Carine Klauberg

Mapping Forest Structure and Composition from Low-Density LiDAR for Informed Forest, Fuel, and Fire Management at Eglin Air Force Base, Florida, USA

Abstract Eglin Air Force Base (AFB) in Florida, in the United States, conserves a large reservoir of native longleaf pine (Pinus palustris Mill.) stands that land managers maintain by using frequent fires. We predicted tree density, basal area, and dominant tree species from 195 forest inventory plots, low-density airborne LiDAR, and Landsat data available across the entirety of Eglin AFB. We used the Random Forests (RF) machine learning algorithm to predict the 3 overstory responses via univariate regression or classification, or multivariate k-NN imputation. Ten predictor variables explained ∼ 50% of variation and were used in all models. Model accuracy and precision statistics were similar among the various RF approaches, so we chose the imputation approach for its advantage of allowing prediction of the ancillary plot attributes of surface fuels and ground cover plant species richness. Maps of the 3 overstory response variables and ancillary attributes were imputed at 30-m resolution and then aggregated to the management block level, where they were significantly correlated with each other and with fire history variables summarized from independent data. We conclude that functional relationships among overstory structure, surface fuels, species richness, and fire history emerge and become more apparent at the block level where management decisions are made.

Impacts of Airborne Lidar Pulse Density on Estimating Biomass Stocks and Changes in a Selectively Logged Tropical Forest

Airborne lidar is a technology well-suited for mapping many forest attributes, including aboveground biomass (AGB) stocks and changes in selective logging in tropical forests. However, trade-offs still exist between lidar pulse density and accuracy of AGB estimates. We assessed the impacts of lidar pulse density on the estimation of AGB stocks and changes using airborne lidar and field plot data in a selectively logged tropical forest located near Paragominas, Para, Brazil. Field-derived AGB was computed at 85 square 50 × 50 m plots in 2014. Lidar data were acquired in 2012 and 2014, and for each dataset the pulse density was subsampled from its original density of 13.8 and 37.5 pulses·m−2 to lower densities of 12, 10, 8, 6, 4, 2, 0.8, 0.6, 0.4 and 0.2 pulses·m−2. For each pulse density dataset, a power-law model was developed to estimate AGB stocks from lidar-derived mean height and corresponding changes between the years 2012 and 2014. We found that AGB change estimates at the plot level were only slightly affected by pulse density. However, at the landscape level we observed differences in estimated AGB change of >20 Mg·ha−1 when pulse density decreased from 12 to 0.2 pulses·m−2. The effects of pulse density were more pronounced in areas of steep slope, especially when the digital terrain models (DTMs) used in the lidar derived forest height were created from reduced pulse density data. In particular, when the DTM from high pulse density in 2014 was used to derive the forest height from both years, the effects on forest height and the estimated AGB stock and changes did not exceed 20 Mg·ha−1. The results suggest that AGB change can be monitored in selective logging in tropical forests with reasonable accuracy and low cost with low pulse density lidar surveys if a baseline high-quality DTM is available from at least one lidar survey. We recommend the results of this study to be considered in developing projects and national level MRV systems for REDD+ emission reduction programs for tropical forests.

Modeling and mapping basal area of Pinus taeda L. plantation using airborne LiDAR data

Basal area (BA) is a good predictor of timber stand volume and forest growth. This study developed predictive models using field and airborne LiDAR (Light Detection and Ranging) data for estimation of basal area in Pinus taeda plantation in south Brazil. In the field, BA was collected from conventional forest inventory plots. Multiple linear regression models for predicting BA from LiDAR-derived metrics were developed and evaluated for predictive power and parsimony. The best model to predict BA from a family of six models was selected based on corrected Akaike Information Criterion (AICc) and assessed by the adjusted coefficient of determination (adj. R²) and root mean square error (RMSE). The best model revealed an adj. R²=0.93 and RMSE=7.74%. Leave one out cross-validation of the best regression model was also computed, and revealed an adj. R² and RMSE of 0.92 and 8.31%, respectively. This study showed that LiDAR-derived metrics can be used to predict BA in Pinus taeda plantations in south Brazil with high precision. We conclude that there is good potential to monitor growth in this type of plantations using airborne LiDAR. We hope that the promising results for BA modeling presented herein will stimulate to operate this technology in Brazil.

Remote Sensing of Leaf Area Index from LiDAR Height Percentile Metrics and Comparison with MODIS Product in a Selectively Logged Tropical Forest Area in Eastern Amazonia

Leaf area index (LAI) is an important parameter to describe the capacity of forests to intercept light and thus affects the microclimate and photosynthetic capacity of canopies. In general, tropical forests have a higher leaf area index and it is a challenge to estimate LAI in a forest with a very dense canopy. In this study, it is assumed that the traditional Light Detection and Ranging (LiDAR)-derived fractional vegetation cover (fCover) has weak relationship with leaf area index in a dense forest. We propose a partial least squares (PLS) regression model using the height percentile metrics derived from airborne LiDAR data to estimate the LAI of a dense forest. Ground inventory and airborne LiDAR data collected in a selectively logged tropical forest area in Eastern Amazonia are used to map LAI from the plot level to the landscape scale. The results indicate that the fCover, derived from the first return or the last return, has no significant correlations with the ground-based LAI. The PLS model evaluated by the leave-one-out validation shows that the estimated LAI is significantly correlated with the ground-based LAI with an R2 of 0.58 and a root mean square error (RMSE) of 1.13. A data comparison indicates that the Moderate Resolution Imaging Spectrometer (MODIS) LAI underestimates the landscape-level LAI by about 22%. The MODIS quality control data show that in the selected tile, the cloud state is not the primary factor affecting the MODIS LAI performance; rather, the LAI from the main radiative transfer (RT) algorithm contributes much to the underestimation of the LAI in the tropical forest. In addition, the results show that the LiDAR-based LAI has a better response to the logging activities than the MODIS-based LAI, and that the leaf area reduction caused by logging is about 13%. In contrast, the MODIS-based LAI exhibits no apparent spatial correlation with the LiDAR-based LAI. It is suggested that the main algorithm of MODIS should be improved with regard to tropical forests. The significance of this study is the proposal of a framework to produce ground-based LAI using forest inventory data and determine the plot-level LAI at the airborne and satellite scale using LiDAR data.

Supervised spatial classification of multispectral LiDAR data in urban areas

Multispectral LiDAR (light detection and ranging) data have been initially used for land cover classification. However, there are still high classification uncertainties, especially in urban areas, where objects are often mixed and confounded. This study investigated the efficiency of combining advanced statistical methods and LiDAR metrics derived from multispectral LiDAR data for improving land cover classification accuracy in urban areas. The study area is located in Oshawa, Ontario, Canada, on the Lake Ontario shoreline. Multispectral Optech Titan LiDAR data over the study area were acquired on 3 September 2014 in a single strip of 3 km2. Using the channels at 1,550 nm (C1), 1,064 nm (C2) and 532 nm (C3), LiDAR intensity data, normalized digital surface model (nDSM), pseudo normalized difference vegetation index (PseudoNDVI), morphological profiles (MP), and a novel hierarchical morphological profiles (HMP) were derived and used as features for the classification. A support vector machine classifier with a radial basis function (RBF) kernel was applied in the classification stage, where the optimal parameters for the classifier were selected by a grid search procedure. The combination of intensity, pseudoNDVI, nDSM and HMP resulted in the best land cover classification, with an overall accuracy of 93.28%.

Zero-spiked regression models generated by gamma random variables with application in the resin oil production

ABSTRACT Zero-inflated data are more frequent when the data represent counts. However, there are practical situations in which continuous data contain an excess of zeros. In these cases, the zero-inflated Poisson, binomial or negative binomial models are not suitable. In order to reduce this gap, we propose the zero-spiked gamma-Weibull (ZSGW) model by mixing a distribution which is degenerate at zero with the gamma-Weibull distribution, which has positive support. The model attempts to estimate simultaneously the effects of explanatory variables on the response variable and the zero-spiked. We consider a frequentist analysis and a non-parametric bootstrap for estimating the parameters of the ZSGW regression model. We derive the appropriate matrices for assessing local influence on the model parameters. We illustrate the performance of the proposed regression model by means of a real data set (copaiba oil resin production) from a study carried out at the Department of Forest Science of the Luiz de Queiroz School of Agriculture, University of São Paulo. Based on the ZSGW regression model, we determine the explanatory variables that can influence the excess of zeros of the resin oil production and identify influential observations. We also prove empirically that the proposed regression model can be superior to the zero-adjusted inverse Gaussian regression model to fit zero-inflated positive continuous data.

Use of ordinary kriging and Gaussian conditional simulation to interpolate airborne fire radiative energy density estimates

Fire radiative energy density (FRED, J m−2) integrated from fire radiative power density (FRPD, W m−2) observations of landscape-level fires can present an undersampling problem when collected from fixed-wing aircraft. In the present study, the aircraft made multiple passes over the fire at ~3 min intervals, thus failing to observe most of the FRPD emitted as the flame front spread. We integrated the sparse FRPD time series to obtain pixel-level FRED estimates, and subsequently applied ordinary kriging (OK) and Gaussian conditional simulation (GCS) to interpolate across data voids caused by the undersampling. We compared FRED interpolated via OK and GCS with FRED estimated independently from ground measurements of biomass consumed from five prescribed burns at Eglin Air Force Base, Florida, USA. In four of five burns considered where undersampling prevailed, OK and GCS effectively interpolated FRED estimates across the data voids, improving the spatial distribution of FRED across the burning event and its overall mean. In a fifth burn, the burning characteristics were such that undersampling did not present a problem needing to be fixed. We also determined where burning and FRPD sampling characteristics merited applying OK and CGS only to the highest FRED estimates to interpolate more accurate FRED maps.

Comparing the Performance of Ground Filtering Algorithms for Terrain Modeling in a Forest Environment Using Airborne LiDAR Data

The aim of this study was to evaluate the performance of four ground filtering algorithms to generate digital terrain models (DTMs) from airborne light detection and ranging (LiDAR) data. The study area is a forest environment located in Washington state, USA with distinct classes of land use and land cover (e.g., shrubland, grassland, bare soil, and three forest types according to tree density and silvicultural interventions: closed-canopy forest, intermediate-canopy forest, and open-canopy forest). The following four ground filtering algorithms were assessed: Weighted Linear Least Squares (WLS), Multi-scale Curvature Classification (MCC), Progressive Morphological Filter (PMF), and Progressive Triangulated Irregular Network (PTIN). The four algorithms performed well across the land cover, with the PMF yielding the least number of points classified as ground. Statistical differences between the pairs of DTMs were small, except for the PMF due to the highest errors. Because the forestry sector requires constant updating of topographical maps, open-source ground filtering algorithms, such as WLS and MCC, performed very well on planted forest environments. However, the performance of such filters should also be evaluated over complex native forest environments.

Estimating Stand Height and Tree Density in Pinus taeda plantations using in-situ data, airborne LiDAR and k-Nearest Neighbor Imputation

Accurate forest inventory is of great economic importance to optimize the entire supply chain management in pulp and paper companies. The aim of this study was to estimate stand dominate and mean heights (HD and HM) and tree density (TD) of Pinus taeda plantations located in South Brazil using in-situ measurements, airborne Light Detection and Ranging (LiDAR) data and the non- k-nearest neighbor (k-NN) imputation. Forest inventory attributes and LiDAR derived metrics were calculated at 53 regular sample plots and we used imputation models to retrieve the forest attributes at plot and landscape-levels. The best LiDAR-derived metrics to predict HD, HM and TD were H99TH, HSD, SKE and HMIN. The Imputation model using the selected metrics was more effective for retrieving height than tree density. The model coefficients of determination (adj.R2) and a root mean squared difference (RMSD) for HD, HM and TD were 0.90, 0.94, 0.38m and 6.99, 5.70, 12.92%, respectively. Our results show that LiDAR and k-NN imputation can be used to predict stand heights with high accuracy in Pinus taeda. However, furthers studies need to be realized to improve the accuracy prediction of TD and to evaluate and compare the cost of acquisition and processing of LiDAR data against the conventional inventory procedures.