Automated tree crown detection and size estimation using multi-scale analysis of high-resolution satellite imagery

We tested an automated multi-scale approach for detecting individual trees and estimating tree crown geometry using high spatial resolution satellite imagery. Individual tree crowns are identified as local extrema points in the Laplacian-of-Gaussian scale-space pyramid that is constructed based on linear scale-space theory. The approach simultaneously detects tree crown centres and estimates tree crown sizes (radiuses). We evaluated our method using two 0.6-m resolution QuickBird images of a forest site that underwent a large shift in tree density between image captures due to drought-associated mortality. The automated multi-scale approach produced tree count estimates with an accuracy of 54% and 73% corresponding to the dense and sparse forests, respectively. Estimated crown diameters were linearly correlated with field-measured crown diameters (r = 0.73–0.86). Tree count accuracies and size estimates were comparable with alternative methods. Future use of the presented approach is merited based on the results of our study, but requires further investigation in a broader range of forest types.     

Estimation of forest stand diameter class using airborne lidar and field data

In this study, a method for estimating the stand diameter at breast height (DBH) classes in a South Korea forest using airborne lidar and field data was proposed. First, a digital surface model (DSM) and digital terrain model (DTM) were generated from the lidar data that have a point density of 4.3 points/m², then a tree canopy model (TCM) was created by subtracting the DTM from the DSM. The tree height and crown diameter were estimated from the rasterized TCM using local maximum points, minimum points and a circle fitting algorithm. Individual tree heights and crown diameters were converted into DBH using the allometric equations obtained from the field survey data. We calculated the proportion of the total number of individual trees belonging to each DBH class in each stand to determine the stand DBH class according to the standard guidelines. More than 60% of the stand DBH classes were correctly estimated by the proposed method, and their area occupied over 80% of the total forest area. The proposed method generated more accurate results compared to the digital forest type map provided by the government.     

Relationship between canopy height and Landsat ETM+ response in lowland Amazonian rainforest

This letter investigates the influence of within-pixel variation of canopy height on the spectral response recorded in Landsat Enhanced Thematic Mapper (ETM+) data for tropical rainforest. Forest canopy height is derived from airborne, small-footprint LiDAR data acquired using a Leica ALS50 II system. The field site is in the Tambopata National Reserve, in Peruvian Amazonia, where forest types include regenerating, swamp, floodplain and terra firme. For individual Landsat ETM+?bands, the strongest correlation for maximum, mean and standard deviation of canopy height occurred with ETM+?Band 4 (near infrared) for regenerating, floodplain and terra firme forest, and with ETM+?Band 5 (middle infrared) for swamp forest. For normalized difference band indices, ND42 and ND43 (i.e. the Normalized Difference Vegetation Index, NDVI) showed strong correlation with both mean and maximum canopy height for regenerating and terra firme forest, and with maximum and standard deviation of canopy height for floodplain forest. The palm-dominated swamp forest showed weaker relationships, with the strongest occurring for ND45 and ND52 with mean canopy height. Many papers have identified middle-infrared bands as being most sensitive to tropical rainforest structure, although these have often focussed on young regenerative forests. By focussing on older regenerative forest (of >25 years since land abandonment) and mature rainforest types, this work has shown that there is considerable variation with how structure may influence spectral reflectance and lends support to the hypothesis that canopy height distribution and shadowing effects caused by canopy complexity and the presence of emergent trees is what most significantly influences spectral response for tropical rainforests.     

A refined method for estimating capsule crops in individual jarrah (Eucalyptus marginata) crowns

An accurate measure of the number of capsules in the crowns of jarrah (Eucalyptus marginata) trees is needed to assess the potential for seedling regeneration prior to silvicultural treatment in jarrah forests. The current method of estimating capsule crops on jarrah trees uses stem diameter and estimates of capsule density in the crown, but has not been fully validated. In this study, we sought to develop an accurate and practical method of assessing capsule crops in the crowns of individual jarrah trees. We did this by measuring a number of tree characteristics prior to felling them. A total of 24 trees were selected, spanning a range of sizes and crown conditions, and the total number of capsules was counted for each tree. Multiple linear regression was used to model capsule number against various combinations of eight different tree/crown variables, with model fit compared using an adjusted coefficient of determination (adjR²). The final model recommended for field use included three easily measured variables (stem diameter, subjective assessment of capsule density, and subjective assessment of capsule clump distribution in the crown) and had a high degree of predictability (adjR² = 0.83), which was the same as that of the full model. This method substantially improved estimates of crown capsule numbers compared with the method currently used (adjR² increased from 0.29 to 0.83), which tended to underestimate canopy capsule numbers.     

Mapping forest functional type in a forest-shrubland ecotone using SPOT imagery and predictive habitat distribution modelling

The availability of land cover data at local scales is an important component in forest management and monitoring efforts. Regional land cover data seldom provide detailed information needed to support local management needs. Here we present a transferable framework to model forest cover by major plant functional type using aerial photos, multi-date Système Pour l’Observation de la Terre (SPOT) imagery, and topographic variables. We developed probability of occurrence models for deciduous broad-leaved forest and needle-leaved evergreen forest using logistic regression in the southern portion of the Wyoming Basin Ecoregion. The model outputs were combined into a synthesis map depicting deciduous and coniferous forest cover type. We evaluated the models and synthesis map using a field-validated, independent data source. Results showed strong relationships between forest cover and model variables, and the synthesis map was accurate with an overall correct classification rate of 0.87 and Cohen’s kappa value of 0.81. The results suggest our method adequately captures the functional type, size, and distribution pattern of forest cover in a spatially heterogeneous landscape.     

A note on the height variation of foliage clumping: comparison with remote sensing retrievals

Clumping index (CI), quantifying the level of foliage grouping within distinct canopy structures relative to a random distribution, is a key structural parameter of plant canopies and is very useful in ecological and meteorological models. In this letter, we report on validating the global foliage clumping map derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) data at 500 m resolution using new information about vertical profiles of foliage clumping in a wide range of forest type stands. We report that in moderate to dense forest stands with developed undergrowth layer, in situ measurements near the ground surface may considerably underestimate the overall canopy-level clumping effect. This is because the large gaps between tree crowns at upper levels of the canopy may not be all measured near the ground due to obscurity by lower vegetation of branches. This information about height variation of CI is shown to be important for correct estimating and validating the foliage clumping from airborne/satellite remote sensing.     

A study on the drivers of canopy reflectance variability in a boreal forest

The degree of which the observable canopy bidirectional reflectance factors (BRF) express plant trait variation at leaf and canopy scales is the fundamental physical basis underlying the use of optical remote sensing data for discriminating tree species and estimating forest biophysical variables. In this study, we quantified the relative contribution of variations in leaf optical properties (LOP), canopy structural properties, and understory reflectance, to canopy BRF variability in a boreal forest, at the spatial and spectral resolutions of Sentinel-2 (S2) Multi-Spectral Instrument. Our approach was based on physically-based forest reflectance model and global sensitivity analysis (SA) parameterized entirely with field measurements. Results showed LOP had dominant contribution to canopy BRF in shortwave infrared (SWIR) in multispecies forest areas, while canopy gap fraction in sensor’s view direction (i.e. nadir) was consistently found as the main driver of canopy BRF in red. This implies the satellite-measured BRF in red is the most robust predictor of effective canopy cover (ECC), while BRF in SWIR are optimal for tree species classification based on interspecific differences in mean leaf traits.     

Development of an automated individual tree detection model using point cloud LiDAR data for accurate tree counts in a Pinus radiata plantation

We develop and evaluate a new individual tree detection (ITD) algorithm to automatically locate and estimate the number of individual trees within a Pinus radiata plantation from relatively sparse airborne LiDAR point cloud data. The area of interest comprised stands covering a range of age classes and stocking levels. Our approach is based on local maxima (LM) filtering that tackles the issue of selecting the optimal search radius from the LiDAR point cloud for every potential LM using metrics derived from local neighbourhood data points; thus, it adapts to the local conditions, irrespective of canopy variability. This was achieved through two steps: (i) logistic regression model development using simulated stands composed of individual trees derived from real LiDAR point cloud data and (ii) application testing of the model using real plantation LiDAR point cloud data and geolocated, tree-level reference crowns that were manually identified in the LiDAR imagery. Our ITD algorithm performed well compared with previous studies, producing RMSE of 5.7% and a bias of only ?2.4%. Finally, we suggest that the ITD algorithm can be used for accurately estimating stocking and tree mapping, which in turn could be used to derive the plot-level metrics for an area-based approach for enhancing estimates of stand-level inventory attributes based on plot imputation.     

Bias in lidar-based canopy gap fraction estimates

Leaf area index and canopy gap fraction (GF) provide important information to forest managers regarding the ecological functioning and productivity of forest resources. Traditional measurements such as those obtained from hemispherical photography (HP) measure solar irradiation, penetrating the forest canopy, but do not provide information regarding the three-dimensional canopy structure. Terrestrial laser scanning (TLS) is an active sensor technology able to describe structural forest attributes by measuring interceptions of emitted laser pulses with the canopy and is able to record the spatial distribution of the foliage in three dimensions. However, due to the beam area of the laser, interceptions are detected more frequently than using conventional HP, and GF is typically underestimated. This study investigates the effects of laser beam area on the retrieval of GF by using morphological image processing to describe estimation bias as a function of canopy perimeters. The results show that, using canopy perimeter, improvements in correlation between HP and TLS can be achieved with an increase in the coefficient of determination R ² up to 28% (from an original R ² of 0.66 to an adjusted R ² of 0.85).     

Evaluating forest biometrics obtained from ground lidar in complex riparian forests

Terrestrial laser scanning is a technique that has been used increasingly in extracting forest biometrics such as trunk diameter and tree heights. Its potential, however, has not been fully explored in complex forested ecosystems, especially in riparian forests, considered among the most dynamic and complex portions of the Earth's biosphere. In this study, forest inventory data and multiple ground scans were obtained in a sparse managed and dense natural riparian forest on the immediate banks of the mid-lower portion of the Garonne River in Southern France, dominated by black poplar (Populus nigra) and commercial hybrid poplars (Populus × euramericana). Overall, the ground-based laser-scanning analysis successfully estimated trunk diameters, tree heights and crown radii from both managed and natural riparian forests. However, the ground scanner analysis was not as successful in identifying all of the trunks in the dense natural riparian forest, with only 141 trunks identified from a total of 234. This also results in allometric scaling exponents for ground scanning, which are significantly different from field-derived exponents. This study thus shows that there may be a limit to the number of trees detected in higher density forests, even with multiple scans.     

How much does the time lag between wildlife field-data collection and LiDAR-data acquisition matter for studies of animal distributions? A case study using bird communities

Vegetation structure quantified by light detection and ranging (LiDAR) can improve understanding of wildlife occupancy and species-richness patterns. However, there is often a time lag between the collection of LiDAR data and wildlife data. We investigated whether a time lag between the LiDAR acquisition and field-data acquisition affected mapped wildlife distributions ranging from an individual species distribution to total avian species richness in a conifer forest. We collected bird and LiDAR data in 2009 across a 20,000 ha forest in northern Idaho. Using the 2009 LiDAR data, we modelled the probability of occurrence for the brown creeper (Certhia americana). Using the same 2009 LiDAR data, we additionally modelled total avian species richness and richness of three different bird nesting guilds (ground/understory, mid/upper canopy and cavity). We mapped brown creeper occupancy probability and species richness using the 2009 models, and then compared these maps with maps based on the same models applied to a 2003-LiDAR dataset. A prior study identified areas harvested between 2003 and 2009. There was on average a 5% absolute decrease in mapped probabilities of brown creeper occurrence in non-harvest areas between 2003 and 2009. Species richness changed by less than one species in all cases within non-harvest areas between the 2003 and 2009 maps. Although these comparisons were statistically significant at the p < 0.0001 level, it is likely that the high number of map cells (~480,000) influenced this result. Similar patterns between our 2003 and 2009 maps in non-harvest areas for this suite of avian responses suggests that a 6-year difference between field-data collection and LiDAR-data collection has a minimal effect on mapped avian patterns in an undisturbed coniferous forest. However, because this is one case study in one ecosystem, additional work examining the effect of temporal lags between LiDAR and field-data collection on mapping wildlife distributions is warranted in additional ecosystems.     

A new method for generating canopy height models from discrete-return LiDAR point clouds

Tree height underestimation and occurrence of visible data pits are two major problems in light detection and ranging (LiDAR)-derived digital surface models and canopy height models (CHMs) in forested areas. To address the two major problems, a new method is proposed for generating CHMs from discrete-return LiDAR point clouds using a selecting and sorting mechanism based on a circle centred at the target point, followed by spatial interpolation. Test results from simulated and real LiDAR point clouds show that the new method outperformed three other methods in terms of treetop approximation, crown surface representation and data pit removal.     

The potential of Sentinel-2 data for estimating biophysical variables in a boreal forest: a simulation study

In this study we use ground reference data from 962 forest plots to demonstrate the potential of Sentinel-2 (S2) bands in estimating canopy biophysical properties in boreal forests in Finland. We simulated canopy bidirectional reflectance factors (BRFs) using the PARAS model, which applies photon recollision probability. Results showed that the highest correlation between simulated S2 BRFs and fraction of absorbed photosynthetically active radiation (fPAR) was for the band combination band 7/band 9 (wavelengths 773–793 nm and 935–955 nm, respectively) (the coefficient of determination (R²) was 0.93). For effective leaf area index (LAI_e) the best band combination was band 8/band 4 (wavelengths 785–900 nm and 650–680 nm, respectively) (R² = 0.93). Based on this study, the above-ground biomass (AGB) and S2 band combinations did not show strong relationships (R² = 0.24). The new inverted red-edge chlorophyll index (IRECI) and Sentinel-2 red-edge position – index (S2REP) showed moderate relationships with fPAR (R² = 0.61 and R² = 0.45, respectively) and LAI_e (R² = 0.56 and R² = 0.30, respectively). This study demonstrated the potential of the S2 data to estimate canopy biophysical properties.     

A simple technique for co-registration of terrestrial LiDAR observations for forestry applications

Light detection and ranging (LiDAR) from terrestrial platforms provides unprecedented detail about the three-dimensional structure of forest canopies. Although airborne laser scanning is designed to yield a relatively homogeneous distribution of returns, the radial perspective of terrestrial laser scanning (TLS) results in a rapid decrease of number of returns with increasing distance from the instrument. Additionally, when used in forested environments, significant parts of the area under investigation may be obscured by tree trunks and understorey. A possible approach to mitigate this effect is to combine TLS observations acquired at different locations to obtain multiple perspectives of an area under investigation. The denser and more evenly distributed observations then allow a spatially explicit and more comprehensive study of forest characteristics. This study demonstrates a simple approach to combine TLS observations made at multiple locations using bright reference targets as tie-points. Results show this technique was able to accurately combine the different TLS data sets (root mean square error (RMSE): 0.04–0.7 m, coefficient of determination (R ²): 0.70–0.99). Terrain elevations from TLS system were highly correlated with field-measured terrain heights (R ²: 0.70–0.98).     

Wildfires in Russia in 2000–2008: estimates of burnt areas using the satellite MODIS MCD45 data

Fire activity in Russia in 2000–2008 was analysed on the basis of the newest satellite MODIS MCD45 Burned Area data. The estimated annual burned areas exhibited strong variability in magnitude, geographical location and affected vegetation, varying in total from 5.4 Mha in 2000 to 33.0 Mha in 2008. The total area burned in 2000–2008 in forests and woodlands (the areas with percent tree canopy cover >40%) was 43.4 Mha, of which about 90% was burned in the eastern Siberia and the Far East, and 53% was burned during the large fire years 2003 and 2008. The forest burned areas were mainly localized within 50–55°N, approximately along the southern edge of the area of boreal forests. The reliability of the resultant estimates was tested by comparison with the known early published results. Some applications of the results in problems of atmospheric composition and air quality are also discussed.     

Comparing yield estimates derived from LiDAR and aerial photogrammetric point-cloud data with cut-to-length harvester data in a Pinus radiata plantation in Tasmania

Accurate mapping of timber resources in commercial forestry is essential to support planning and management operations of forest growers. Over the last two decades, Light Detection and Ranging (LiDAR) systems have been successfully deployed for the collection of point-cloud data for accurate modelling of forest attributes that are traditionally obtained from plot-based inventory. In recent years, studies conducted in North America and Scandinavia have shown that three-dimensional point clouds derived from digital aerial photogrammetric (AP) data can be used to model forest attributes with a level of accuracy similar to traditional LiDAR-based approaches. A comparative analysis of the performance of the two point-cloud technologies has never been attempted in Australian plantations. In this study, we compared the performance of LiDAR-based and AP-based point clouds for estimating total recoverable volume in a Pinus radiata plantation at Springfield in north-eastern Tasmania, using volume data collected by harvesting machines as a reference. Our results showed that AP point clouds can be used for mapping total recoverable volume in P. radiata plantations with levels of accuracy that are comparable to LiDAR-based estimates. Plot-level relative root mean squared error (RMSE%) values were 23.85% for LiDAR and ranged from 22.07% to 27.10% for the three AP dense point-cloud settings evaluated. At the stand level, RMSE% decreased to 9.86% and 8.91% for LiDAR and AP, respectively. Both LiDAR-based and AP-based modelled volumes showed a close agreement with volumes measured using harvester head data, demonstrating the potential of AP technology for the management and planning of forestry operations in softwood plantations.     

A global corrected SRTM DEM product for vegetated areas

It has been found that the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) is systematically higher than the actual land surface in vegetated areas. This study developed a new globally corrected SRTM DEM through reducing its systematic bias in vegetated areas. Over 150, 000 km² airborne light detection and ranging (LiDAR) data along with spaceborne LiDAR, global canopy height data, global canopy cover data and global land cover data were collected to correct the SRTM DEM. A linear regression based method was used to estimate the original SRTM DEM error and therefore correct the SRTM DEM data. The results show that the original SRTM DEM data is around 6 m higher than the actual land surfaces on average across all vegetation types. Our corrected SRTM DEM data can significantly reduce the significant bias to near zero, and can also reduce the root-mean-square error by 1 m.     

Retrieving leaf area index using ICESat/GLAS full-waveform data

Leaf area index (LAI) is an important parameter controlling many biological and physical processes associated with vegetation on the Earth's surface. In this study, an algorithm for estimating LAI from the ICESat (Ice, Cloud and land Elevation Satellite)/GLAS (Geoscience Laser Altimeter System) data was proposed and applied to a forest area in the Tibetan Plateau. First, Gaussian decomposition of the GLAS waveform was implemented to identify the ground peaks and calculate the ground and canopy return energy. Second, the ground-to-total energy ratio (E _r) was computed as the ratio of the ground return energy to the total waveform return energy for each GLAS footprint. Third, a regression model between the E _r and the field-measured LAI was established based on the Beer–Lambert law. The coefficient of determination (R ²) of the model was 0.81 and the root mean square error (RMSE) is 0.35 (n?=?23, p < 0.001). Finally, the leave-one-out cross-validation procedure was used to assess the constructed regression model. The results indicate that the regression model is not overfitting the data and has a good generalization capability. We validated the accuracy of the GLAS-predicted LAIs using the other 15 field-measured LAIs (R ² = 0.84), and the result shows that the accuracy of the GLAS-predicted LAI is high (RMSE = 0.31).     

Use of ground and LiDAR data to model the NPP of a Mediterranean pine forest

This letter presents the application of a recently developed methodology to predict the net primary production (NPP) of a Mediterranean pine forest in San Rossore, Central Italy. The methodology is based on the use of two models, C-Fix and BIOME-BGC, whose outputs are combined with estimates of stand volume and age to describe the actual trophic status of the examined ecosystems. This work investigates the possibility of deriving these two forest attributes from airborne high-resolution Light Detection and Ranging (LiDAR) data. Specifically, stand volume and age estimates are obtained by linear transformation of LiDAR average and maximum stand canopy heights, respectively. The NPP modelling strategy driven by these estimates yields results that are evaluated by comparison with ground measurements of current annual increment. The success of this test opens the way to integrate ecosystem modelling techniques and LiDAR data for simulating net forest carbon fluxes.