) for obtaining information on urban forests in Philadelphia, United States and Beijing, China. We determined the spatial locations of all urban trees in the two cities from high-resolution aerial images. We ran simulated surveys on the data set using the four sampling designs with varied sample sizes to estimate structural variables of urban forests, including the total number of trees, species richness, and species abundance. We conducted accuracy assessments by comparing the estimates to the real values. In the end, we located 1,381,903 and 2,295,794 urban trees in Philadelphia and Beijing, respectively. The results of simulated surveys showed that SRS had a more stable performance than other sampling designs when used for estimating the total number of trees. STR and GRTS performed better than SRS and SS in estimating species richness and detecting rare species. For estimating species abundance, SRS again displayed a stable performance among all designs. Also, we found that small sample sizes (≤200) used in existing studies would lead to estimates of urban forest structure with low accuracy. Our findings provide useful guidance to researchers and practitioners on sampling urban forests.

Utility of hyperspectral compared to multispectral remote sensing data in estimating forest biomass and structure variables in Finnish boreal forest

Three-quarters of Finland’s land surface area is filled with forests, which compose a great part of the country’s biomass, carbon pools and carbon sinks. In order to acquire up-to-date information on the forests, optical remote sensing techniques are commonly used. Moreover, in the future hyperspectral satellite missions will start providing data to support the needs of natural resource management practices, such as forestry. It is, however, unclear what would be the additional value from using hyperspectral data compared to multispectral in quantifying forest variables of Finnish boreal forest. In this study, we used the remote sensing data by hyperspectral AISA imager (128 bands, 400–1000 nm, resolution 0.7 m) and Sentinel-2 (10 bands, resolution 10 m) to assess the possible benefits of higher spectral resolution. As reference data, we used a new nationwide forest resource dataset (stand-level data), which has a high potential in further remote sensing applications. In addition, we used a set of independent in situ measurements (plot-level data) for validation. We applied two kernel-based machine learning regression algorithms (Gaussian process and support vector regression) to relate boreal forest variables with the remote sensing data. The variables of interest were mean height, basal area, leaf area index (LAI), stem biomass and main tree species. The regression algorithms were trained with stand-level data and estimations were evaluated with stand- and plot-level holdout sets. The estimation accuracies were examined with absolute and relative root-mean-square errors. Successful variable estimations showed that kernel-based regression algorithms are suitable tools for forest structure estimation. Based on the results, the additional value of hyperspectral remote sensing data in forest variable estimation in Finnish boreal forest is mainly related to variables with species-specific information, such as main tree species and LAI. The more interesting variables for forestry industry, such as mean height, basal area and stem biomass, can also be estimated accurately with more traditional multispectral remote sensing data.

Effects of topography on tropical forest structure depend on climate context

AbstractTopography affects abiotic conditions which can influence the structure, function and dynamics of ecological communities. An increasing number of studies have demonstrated biological consequences of fine‐scale topographic heterogeneity but we have a limited understanding of how these effects depend on the climate context.We merged high‐resolution (1 m2) data on topography and canopy height derived from airborne lidar with ground‐based data from 15 forest plots in Puerto Rico distributed along a precipitation gradient spanningc. 800–3,500 mm/year. Ground‐based data included species composition, estimated above‐ground biomass (AGB), and two key functional traits (wood density and leaf mass per area, LMA) that reflect resource‐use strategies and a trade‐off between hydraulic safety and hydraulic efficiency. We used hierarchical Bayesian models to evaluate how the interaction between topography × climate is related to metrics of forest structure (i.e. canopy height and AGB), as well as taxonomic and functional alpha‐ and beta‐diversity.Fine‐scale topography (characterized with the topographic wetness index, TWI) significantly affected forest structure and the strength (and in some cases direction) of these effects varied across the precipitation gradient. In all plots, canopy height increased with topographic wetness but the effect was much stronger in dry compared to wet forest plots. In dry forest plots, topographically wetter microsites also had higher levels of AGB but in wet forest plots, topographically drier microsites had higher AGB.Fine‐scale topography influenced functional composition but had only weak or non‐significant effects on taxonomic and functional alpha‐ and beta‐diversity. For instance, community‐weighted wood density followed a similar pattern to AGB across plots. We also found a marginally significant association between variation of wood density and topographic heterogeneity that depended on climate context.Synthesis. The effects of fine‐scale topographic heterogeneity on tropical forest structure and composition depend on the climate context. Our study demonstrates how a stronger integration of topographic heterogeneity across precipitation gradients could improve estimates of forest structure and biomass, and may provide insight to the ways that topography might mediate species responses to drought and climate change.

Early Lessons on Combining Lidar and Multi-baseline SAR Measurements for Forest Structure Characterization

The estimation and monitoring of 3D forest structure at large scales strongly rely on the use of remote sensing techniques. Today, two of them are able to provide 3D forest structure estimates: lidar and synthetic aperture radar (SAR) configurations. The differences in wavelength, imaging geometry, and technical implementation make the measurements provided by the two configurations different and, when it comes to the sensitivity to individual 3D forest structure components, complementary. Accordingly, the potential of combining lidar and SAR measurements toward an improved 3D forest structure estimation has been recognised from the very beginning. However, until today there is no established framework for this combination. This paper attempts to review differences, commonalities, and complementarities of lidar and SAR measurements. First, vertical lidar reflectance and SAR reflectivity profiles at different wavelengths are compared in different forest types. Then, current perspectives on their combination for the generation of enhanced structure products are discussed. Two promising frameworks for combining lidar and SAR measurements are reviewed. The first one is a model-based framework where lidar-derived parameters are used to initialize SAR scattering models, and relies on both the validity of the models and on the physical equivalence of the used lidar and SAR parameters. The second one is a structure-based framework based on the ability of lidar and SAR measurements to express physical forest structure by means of appropriate indices. These indices can then be used to establish a link between the two kind of measurements. The review is supported by experimental results achieved using space- and airborne data acquired in recent relevant mission and campaigns.

Prediction of stem diameter and biomass at individual tree crown level with advanced machine learning techniques

Knowledge about the aboveground biomass (AGB) and the diameters at breast height (DBH) distribution can lead to a precise estimation of carbon density and forest structure which can be very important for ecology studies especially for those concerning climate change. In this study, we propose to predict DBH and AGB of individual trees using tree height (H) and crown diameter (CD), and other metrics extracted from airborne laser scanning (ALS) data as input. In the proposed approach, regression methods, such us support vector machine for regression (SVR) and random forests (RF), were used to find a transformation or a transfer function that links the input parameters (H, CD, and other ALS metrics) with the output (DBH and AGB). The developed approach was tested on two datasets collected in southern Norway comprising 3970 and 9467 recorded trees, respectively. The results demonstrate that the developed approach provides better results compared to a state-of-the-art work (based on a linear model with the standard least-squares method) with RMSE equal to 81.4 kg and 92.0 kg, respectively (compared to 94.2 kg and 110.0 kg) for the prediction of AGB, and 5.16 cm and 4.93 cm, respectively (compared to 5.49 cm and 5.30 cm) for DBH.

Estimating Canopy Fuel Attributes from Low-Density LiDAR

Simulations of wildland fire risk are dependent on the accuracy and relevance of spatial data inputs describing drivers of wildland fire, including canopy fuels. Spatial data are freely available at national and regional levels. However, the spatial resolution and accuracy of these types of products often are insufficient for modeling local conditions. Fortunately, active remote sensing techniques can produce accurate, high-resolution estimates of forest structure. Here, low-density LiDAR and field-based data were combined using randomForest k-nearest neighbor imputation (RF-kNN) to estimate canopy bulk density, canopy base height, and stand age across the Boundary Waters Canoe Area in Minnesota, USA. RF-kNN models produced strong relationships between estimated canopy fuel attributes and field-based data for stand age (Adj. R2 = 0.81, RMSE = 10.12 years), crown fuel base height (Adj. R2 = 0.78, RMSE = 1.10 m), live crown base height (Adj. R2 = 0.7, RMSE = 1.60 m), and canopy bulk density (Adj. R2 = 0.48, RMSE = 0.09kg/m3). These results suggest that low-density LiDAR can help estimate canopy fuel attributes in mixed forests, with robust model accuracies and high spatial resolutions compared to currently utilized fire behavior model inputs. Model map outputs provide a cost-efficient alternative for data required to simulate fire behavior and support local management.

Chimera: A Multi-Task Recurrent Convolutional Neural Network for Forest Classification and Structural Estimation

More consistent and current estimates of forest land cover type and forest structural metrics are needed to guide national policies on forest management, carbon sequestration, and ecosystem health. In recent years, the increased availability of high-resolution (<30 m) imagery and advancements in machine learning algorithms have opened up a new opportunity to fuse multiple datasets of varying spatial, spectral, and temporal resolutions. Here, we present a new model, based on a deep learning architecture, that performs both classification and regression concurrently, thereby consolidating what was previously several independent tasks and models into one stream. The model, a multi-task recurrent convolutional neural network that we call the Chimera, integrates varying resolution, freely available aerial and satellite imagery, as well as relevant environmental factors (e.g., climate, terrain) to simultaneously classify five forest cover types (`conifer’, `deciduous’, `mixed’, `dead’, `none’ (non-forest)) and to estimate four continuous forest structure metrics (above ground biomass, quadratic mean diameter, basal area, canopy cover). We demonstrate the performance of our approach by training an ensemble of Chimera models on 9967 georeferenced (true locations) Forest Inventory and Analysis field plots from the USDA Forest Service within California and Nevada. Classification diagnostics for the Chimera ensemble on an independent test set produces an overall average precision, recall, and F1-score of 0.92, 0.92, and 0.92. Class-wise F1-scores were high for `none’ (0.99) and `conifer’ (0.85) cover classes, and moderate for the `mixed’ (0.74) class samples. This demonstrates a strong ability to discriminate locations with and without trees. Regression diagnostics on the test set indicate very high accuracy for ensembled estimates of above ground biomass ( R 2 = 0 . 84 , RMSE = 37 . 28 Mg/ha), quadratic mean diameter ( R 2 = 0 . 81 , RMSE = 3 . 74 inches), basal area ( R 2 = 0 . 87 , RMSE = 25 . 88 ft 2 /ac), and canopy cover ( R 2 = 0 . 89 , RMSE = 8 . 01 percent). Comparative analysis of the Chimera ensemble versus support vector machine and random forest approaches demonstrates increased performance over both methods. Future implementations of the Chimera ensemble on a distributed computing platform could provide continuous, annual estimates of forest structure for other forested landscapes at regional or national scales.

Linking Terrestrial LiDAR Scanner and Conventional Forest Structure Measurements with Multi-Modal Satellite Data

Obtaining information on vertical forest structure requires detailed data acquisition and analysis which is often performed at a plot level. With the growing availability of multi-modal satellite remote sensing (SRS) datasets, their usability towards forest structure estimation is increasing. We assessed the relationship of PlanetScope-, Sentinel-2-, and Landsat-7-derived vegetation indices (VIs), as well as ALOS-2 PALSAR-2- and Sentinel-1-derived backscatter intensities with a terrestrial laser scanner (TLS) and conventionally measured forest structure parameters acquired from 25 field plots in a tropical montane cloud forest in Kafa, Ethiopia. Results showed that canopy gap-related forest structure parameters had their highest correlation (|r| = 0.4 − 0.48) with optical sensor-derived VIs, while vegetation volume-related parameters were mainly correlated with red-edge- and short-wave infrared band-derived VIs (i.e., inverted red-edge chlorophyll index (IRECI), normalized difference moisture index), and synthetic aperture radar (SAR) backscatters (|r| = −0.57 − 0.49). Using stepwise multi-linear regression with the Akaike information criterion as evaluation parameter, we found that the fusion of different SRS-derived variables can improve the estimation of field-measured structural parameters. The combination of Sentinel-2 VIs and SAR backscatters was dominant in most of the predictive models, while IRECI was found to be the most common predictor for field-measured variables. The statistically significant regression models were able to estimate cumulative plant area volume density with an R2 of 0.58 and with the lowest relative root mean square error (RRMSE) value (0.23). Mean gap and number of gaps were also significantly estimated, but with higher RRMSE (R2 = 0.52, RRMSE = 1.4, R2 = 0.68, and RRMSE = 0.58, respectively). The models showed poor performance in predicting tree density and number of tree species (R2 = 0.28, RRMSE = 0.41, and R2 = 0.21, RRMSE = 0.39, respectively). This exploratory study demonstrated that SRS variables are sensitive to retrieve structural differences of tropical forests and have the potential to be used to upscale biodiversity relevant field-based forest structure estimates.

Information Needs of Next-Generation Forest Carbon Models: Opportunities for Remote Sensing Science

Forests are integral to the global carbon cycle, and as a result, the accurate estimation of forest structure, biomass, and carbon are key research priorities for remote sensing science. However, estimating and understanding forest carbon and its spatiotemporal variations requires diverse knowledge from multiple research domains, none of which currently offer a complete understanding of forest carbon dynamics. New large-area forest information products derived from remotely sensed data provide unprecedented spatial and temporal information about our forests, which is information that is currently underutilized in forest carbon models. Our goal in this communication is to articulate the information needs of next-generation forest carbon models in order to enable the remote sensing community to realize the best and most useful application of its science, and perhaps also inspire increased collaboration across these research fields. While remote sensing science currently provides important contributions to large-scale forest carbon models, more coordinated efforts to integrate remotely sensed data into carbon models can aid in alleviating some of the main limitations of these models; namely, low sample sizes and poor spatial representation of field data, incomplete population sampling (i.e., managed forests exclusively), and an inadequate understanding of the processes that influence forest carbon accumulation and fluxes across spatiotemporal scales. By articulating the information needs of next-generation forest carbon models, we hope to bridge the knowledge gap between remote sensing experts and forest carbon modelers, and enable advances in large-area forest carbon modeling that will ultimately improve estimates of carbon stocks and fluxes.

Using LiDAR to develop high-resolution reference models of forest structure and spatial pattern

Successful restoration of degraded forest landscapes requires reference models that adequately capture structural heterogeneity at multiple spatial scales and for specific landforms. Despite this need, managers often lack access to reliable reference information, in large part because field-based methods for assessing variation in forest structure are costly and inherently suffer from limited replication and spatial coverage and, therefore, yield limited insights about the ecological structure of reference forests at landscape scales. LiDAR is a cost-effective alternative that can provide high-resolution characterizations of variation in forest structure among landform types. However, managers and researchers have been reluctant to use LiDAR for characterizing structure because of low confidence in its capacity to approximate actual tree distributions. By calculating bias in LiDAR estimates for a range of tree-height cutoffs, we improved LiDAR’s ability to capture structural variability in terms of individual trees. We assessed bias in the processed LiDAR data by comparing datasets of field-measured and LiDAR-detected trees of various height classes in terms of overall number of trees and estimates of structure and spatial pattern in an important contemporary reference forest, the Sierra de San Pedro Martir National Park, Baja California, Mexico. Agreement between LiDAR- and field-based estimates of tree density, as well as estimates of forest structure and spatial pattern, was maximized by removing trees less than 12 m tall. We applied this height cutoff to LiDAR-detected trees of our study landscape, and asked if forest structure and spatial pattern varied across topographic settings. We found that canyons, shallow northerly, and shallow southerly slopes were structurally similar; each had a greater number of all trees, large trees, and large tree clumps than steep southerly slopes and ridges. Steep northerly slopes supported unique structures, with taller trees than ridges and shorter trees than canyons and shallow southerly slopes. Our results show that characterizations of forest structure based on LiDAR-detected trees are reasonably accurate when the focus is narrowed to the overstory. In addition, our finding of strong variation of forest structure and spatial pattern across topographic settings demonstrates the importance of developing reference models at the landscape scale, and highlights the need for replicated sampling among stands and landforms. Methods developed here should be useful to managers interested in using LiDAR to characterize distributions of medium and large overstory trees, particularly for the development of landscape-scale reference models.

Improved forest height estimation by fusion of simulated GEDI Lidar data and TanDEM-X InSAR data

Interferometric Synthetic Aperture Radar (InSAR) and lidar are increasingly used active remote sensing techniques for forest structure observation. The TanDEM-X (TDX) InSAR mission of German Aerospace Center (DLR) and the upcoming Global Ecosystem Dynamics Investigation (GEDI) of National Aeronautics and Space Administration (NASA) together may provide more accurate estimates of global forest structure and biomass via their synergic use. In this paper, we explored the efficacy of simulated GEDI data in improving height estimates from TDX InSAR data. Our study sites span three major forest types: a temperate forest, a mountainous conifer forest, and a tropical rainforest. The GEDI lidar coverage was simulated for the full nominal two-year mission duration, under both cloud-free and 50%-cloud conditions. We then used these GEDI data to parameterize the Random Volume over Ground (RVoG) model driven by TDX imagery. In particular, we explored the following three strategies for forest structure estimation: 1) TDX data alone; 2) TDX + GEDI-derived digital terrain model (DTM); and 3) TDX + GEDI DTM + GEDI canopy height. We then validated the retrieved forest heights against wall-to-wall airborne lidar measurements. We found relatively large biases at 90 [m] spatial resolution, from 4.2–11.9 [m], and root mean square errors (RMSEs), from 7.9–12.7 [m] when using TDX data alone under constrained RVoG assumptions of a fixed extinction coefficient (σ) and a zero ground-to-volume amplitude ratio (μ = 0). Results improved significantly with the aid of a DTM derived from GEDI data which enabled estimation of spatially-varying σ values (vs. fixed extinction) under a μ = 0 assumption, with biases reduced to 1.7–4.2 [m] and RMSEs to 4.9–8.6 [m] across cloudy and cloud-free cases. The best agreement was achieved in the third strategy by also incorporating information of GEDI-derived canopy height to further enhance the RVoG parameters. The improved model, when still assuming μ = 0, reduced biases to less than or close to 1 m and further reduced RMSEs to 4.0–6.7 [m]. Finally, we used GEDI data to estimate spatially-varying μ in the RVoG model. We found biases of between −0.7–0.9 [m] and RMSEs in the range from 2.6–7.1 [m] over the three sites. Our results suggest that use of GEDI data improves height inversion from TDX, providing heights at more accuracy than can be achieved by TDX alone, and enabling wall-to-wall height estimation at much finer spatial resolution than can be achieved by GEDI alone.

Foreword to the Special Issue on Forest Structure Estimation in Remote Sensing

FORESTS are important ecosystems to all life on Earth. They provide shelter for many flora and fauna, rich resources for anthropogenic development and an important sink in the global carbon cycle. Satellite earth observation provides a critical capacity to observe this ecosystem wall to wall. Especially, synthetic aperture radar (SAR) and Lidar technologies hold large potential to measure forest structure properties such as forest height and biomass and are consequently used in existing observation systems. In addition to the existing observation systems space agencies are implementing or planning dedicated satellite missions capitalizing on these technologies, such as ESAs BIOMASS, NASAs GEDI, and NISAR missions or DLRs Tandem-L mission concept. Despite the sensitivity of SAR and Lidar technologies to forest structure the actual interpretation of the measurements is challenging. Forest structure, environmental effects and system parameters have a confounding effect on the measurements and need to be accounted for in the biophysical parameter retrieval. The impact of these factors is often not well understood and hampers our ability to make optimum use of newly available technologies and their synergies. The most striking development in the last years was in the three-dimensional (3-D) radar data processing. The evolution and refinement of coherent multibaseline (tomographic) processing techniques dramatically increased the quality of the reconstructed 3-D radar reflectivity allowing a deeper insight in the 3-D scattering processes at different polarizations and frequencies and of the opportunities arising from this new measurement technique to monitor and investigate 3-D forest structure. This special issue collect and highlight the latest results, Digital Object Identifier 10.1109/JSTARS.2018.2870266 data processing developments and lessons learned from existing satellite data, airborne data and tower based research campaigns.

Three decades of forest structural dynamics over Canada's forested ecosystems using Landsat time-series and lidar plots

Attributes that describe forest structure, such as height, canopy cover, volume, and biomass, are required to inform forest inventories and monitoring programs. Light Detection and Ranging (lidar) has been successfully demonstrated as a means to derive a suite of forest structure attributes at the plot level; however, these acquisitions are often constrained to limited spatial extents and to a given point in time. Sample based approaches for model development can accommodate the spatial limitations of lidar acquisitions when characterizing large areas. The combination of lidar plot data and time-series satellite imagery is well suited to provide spatially extensive, and temporally dense, information on forest structure and related dynamics over very large areas. In this research, we combine lidar plot-derived information with Landsat pixel-based composites to produce annual forest structure estimates from 1984 to 2016 over 650 million ha of Canada's forest ecosystems using a nearest neighbor imputation approach with a Random Forests-based distance metric. Imputed variables included lidar metrics of height (e.g., mean height, standard deviation of height) and cover, as well as area-based modelled inventory estimates of Lorey's height, basal area, stem volume, and biomass. Models were validated using reserved validation plots, with model R2 ranging from 0.62 to 0.64 for lidar metrics of height and cover, and R2 of 0.67, 0.68, 0.71, and 0.70 for Lorey's height, basal area, volume and biomass, respectively. Unique to this study was the assessment of model extension through time, with model performance for imputing lidar metrics evaluated at the forest stand-level using independent lidar data representing a latitudinal gradient of forest conditions and that was not used in model development. The period evaluated was 2006–2012, with R2 values ranging from 0.36 to 0.66 for height metrics, and 0.47–0.77 for cover metrics. Ultimately, we show how deriving forest structural estimates on an annual basis enables the analysis of both the dynamics and regional trends of undisturbed forest, as well as regenerating stands following stand-replacing disturbances (i.e., fire, harvesting).

Forest Structure Estimation from a UAV-Based Photogrammetric Point Cloud in Managed Temperate Coniferous Forests

Here, we investigated the capabilities of a lightweight unmanned aerial vehicle (UAV) photogrammetric point cloud for estimating forest biophysical properties in managed temperate coniferous forests in Japan, and the importance of spectral information for the estimation. We estimated four biophysical properties: stand volume (V), Lorey’s mean height (HL), mean height (HA), and max height (HM). We developed three independent variable sets, which included a height variable, a spectral variable, and a combined height and spectral variable. The addition of a dominant tree type to the above data sets was also tested. The model including a height variable and dominant tree type was the best for all biophysical property estimations. The root-mean-square errors (RMSEs) for the best model for V, HL, HA, and HM, were 118.30, 1.13, 1.24, and 1.24, respectively. The model including a height variable alone yielded the second highest accuracy. The respective RMSEs were 131.74, 1.21, 1.31, and 1.32. The model including a spectral variable alone yielded much lower estimation accuracy than that including a height variable. Thus, a lightweight UAV photogrammetric point cloud could accurately estimate forest biophysical properties, and a spectral variable was not necessarily required for the estimation. The dominant tree type improved estimation accuracy.

A hyperspectral image can predict tropical tree growth rates in single-species stands.

Remote sensing is increasingly needed to meet the critical demand for estimates of forest structure and composition at landscape to continental scales. Hyperspectral images can detect tree canopy properties, including species identity, leaf chemistry and disease. Tree growth rates are related to these measurable canopy properties but whether growth can be directly predicted from hyperspectral data remains unknown. We used a single hyperspectral image and light detection and ranging-derived elevation to predict growth rates for 20 tropical tree species planted in experimental plots. We asked whether a consistent relationship between spectral data and growth rates exists across all species and which spectral regions, associated with different canopy chemical and structural properties, are important for predicting growth rates. We found that a linear combination of narrowband indices and elevation is correlated with standardized growth rates across all 20 tree species (R2 =53.70%). Although wavelengths from the entire visible-to-shortwave infrared spectrum were involved in our analysis, results point to relatively greater importance of visible and near-infrared regions for relating canopy reflectance to tree growth data. Overall, we demonstrate the potential for hyperspectral data to quantify tree demography over a much larger area than possible with field-based methods in forest inventory plots.

Optimal Altitude, Overlap, and Weather Conditions for Computer Vision UAV Estimates of Forest Structure

Ecological remote sensing is being transformed by three-dimensional (3D), multispectral measurements of forest canopies by unmanned aerial vehicles (UAV) and computer vision structure from motion (SFM) algorithms. Yet applications of this technology have out-paced understanding of the relationship between collection method and data quality. Here, UAV-SFM remote sensing was used to produce 3D multispectral point clouds of Temperate Deciduous forests at different levels of UAV altitude, image overlap, weather, and image processing. Error in canopy height estimates was explained by the alignment of the canopy height model to the digital terrain model (R2 = 0.81) due to differences in lighting and image overlap. Accounting for this, no significant differences were observed in height error at different levels of lighting, altitude, and side overlap. Overall, accurate estimates of canopy height compared to field measurements (R2 = 0.86, RMSE = 3.6 m) and LIDAR (R2 = 0.99, RMSE = 3.0 m) were obtained under optimal conditions of clear lighting and high image overlap (>80%). Variation in point cloud quality appeared related to the behavior of SFM ‘image features’. Future research should consider the role of image features as the fundamental unit of SFM remote sensing, akin to the pixel of optical imaging and the laser pulse of LIDAR.

Comparison of small-footprint discrete return and full waveform airborne lidar data for estimating multiple forest variables

The quantification of forest ecosystems is important for a variety of purposes, including the assessment of wildlife habitat, nutrient cycles, timber yield and fire propagation. This research assesses the estimation of forest structure, composition and deadwood variables from small-footprint airborne lidar data, both discrete return (DR) and full waveform (FW), acquired under leaf-on and leaf-off conditions. The field site, in the New Forest, UK, includes managed plantation and ancient, semi-natural, coniferous and deciduous woodland. Point clouds were rendered from the FW data through Gaussian decomposition. An area-based regression approach (using Akaike Information Criterion analysis) was employed, separately for the DR and FW data, to model 23 field-measured forest variables. A combination of plot-level height, intensity/amplitude and echo-width variables (the latter for FW lidar only) generated from both leaf-on and leaf-off point cloud data were utilised, together with individual tree crown (ITC) metrics from rasterised leaf-on height data. Statistically significant predictive models (p b 0.05) were generated for all 23 forest metrics using both the DR and FW lidar datasets, with R2 values for the best fit models in the range R2= 0.43–0.94 for the DR data and R2= 0.28–0.97 for the FW data (with normalised RMSE values being 18%–66% and 16%–48% respectively). For all but two forest metrics the difference between the NRMSE of the best performing DR and FW models was ≤7%, and there was an even split (11:12) as to which lidar dataset (DR or FW) generated the best model per forest metric. Overall, the DR data performed better at modelling structure variables, whilst the FW data performed better at modelling composition and deadwood variables. Neither showed a clear advantage at modelling variables from a particular vegetation layer (canopy, shrub or ground). Height, intensity/ amplitude, and ITC-derived crown variables were shown to be important inputs across the best performing models (DR or FW), but the additional echo-width variables available from FW point data were relatively unimportant. Of perhaps greater significance to the choice between lidar data type (i.e. DR or FW) in determining the predictive power of the best performing models was the selection of leaf-on and/or leaf-off data. Of the 23 best models, 10 contained both leaf-on and leaf-off lidar variables, whilst 11 contained only leaf-on and two only leaf-off data. We therefore conclude that although FW lidar has greater vertical profile information than DR lidar, the greater complimentary information about the entire forest canopy profile that is available from both leaf-on and leaf-off data is of more benefit to forest inventory, in general, than the selection between DR or FW lidar.

Structural Dynamics of Tropical Moist Forest Gaps.

Gap phase dynamics are the dominant mode of forest turnover in tropical forests. However, gap processes are infrequently studied at the landscape scale. Airborne lidar data offer detailed information on three-dimensional forest structure, providing a means to characterize fine-scale (1 m) processes in tropical forests over large areas. Lidar-based estimates of forest structure (top down) differ from traditional field measurements (bottom up), and necessitate clear-cut definitions unencumbered by the wisdom of a field observer. We offer a new definition of a forest gap that is driven by forest dynamics and consistent with precise ranging measurements from airborne lidar data and tall, multi-layered tropical forest structure. We used 1000 ha of multi-temporal lidar data (2008, 2012) at two sites, the Tapajos National Forest and Ducke Reserve, to study gap dynamics in the Brazilian Amazon. Here, we identified dynamic gaps as contiguous areas of significant growth, that correspond to areas > 10 m2, with height <10 m. Applying the dynamic definition at both sites, we found over twice as much area in gap at Tapajos National Forest (4.8 %) as compared to Ducke Reserve (2.0 %). On average, gaps were smaller at Ducke Reserve and closed slightly more rapidly, with estimated height gains of 1.2 m y-1 versus 1.1 m y-1 at Tapajos. At the Tapajos site, height growth in gap centers was greater than the average height gain in gaps (1.3 m y-1 versus 1.1 m y-1). Rates of height growth between lidar acquisitions reflect the interplay between gap edge mortality, horizontal ingrowth and gap size at the two sites. We estimated that approximately 10 % of gap area closed via horizontal ingrowth at Ducke Reserve as opposed to 6 % at Tapajos National Forest. Height loss (interpreted as repeat damage and/or mortality) and horizontal ingrowth accounted for similar proportions of gap area at Ducke Reserve (13 % and 10 %, respectively). At Tapajos, height loss had a much stronger signal (23 % versus 6 %) within gaps. Both sites demonstrate limited gap contagiousness defined by an increase in the likelihood of mortality in the immediate vicinity (~6 m) of existing gaps.

Airborne lidar-based estimates of tropical forest structure in complex terrain: opportunities and trade-offs for REDD+

BackgroundCarbon stocks and fluxes in tropical forests remain large sources of uncertainty in the global carbon budget. Airborne lidar remote sensing is a powerful tool for estimating aboveground biomass, provided that lidar measurements penetrate dense forest vegetation to generate accurate estimates of surface topography and canopy heights. Tropical forest areas with complex topography present a challenge for lidar remote sensing.ResultsWe compared digital terrain models (DTM) derived from airborne lidar data from a mountainous region of the Atlantic Forest in Brazil to 35 ground control points measured with survey grade GNSS receivers. The terrain model generated from full-density (~20 returns m−2) data was highly accurate (mean signed error of 0.19 ± 0.97 m), while those derived from reduced-density datasets (8 m−2, 4 m−2, 2 m−2 and 1 m−2) were increasingly less accurate. Canopy heights calculated from reduced-density lidar data declined as data density decreased due to the inability to accurately model the terrain surface. For lidar return densities below 4 m−2, the bias in height estimates translated into errors of 80–125 Mg ha−1 in predicted aboveground biomass.ConclusionsGiven the growing emphasis on the use of airborne lidar for forest management, carbon monitoring, and conservation efforts, the results of this study highlight the importance of careful survey planning and consistent sampling for accurate quantification of aboveground biomass stocks and dynamics. Approaches that rely primarily on canopy height to estimate aboveground biomass are sensitive to DTM errors from variability in lidar sampling density.Electronic supplementary materialThe online version of this article (doi:10.1186/s13021-015-0013-x) contains supplementary material, which is available to authorized users.