Exploring LiDAR-RaDAR synergy—predicting aboveground biomass in a southwestern ponderosa pine forest using LiDAR, SAR and InSAR

Scanning Light Detecting and Ranging (LiDAR), Synthetic Aperture Radar (SAR) and Interferometric SAR (InSAR) were analyzed to determine (1) which of the three sensor systems most accurately predicted forest biomass, and (2) if LiDAR and SAR/InSAR data sets, jointly considered, produced more accurate, precise results relative to those same data sets considered separately. LiDAR ranging measurements, VHF-SAR cross-sectional returns, and X- and P-band cross-sectional returns and interferometric ranges were regressed with ground-estimated (from dbh) forest biomass in ponderosa pine forests in the southwestern United States. All models were cross-validated. Results indicated that the average canopy height measured by the scanning LiDAR produced the best predictive equation. The simple linear LiDAR equation explained 83% of the biomass variability (n = 52 plots) with a cross-validated root mean square error of 26.0 t/ha. Additional LiDAR metrics were not significant to the model. The GeoSAR P-band (λ = 86 cm) cross-sectional return and the GeoSAR/InSAR canopy height (X-P) captured 30% of the forest biomass variation with an average predictive error of 52.5 t/ha. A second RaDAR-FOPEN collected VHF (λ ∼ 7.8 m) and cross-polarized P-band (λ = 88 cm) cross-sectional returns, none of which proved useful for forest biomass estimation (cross-validated R² = 0.09, RMSE = 63.7 t/ha). Joint consideration of LiDAR and RaDAR measurements produced a statistically significant, albeit small improvement in biomass estimation precision. The cross-validated R² increased from 83% to 84% and the prediction error decreased from 26.0 t/ha to 24.9 t/ha when the GeoSAR X-P interferometric height is considered along with the average LiDAR canopy height. Inclusion of a third LiDAR metric, the 60th decile height, further increased the R² to 85% and decreased the RMSE to 24.1 t/ha. On this 11 km² ponderosa pine study area, LiDAR data proved most useful for predicting forest biomass. RaDAR ranging measurements did not improve the LiDAR estimates.     

Investigating RaDAR-LiDAR synergy in a North Carolina pine forest

A low frequency (80-120 MHz) VHF RaDAR, BioSAR, specifically designed for forest biomass estimation and a profiling LiDAR, PALS, were flown over loblolly pine plantations in the southeastern United States. LiDAR-only, RaDAR-only, and joint LiDAR-RaDAR linear models were developed to determine if returns from two sensors could be used to estimate pine biomass more accurately and precisely than returns from either sensor alone. The best five-variable RaDAR model explained 81.8% (R²) of the stem green biomass variability, with a regression RMSE of 57.5 t/ha. The best one-variable LiDAR model explained 93.3% of the biomass variation (RMSE = 33.9 t/ha). Combining the RaDAR normalized volumetric returns with the profiling LiDAR ranging measurements did little to improve the best LiDAR-only model. The best LiDAR-RaDAR model explained 93.8% of the biomass variation (RSME = 32.7 t/ha). Cross-validation and training/test validation procedures demonstrated (1) that all models are unbiased and (2) the increased precision of the LiDAR-only and LiDAR-RaDAR models. The results of this investigation and a companion study indicate that there is little to be gained combining VHF-RaDAR volumetric returns and profiling LiDAR ranging measurements in pine forests; a LiDAR ranging system is sufficient for accurate, precise biomass estimation.     

Aboveground biomass retrieval in tropical forests — The potential of combined X- and L-band SAR data use

In the context of reducing emissions from deforestation and forest degradation (REDD) and the international effort to reduce anthropogenic greenhouse gas emissions, a reliable assessment of aboveground forest biomass is a major requirement. Especially in tropical forests which store huge amounts of carbon, a precise quantification of aboveground biomass is of high relevance for REDD activities. This study investigates the potential of X- and L-band SAR data to estimate aboveground biomass (AGB) in intact and degraded tropical forests in Central Kalimantan, Borneo, Indonesia. Based on forest inventory data, aboveground biomass was first estimated using LiDAR data. These results were then used to calibrate SAR backscatter images and to upscale the biomass estimates across large areas and ecosystems. This upscaling approach not only provided aboveground biomass estimates over the whole biomass range from woody regrowth to mature pristine forest but also revealed a spatial variation due to varying growth condition within specific forest types. Single and combined frequencies, as well as mono- and multi-temporal TerraSAR-X and ALOS PALSAR biomass estimation models were analyzed for the development of accurate biomass estimations. Regarding the single frequency analysis overall ALOS PALSAR backscatter is more sensitive to AGB than TerraSAR-X, especially in the higher biomass range (> 100 t/ha). However, ALOS PALSAR results were less accurate in low biomass ranges due to a higher variance. The multi-temporal L- and X-band combined model achieved the best result and was therefore tested for its temporal and spatial transferability. The achieved accuracy for this model using nearly 400 independent validation points was r² = 0.53 with an RMSE of 79 t/ha. The model is valid up to 307 t/ha with an accuracy requirement of 50 t/ha and up to 614 t/ha with an accuracy requirement of 100 t/ha in flat terrain. The results demonstrate that direct biomass measurements based on the synergistic use of L- and X-band SAR can provide large-scale AGB estimations for tropical forests. In the context of REDD monitoring the results can be used for the assessment of the spatial distribution of the biomass, also indicating trends in high biomass ranges and the characterization of the spatial patterns in different forest types.     

Biomass estimation over a large area based on standwise forest inventory data and ASTER and MODIS satellite data: A possibility to verify carbon inventories

According to the IPCC GPG (Intergovernmental Panel on Climate Change, Good Practice Guidance), remote sensing methods are especially suitable for independent verification of the national LULUCF (Land Use, Land-Use Change, and Forestry) carbon pool estimates, particularly the aboveground biomass. In the present study, we demonstrate the potential of standwise (forest stand is a homogenous forest unit with average size of 1-3 ha) forest inventory data, and ASTER and MODIS satellite data for estimating stand volume (m³ ha^− 1) and aboveground biomass (t ha^− 1) over a large area of boreal forests in southern Finland. The regression models, developed using standwise forest inventory data and standwise averages of moderate spatial resolution ASTER data (15 m × 15 m), were utilized to estimate stand volume for coarse resolution MODIS pixels (250 m × 250 m). The MODIS datasets for three 8-day periods produced slightly different predictions, but the averaged MODIS data produced the most accurate estimates. The inaccuracy in radiometric calibration between the datasets, the effect of gridding and compositing artifacts and phenological variability are the most probable reasons for this variability. Averaging of the several MODIS datasets seems to be one possibility to reduce bias. The estimates obtained were significantly close to the district-level mean values provided by the Finnish National Forest Inventory; the relative RMSE was 9.9%. The use of finer spatial resolution data is an essential step to integrate ground measurements with coarse spatial resolution data. Furthermore, the use of standwise forest inventory data reduces co-registration errors and helps in solving the scaling problem between the datasets. The approach employed here can be used for estimating the stand volume and biomass, and as required independent verification data.     

Improved forest biomass estimates using ALOS AVNIR-2 texture indices

Optical remote sensing is still one of the most attractive choices for obtaining biomass information, as new sensors are available with fine spatial and spectral resolutions. Better biomass estimates may be possible if suitable processing techniques for these sensors can be demonstrated. This research investigates the potential of high resolution optical data from the ALOS AVNIR-2 sensor for biomass estimation in a mountainous, subtropical forested region using four different types of image processing techniques including i) spectral reflectance and simple spectral band ratio, ii) commonly used vegetation indices, iii) texture parameters and iv) ratio of texture parameters. Simple linear and stepwise multiple regression models were developed between biomass data from 50 field plots, and image parameters derived from these techniques.Results indicate that spectral reflectance, the simple band ratio, and commonly used vegetation indices have relatively low potential for biomass estimation, as only about 58% of the variability in the field data was explained by the model (adjusted r² = 0.58 and RMSE = 64 t/ha). However, the texture parameters of spectral bands were found to be effective for biomass estimation with an explained variability of ca. 76% (adjusted r² = 0.76 and RMSE = 46 t/ha). The result was further improved to adjusted r² = 0.88 (RMSE = 32 t/ha) using the simple ratio of texture parameters. The results suggest that the performance of biomass estimation can be improved significantly using the texture parameters of high resolution optical data, and further improvement can be obtained using the ratio of texture parameters, as this combines the advantages of both texture and ratio.     

Estimating forest canopy height and terrain relief from GLAS waveform metrics

Quantifying aboveground biomass in forest ecosystems is required for carbon stock estimation, aspects of forest management, and further developing a capacity for monitoring carbon stocks over time. Airborne Light Detection And Ranging (LiDAR) systems, of all remote sensing technologies, have been demonstrated to yield the most accurate estimates of aboveground biomass for forested areas over a wide range of biomass values. However, these systems are limited by considerations including large data volumes and high costs. Within the constraints imposed by the nature of the satellite mission, the GeoScience Laser Altimeter System (GLAS) aboard ICESat has provided data conferring information regarding forest vertical structure for large areas at a low end user cost. GLAS data have been demonstrated to accurately estimate forest height and aboveground biomass especially well in topographically smooth areas with homogeneous forested conditions. However in areas with dense forests, high relief, or heterogeneous vegetation cover, GLAS waveforms are more complex and difficult to consistently characterize. We use airborne discrete return LiDAR data to simulate GLAS waveforms and to subsequently deconstruct coregistered GLAS waveforms into vegetation and ground returns. A series of waveform metrics was calculated and compared to topography and vegetation information gleaned from the airborne data. A model to estimate maximum relief directly from waveform metrics was developed with an R² of 0.76 (n = 110), and used for the classification of the maximum relief of the areas sensed by GLAS. Discriminant analysis was also conducted as an alternative classification technique. A model was also developed estimating forest canopy height from waveform metrics for all of the data (R² = 0.81, n = 110) and for the three separate relief classes; maximum relief 0-7 m (R² = 0.83, n = 44), maximum relief 7-15 m (R² = 0.88, n = 41) and maximum relief > 15 m (R² = 0.75, n = 25). The moderate relief class model yielded better predictions of forest height than the low relief class model which is attributed to the increasing variability of waveform metrics with terrain relief. The moderate relief class model also yielded better predictions than the high relief class model because of the mixing of vegetation and terrain signals in waveforms from high relief footprints. This research demonstrates that terrain can be accurately modeled directly from GLAS waveforms enabling the inclusion of terrain relief, on a waveform specific basis, as supplemental model input to improve estimates of canopy height.     

Estimating aboveground forest biomass from canopy reflectance model inversion in mountainous terrain

The amount and spatial distribution of aboveground forest biomass (AGB) are required inputs to forest carbon budgets and ecosystem productivity models. Satellite remote sensing offers distinct advantages for large area and multi-temporal applications, however, conventional empirical methods for estimating forest canopy structure and AGB can be difficult in areas of high relief and variable terrain. This paper introduces a new method for obtaining AGB from forest structure estimates using a physically-based canopy reflectance (CR) model inversion approach. A geometric-optical CR model was run in multiple forward mode (MFM) using SPOT-5 imagery to derive forest structure and biomass at Kananaskis, Alberta in the Canadian Rocky Mountains. The approach first estimates tree crown dimensions and stem density for satellite image pixels which are then related to tree biomass and AGB using a crown spheroid surface area approach. MFM estimates of AGB were evaluated for 36 deciduous (trembling aspen) and conifer (lodgepole pine) field validation sites and compared against spectral mixture analysis (SMA) and normalised difference vegetation index (NDVI) biomass predictions from atmospherically and topographically corrected (SCS+C) imagery. MFM provided the lowest error for all validation plots of 31.7 tonnes/hectare (t/ha) versus SMA (32.6 t/ha error) and NDVI (34.7 t/ha) as well as for conifer plots (MFM: 23.0 t/ha; SMA 27.9 t/ha; NDVI 29.7 t/ha) but had higher error than SMA and NDVI for deciduous plots (by 4.5 t/ha and 2.1 t/ha, respectively). The MFM approach was considerably more stable over the full range of biomass values (67 to 243 t/ha) measured in the field. Field plots with biomass > 1 standard deviation from the field mean (over 30% of plots) had biomass estimation errors of 37.9 t/ha using MFM compared with 65.5 t/ha and 67.5 t/ha error from SMA and NDVI, respectively. In addition to providing more accurate overall results and greater stability over the range of biomass values, the MFM approach also provides a suite of other biophysical structural outputs such as density, crown dimensions, LAI, height and sub-pixel scale fractions. Its explicit physical-basis and minimal ground data requirements are also more appropriate for larger area, multi-scene, multi-date applications with variable scene geometry and in high relief terrain. MFM thus warrants consideration for applications in mountainous and other, less complex terrain for purposes such as forest inventory updates, ecological modeling and terrestrial biomass and carbon monitoring studies.     

Mapping biomass and stress in the Sierra Nevada using lidar and hyperspectral data fusion

In this paper, we explored fusion of structural metrics from the Laser Vegetation Imaging Sensor (LVIS) and spectral characteristics from the Airborne Visible Infrared Imaging Spectrometer (AVIRIS) for biomass estimation in the Sierra Nevada. In addition, we combined the two sensors to map species-specific biomass and stress at landscape scale. Multiple endmember spectral mixture analysis (MESMA) was used to classify vegetation from AVIRIS images and obtain sub-pixel fractions of green vegetation, non-photosynthetic vegetation, soil, and shade. LVIS metrics, AVIRIS spectral indices, and MESMA fractions were compared with field measures of biomass using linear and stepwise regressions at stand (1 ha) level. AVIRIS metrics such as water band indices and shade fractions showed strong correlation with LVIS canopy height (r² = 0.69, RMSE = 5.2 m) and explained around 60% variability in biomass. LVIS variables were found to be consistently good predictors of total and species specific biomass (r² = 0.77, RMSE = 70.12 Mg/ha). Prediction by LVIS after species stratification of field data reduced errors by 12% (r² = 0.84, RMSE = 58.78 Mg/ha) over using LVIS metrics alone. Species-specific biomass maps and associated errors created from fusion were different from those produced without fusion, particularly for hardwoods and pines, although mean biomass differences between the two techniques were not statistically significant. A combined analysis of spatial maps from LVIS and AVIRIS showed increased water and chlorophyll stress in several high biomass stands in the study area. This study provides further evidence that lidar is better suited for biomass estimation, per se, while the best use of hyperspectral data may be to refine biomass predictions through a priori species stratification, while also providing information on canopy state, such as stress. Together, the two sensors have many potential applications in carbon dynamics, ecological and habitat studies.     

Forest carbon densities and uncertainties from Lidar, QuickBird, and field measurements in California

Greenhouse gas inventories and emissions reduction programs require robust methods to quantify carbon sequestration in forests. We compare forest carbon estimates from Light Detection and Ranging (Lidar) data and QuickBird high-resolution satellite images, calibrated and validated by field measurements of individual trees. We conducted the tests at two sites in California: (1) 59 km² of secondary and old-growth coast redwood (Sequoia sempervirens) forest (Garcia-Mailliard area) and (2) 58 km² of old-growth Sierra Nevada forest (North Yuba area). Regression of aboveground live tree carbon density, calculated from field measurements, against Lidar height metrics and against QuickBird-derived tree crown diameter generated equations of carbon density as a function of the remote sensing parameters. Employing Monte Carlo methods, we quantified uncertainties of forest carbon estimates from uncertainties in field measurements, remote sensing accuracy, biomass regression equations, and spatial autocorrelation. Validation of QuickBird crown diameters against field measurements of the same trees showed significant correlation (r = 0.82, P < 0.05). Comparison of stand-level Lidar height metrics with field-derived Lorey's mean height showed significant correlation (Garcia-Mailliard r = 0.94, P < 0.0001; North Yuba R = 0.89, P < 0.0001). Field measurements of five aboveground carbon pools (live trees, dead trees, shrubs, coarse woody debris, and litter) yielded aboveground carbon densities (mean ± standard error without Monte Carlo) as high as 320 ± 35 Mg ha^− 1 (old-growth coast redwood) and 510 ± 120 Mg ha^− 1 (red fir [Abies magnifica] forest), as great or greater than tropical rainforest. Lidar and QuickBird detected aboveground carbon in live trees, 70-97% of the total. Large sample sizes in the Monte Carlo analyses of remote sensing data generated low estimates of uncertainty. Lidar showed lower uncertainty and higher accuracy than QuickBird, due to high correlation of biomass to height and undercounting of trees by the crown detection algorithm. Lidar achieved uncertainties of < 1%, providing estimates of aboveground live tree carbon density (mean ± 95% confidence interval with Monte Carlo) of 82 ± 0.7 Mg ha^− 1 in Garcia-Mailliard and 140 ± 0.9 Mg ha^− 1 in North Yuba. The method that we tested, combining field measurements, Lidar, and Monte Carlo, can produce robust wall-to-wall spatial data on forest carbon.     

Measuring forest structure and biomass in New England forest stands using Echidna ground-based lidar

A ground-based, upward-scanning, near-infrared lidar, the Echidna® validation instrument (EVI), built by CSIRO Australia, retrieves forest stand structural parameters, including mean diameter at breast height (DBH), stem count density (stems/area), basal area, and above-ground woody biomass with very good accuracy in six New England hardwood and conifer forest stands. Comparing forest structural parameters retrieved using EVI data with extensive ground measurements, we found excellent agreement at the site level using five EVI scans (plots) per site (R² = 0.94-0.99); very good agreement at the plot level for stem count density and biomass (R² = 0.90-0.85); and good agreement at the plot level for mean DBH and basal area (R² = 0.48-0.66). The observed variance at site and plot levels suggest that a sample area of at least 1 ha (10⁴ m²) is required to estimate these parameters accurately at the stand level using either lidar-based or conventional methods. The algorithms and procedures used to retrieve these structural parameters are dependent on the unique ability of the Echidna® lidar to digitize the full waveform of the scattered lidar pulse as it returns to the instrument, which allows consistent separation of scattering by trunks and large branches from scattering by leaves. This successful application of ground-based lidar technology opens the door to rapid and accurate measurement of biomass and timber volume in areal sampling scenarios and as a calibration and validation tool for mapping biomass using airborne or spaceborne remotely sensed data.     

Assessment of QuickBird high spatial resolution imagery to detect red attack damage due to mountain pine beetle infestation

High spatial resolution remotely sensed data has the potential to complement existing forest health programs for both strategic planning over large areas, as well as for detailed and precise identification of tree crowns subject to stress and infestation. The area impacted by the current mountain pine beetle (Dendroctonus ponderosae Hopkins) outbreak in British Columbia, Canada, has increased 40-fold over the previous 5 years, with approximately 8.5 million ha of forest infested in 2005. As a result of the spatial extent and intensity of the outbreak, new technologies are being assessed to help detect, map, and monitor the damage caused by the beetle, and to inform mitigation of future beetle outbreaks. In this paper, we evaluate the capacity of high spatial resolution QuickBird multi-spectral imagery to detect mountain pine beetle red attack damage. ANOVA testing of individual spectral bands, as well as the Normalized Difference Vegetation Index (NDVI) and a ratio of red to green reflectance (Red-Green Index or RGI), indicated that the RGI was the most successful (p < 0.001) at separating non-attack crowns from red attack crowns. Based on this result, the RGI was subsequently used to develop a binary classification of red attack and non-attack pixels. The total number of QuickBird pixels classified as having red attack damage within a 50 m buffer of a known forest health survey point were compared to the number of red attack trees recorded at the time of the forest health survey. The relationship between the number of red attack pixels and observed red attack crowns was assessed using independent validation data and was found to be significant (r² = 0.48, p < 0.001, standard error = 2.8 crowns). A comparison of the number of QuickBird pixels classified as red attack, and a broader scale index of mountain pine beetle red attack damage (Enhanced Wetness Difference Index, calculated from a time series of Landsat imagery), was significant (r² = 0.61, p < 0.001, standard error = 1.3 crowns). These results suggest that high spatial resolution imagery, in particular QuickBird satellite imagery, has a valuable role to play in identifying tree crowns with red attack damage. This information could subsequently be used to augment existing detailed forest health surveys, calibrate synoptic estimates of red attack damage generated from overview surveys and/or coarse scale remotely sensed data, and facilitate the generation of value-added information products, such as estimates of timber volume impacts at the forest stand level.     

Distinguishing between live and dead standing tree biomass on the North Rim of Grand Canyon National Park, USA using small-footprint lidar data

Accurate estimation of live and dead biomass in forested ecosystems is important for studies of carbon dynamics, biodiversity, wildfire behavior, and for forest management. Lidar remote sensing has been used successfully to estimate live biomass, but studies focusing on dead biomass are rare. We used lidar data, in conjunction with field measurements from 58 plots to distinguish between and map standing live and dead tree biomass in the mixed coniferous forest of the North Rim of Grand Canyon National Park, USA. Lidar intensity and canopy volume were key variables for estimating live biomass, whereas for dead biomass, lidar intensity alone was critical for accurate estimation. Regression estimates of both live and dead biomass ranged between 0 and 600 Mg ha^− 1, with means of 195.08 Mg ha^− 1 and 65.73 Mg ha^− 1, respectively. Cross validation with field data resulted in correlation coefficients for predicted vs. observed of 0.85 for live biomass (RMSE = 50 Mg ha^− 1 and %RMSE (RMSE as a percent of the mean) = 26). For dead biomass, correlation was 0.79, RMSE was 42 Mg ha^− 1, and %RMSE was 63. Biomass maps revealed interesting patterns of live and dead standing tree biomass. Live biomass was highest in the ponderosa pine zone, and decreased from south to north through the mixed conifer and spruce-fir forest zones. Dead biomass exhibited a background range of values in these mature forests from zero to 100 Mg ha^− 1, with lower values in locations having higher live biomass. In areas with high dead biomass values, live biomass was near zero. These areas were associated with recent wildfires, as indicated by fire maps derived from the Monitoring Trends in Burn Severity Project (MTBS). Combining our dead biomass maps with the MTBS maps, we demonstrated the complementary power of these two datasets, revealing that MTBS burn intensity class can be described quantitatively in terms of dead biomass. Assuming a background range of dead biomass up to 100 Mg ha^− 1, it is possible to estimate and map the contribution to the standing dead tree biomass pool associated with recent wildfire.     

Forest canopy height from the Multiangle Imaging SpectroRadiometer (MISR) assessed with high resolution discrete return lidar

In this study retrievals of forest canopy height were obtained through adjustment of a simple geometric-optical (GO) model against red band surface bidirectional reflectance estimates from NASA's Multiangle Imaging SpectroRadiometer (MISR), mapped to a 250 m grid. The soil-understory background contribution was partly isolated prior to inversion using regression relationships with the isotropic, geometric, and volume scattering kernel weights of a Li-Ross kernel-driven bidirectional reflectance distribution function (BRDF) model. The height retrievals were assessed using discrete return lidar data acquired over sites in Colorado as part of the Cold Land Processes Experiment (CLPX) and used with fractional crown cover retrievals to obtain aboveground woody biomass estimates. For all model runs with reasonable backgrounds and initial b/r (vertical to horizontal crown radii) values < 2.0, root mean square error (RMSE) distributions were centered between 2.5 and 3.7 m while R² distributions were centered between 0.4 and 0.7. The MISR/GO aboveground biomass estimates predicted via regression on fractional cover and mean canopy height for the CLPX sites showed good agreement with U.S. Forest Service Interior West map data (adjusted R² = 0.84). The implication is that multiangle sensors such as MISR can provide spatially contiguous retrievals of forest canopy height, cover, and aboveground woody biomass that are potentially useful in mapping distributions of aboveground carbon stocks, tracking disturbance, and in initializing, constraining, and validating ecosystem models. This is important because the MISR record is spatially comprehensive and extends back to the year 2000 and the launch of the NASA Earth Observing System (EOS) Terra satellite; it might thus provide a ~ 10-year baseline record that would enhance exploitation of data from the NASA Deformation, Ecosystem Structure and Dynamics of Ice (DESDynI) mission, as well as furthering realization of synergies with active instruments.     

Estimating montane forest above-ground biomass in the upper reaches of the Heihe River Basin using Landsat-TM data

In this work, the results of above-ground biomass (AGB) estimates from Landsat Thematic Mapper 5 (TM) images and field data from the fragmented landscape of the upper reaches of the Heihe River Basin (HRB), located in the Qilian Mountains of Gansu province in northwest China, are presented. Estimates of AGB are relevant for sustainable forest management, monitoring global change, and carbon accounting. This is particularly true for the Qilian Mountains, which are a water resource protection zone. We combined forest inventory data from 133 plots with TM images and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) global digital elevation model (GDEM) V2 products (GDEM) in order to analyse the influence of the sun-canopy-sensor plus C (SCS+C) topographic correction on estimations of forest AGB using the stepwise multiple linear regression (SMLR) and k-nearest neighbour (k-NN) methods. For both methods, our results indicated that the SCS+C correction was necessary for getting more reliable forest AGB estimates within this complex terrain. Remotely sensed AGB estimates were validated against forest inventory data using the leave-one-out (LOO) method. An optimized k-NN method was designed by varying both mathematical formulation of the algorithm and remote-sensing data input, which resulted in 3000 different model configurations. Following topographic correction, performance of the optimized k-NN method was compared to that of the regression method. The optimized k-NN method (R² = 0.59, root mean square error (RMSE) = 24.92 tonnes ha^–1) was found to perform much better than the regression method (R² = 0.42, RMSE = 29.74 tonnes ha^–1) for forest AGB retrieval over this montane area. Our results indicated that the optimized k-NN method is capable of operational application to forest AGB estimates in regions where few inventory data are available.     

MODIS NDVI time-series allow the monitoring of Eucalyptus plantation biomass

The use of remote sensing is necessary for monitoring forest carbon stocks at large scales. Optical remote sensing, although not the most suitable technique for the direct estimation of stand biomass, offers the advantage of providing large temporal and spatial datasets. In particular, information on canopy structure is encompassed in stand reflectance time series. This study focused on the example of Eucalyptus forest plantations, which have recently attracted much attention as a result of their high expansion rate in many tropical countries. Stand scale time-series of Normalized Difference Vegetation Index (NDVI) were obtained from MODIS satellite data after a procedure involving un-mixing and interpolation, on about 15,000 ha of plantations in southern Brazil. The comparison of the planting date of the current rotation (and therefore the age of the stands) estimated from these time series with real values provided by the company showed that the root mean square error was 35.5 days. Age alone explained more than 82% of stand wood volume variability and 87% of stand dominant height variability. Age variables were combined with other variables derived from the NDVI time series and simple bioclimatic data by means of linear (Stepwise) or nonlinear (Random Forest) regressions. The nonlinear regressions gave r-square values of 0.90 for volume and 0.92 for dominant height, and an accuracy of about 25 m³/ha for volume (15% of the volume average value) and about 1.6 m for dominant height (8% of the height average value). The improvement including NDVI and bioclimatic data comes from the fact that the cumulative NDVI since planting date integrates the interannual variability of leaf area index (LAI), light interception by the foliage and growth due for example to variations of seasonal water stress. The accuracy of biomass and height predictions was strongly improved by using the NDVI integrated over the two first years after planting, which are critical for stand establishment. These results open perspectives for cost-effective monitoring of biomass at large scales in intensively-managed plantation forests.     

Augmented k-ary n-cubes

We define an interconnection network AQ_n,k which we call the augmented k-ary n-cube by extending a k-ary n-cube in a manner analogous to the existing extension of an n-dimensional hypercube to an n-dimensional augmented cube. We prove that the augmented k-ary n-cube AQ_n,k has a number of attractive properties (in the context of parallel computing). For example, we show that the augmented k-ary n-cube AQ_n,k: is a Cayley graph, and so is vertex-symmetric, but not edge-symmetric unless n = 2; has connectivity 4n − 2 and wide-diameter at most max{(n − 1)k − (n − 2), k + 7}; has diameter , when n = 2; and has diameter at most , for n ? 3 and k even, and at most , for n ? 3 and k odd.     

Locally linear reconstruction for instance-based learning

Instance-based learning (IBL), so called memory-based reasoning (MBR), is a commonly used non-parametric learning algorithm. k-nearest neighbor (k-NN) learning is the most popular realization of IBL. Due to its usability and adaptability, k-NN has been successfully applied to a wide range of applications. However, in practice, one has to set important model parameters only empirically: the number of neighbors (k) and weights to those neighbors. In this paper, we propose structured ways to set these parameters, based on locally linear reconstruction (LLR). We then employed sequential minimal optimization (SMO) for solving quadratic programming step involved in LLR for classification to reduce the computational complexity. Experimental results from 11 classification and eight regression tasks were promising enough to merit further investigation: not only did LLR outperform the conventional weight allocation methods without much additional computational cost, but also LLR was found to be robust to the change of k.     

Estimating areal means and variances of forest attributes using the k-Nearest Neighbors technique and satellite imagery

The k-Nearest Neighbor (k-NN) technique has become extremely popular for a variety of forest inventory mapping and estimation applications. Much of this popularity may be attributed to the non-parametric, multivariate features of the technique, its intuitiveness, and its ease of use. When used with satellite imagery and forest inventory plot data, the technique has been shown to produce useful estimates of many forest attributes including forest/non-forest, volume, and basal area. However, variance estimators for quantifying the uncertainty of means or sums of k-NN pixel-level predictions for areas of interest (AOI) consisting of multiple pixels have not been reported. The primary objectives of the study were to derive variance estimators for AOI estimates obtained from k-NN predictions and to compare precision estimates resulting from different approaches to k-NN prediction and different interpretations of those predictions. The approaches were illustrated by estimating proportion forest area, tree volume per unit area, tree basal area per unit area, and tree density per unit area for 10-km AOIs. Estimates obtained using k-NN approaches and traditional inventory approaches were compared and found to be similar. Further, variance estimates based on different interpretations of k-NN predictions were similar. The results facilitate small area estimation and simultaneous and consistent mapping and estimation of multiple forest attributes.     

A direct boosting algorithm for the k-nearest neighbor classifier via local warping of the distance metric

Though the k-nearest neighbor (k-NN) pattern classifier is an effective learning algorithm, it can result in large model sizes. To compensate, a number of variant algorithms have been developed that condense the model size of the k-NN classifier at the expense of accuracy. To increase the accuracy of these condensed models, we present a direct boosting algorithm for the k-NN classifier that creates an ensemble of models with locally modified distance weighting. An empirical study conducted on 10 standard databases from the UCI repository shows that this new Boosted k-NN algorithm has increased generalization accuracy in the majority of the datasets and never performs worse than standard k-NN.