Henrik J. Persson

Forest Variable Estimation Using Radargrammetric Processing of TerraSAR-X Images in Boreal Forests

The last decade has seen launches of radar satellite missions operating in X-band with the sensors acquiring images with spatial resolutions on the order of 1 m. This study uses digital surface models (DSMs) extracted from stereo synthetic aperture radar images and a reference airborne laser scanning digital terrain model to calculate the above-ground biomass and tree height. The resulting values are compared to in situ data. Analyses were undertaken at the Swedish test sites Krycklan (64 degrees N) and Remningstorp (58 degrees N), which have different site conditions. The results showed that, for 459 forest stands in Remningstorp, biomass estimation at the stand level could be performed with 22.9% relative root mean square error, while the height estimation showed 9.4%. Many factors influenced the results and it was found that the topography has a significant effect on the generated DSMs and should therefore be taken into consideration when the stand level mean slope is above four degrees. Different tree species did not have a major effect on the models during leaf-on conditions. Moreover, correct estimation within young forest stands was problematic. The intersection angles resulting in the best results were in the range 8-16 degrees. Based on the results in this study, radargrammetry appears to be a promising potential remote sensing technique for future forest applications.

Estimating forest biomass and height using optical stereo satellite data and a DTM from laser scanning data

This paper investigates the possibilities of improving aboveground forest biomass and basal area weighted mean height estimates from optical multispectral satellite data by adding Canopy Height Models (CHMs) obtained from matching multiple view-angle satellite data. The analysis was carried out using data collected over the Remningstorp test site in southern Sweden from 2008–2011 and used training and validation data from airborne laser scanning and field data. CHMs were produced by subtracting a Digital Elevation Model (based on airborne laser scanning data) from the Digital Surface Models created by matching multiple view-angle SPOT-5 HRS and ALOS PRISM data. By modeling biomass and height using regression analysis on multispectral data from SPOT-5 HRG in combination with height metrics from the CHMs, an improved root mean squared error (RMSE) was attained, compared with using the individual satellite data sources alone. A comparison between SPOT-5 HRS and ALOS PRISM CHMs in combination with multispectral data was made at stand level using biomass and height estimates from laser scanning data as reference data. For biomass, the relative RMSE improved from 32.9% when using only multispectral data, to 29.2% and 22.4% when adding the CHM from SPOT-5 HRS and ALOS PRISM, respectively. The corresponding improvements for height were from 16.1% to 15.3% with the SPOT-5 HRS CHM and to 12.9% with the ALOS PRISM CHM. A further analysis of combining the ALOS PRISM CHM and multispectral data was made at sub-stand level with field measurements as reference data. This combination gave a relative RMSE of 20.6% for biomass and 10.5% for height. In conclusion, the estimation accuracy for aboveground biomass and basal area weighted mean height was improved by adding CHM data to multispectral data from optical satellites.

Comparison between TanDEM-X- and ALS-based estimation of aboveground biomass and tree height in boreal forests

ABSTRACT Interferometric Synthetic Aperture Radar (InSAR) data from TerraSAR-X add-on for Digital Elevation Measurement (TanDEM-X) were used to estimate aboveground biomass (AGB) and tree height with linear regression models. These were compared to models based on airborne laser scanning (ALS) data at two Swedish boreal forest test sites, Krycklan (64°N19°E) and Remningstorp (58°N13°E). The predictions were validated using field data at the stand-level (0.5–26.1 ha) and at the plot-level (10 m radius). Additionally, the ALS metrics percentile 99 (p99) and vegetation ratio, commonly used to estimate AGB and tree height, were estimated in order to investigate the feasibility of replacing ALS data with TanDEM-X InSAR data. Both AGB and tree height could be estimated with about the same accuracy at the stand-level from both TanDEM-X- and ALS-based data. The AGB was estimated with 17.2% and 14.6% root mean square error (RMSE) and the tree height with 7.6% and 4.1% RMSE from TanDEM-X data at the stand-level at the two test sites Krycklan and Remningstorp. The Pearson correlation coefficients between the TanDEM-X height and the ALS height p99 were r = .98 and r = .95 at the two test sites. The TanDEM-X height contains information related to both tree height and forest density, which was validated from several estimation models.

Assessment of boreal forest height from WorldView-2 satellite stereo images

ABSTRACT WorldView-2 (WV2) satellite stereo images were used to derive a digital surface model, which together with a high-resolution digital terrain model from airborne laser scanning (ALS) were used to estimate forest height. Lorey’s mean height (HL) could be estimated with a root mean square error of 1.5 m (8.3%) and 1.4 m (10.4%), using linear regression, at the two Swedish test sites Remningstorp (Lat. 58°30ʹN, Long. 13°40ʹE) and Krycklan (Lat. 64°16ʹN, Long. 19°46ʹE), which contain hemi-boreal and boreal forest. The correlation coefficients were r = 0.94 and r = 0.91, respectively. The 10 m sample plots were 175 in Remningstorp and 282 in Krycklan. It was furthermore found that WV2 data are sometimes unstable for canopy top height estimations (ALS height percentile 100, p100) and that the reconstructed heights are generally located below the actual top height. The WV2 p60 was found to correlate best with ALS p70 in Remningstorp, while WV2 p95 was found to correlate best with ALS p70 in Krycklan, and it moreover reached the highest correlation for all other estimated variables, at both test sites. It was concluded that WV2 p95 height data overall represent approximately the forest height ALS p70. The overall high correlation coefficients above 0.90 at both test sites, with different forest conditions, indicate that stereo matching of WV2 satellite images is suitable for forest height mapping.

Estimation of Boreal Forest Attributes from Very High Resolution Pléiades Data

In this study, the potential of using very high resolution Pleiades imagery to estimate a number of common forest attributes for 10-m plots in boreal forest was examined, when a high-resolution terrain model was available. The explanatory variables were derived from three processing alternatives. Height metrics were extracted from image matching of the images acquired from different incidence angles. Spectral derivatives were derived by performing principal component analysis of the spectral bands and lastly, second order textural metrics were extracted from a gray-level co-occurrence matrix, computed with an 11 × 11 pixels moving window. The analysis took place at two Swedish test sites, Krycklan and Remningstorp, containing boreal and hemi-boreal forest. The lowest RMSE was estimated with 1.4 m (7.7%) for Lorey’s mean height, 1.7 m (10%) for airborne laser scanning height percentile 90, 5.1 m2·ha−1 (22%) for basal area, 66 m3·ha−1 (27%) for stem volume, and 26 tons·ha−1 (26%) for above-ground biomass, respectively. It was found that the image-matched height metrics were most important in all models, and that the spectral and textural metrics contained similar information. Nevertheless, the best estimations were obtained when all three explanatory sources were used. To conclude, image-matched height metrics should be prioritised over spectral metrics when estimation of forest attributes is concerned.

Estimation of boreal forest biomass from two-level model inversion of interferometric TanDEM-X data

A new model for aboveground biomass estimation from forest height and canopy density estimates obtained from the inversion of a two-level model (TLM) is presented and studied using data from the hemi-boreal test site Remningstorp, situated in southern Sweden. Three bistatic-interferometric TanDEM-X acquisitions from the summers of 2011, 2012, and 2013 and with heights-of-ambiguity (HOAs) 49 m, 32 m, and 63 m, respectively, are used. An external, high-resolution digital terrain model (DTM) is used as ground reference during interferogram flattening. Model parameters are estimated for each acquisition separately, and the model is evaluated on all three acquisitions, to examine both its explanatory and predictive values. Residual root-mean-square errors (RMSEs) are 14%-19% and the model explains 67%-84% of the variance in the data. Prediction RMSE is 20% for the two images with the highest HOA, but much higher for the third image.

Detection of thinning and clear-cuts using TanDEM-X data

Interferometric TanDEM-X data from 2011 and 2014 were used to create biomass maps over the Swedish test site Remningstorp. These maps were used to compute the biomass change for four classes; pre-commercial thinning, thinning, clear-cutting, and untouched forest. Field inventory and ALS data from the corresponding years were used as reference data. The biomass change was compared on 12 subjectively chosen plots with 40 m radius for each class. It was found, that pre-commercial thinning was difficult to detect, as the biomass loss was less than the biomass growth during the four vegetation seasons investigated. Thinning could be detected from the biomass change being about zero or slightly negative, while clear-cut plots were obvious to notice, with the biomass withdrawal being several hundreds of tons ha-1. The untouched plots had a biomass growth of about 4 to 6 tons ha-1 year-1. It was concluded, that annual TanDEM-X images can be used to detect also smaller silviculture activities such as thinning, but further research with shorter time periods would be desired.

Detection of forest change and robust estimation of forest height from two-level model inversion of multi-temporal, single-pass InSAR data

In this paper, forest change detection and forest height estimation are studied using two-level model (TLM) inversion of multi-temporal TanDEM-X (TDM) data. Parameter Ah, describing the distance between ground and vegetation levels, is kept constant for all acquisitions, whereas parameter ¡jl, the area-weighted backscatter ratio, changes with acquisition. Two multi-temporal sets of TDM data, acquired over the hemi-boreal test site Remningstorp, situated in southern Sweden, are studied: one consisting of 12 acquisitions made in the summers of 2011, 2012, 2013, and 2014 with heights-of-ambiguity (HOAs) between 32 m and 63 m, and one consisting of 33 acquisitions made between August 2013 and August 2014 with HOAs between 38 m and 195 m. The first dataset is used to show that commercial thinnings and clear-cuts can be detected by studying the canopy density estimate r/o = 1/(1 + /x). The second dataset is used to show that seasonal change can be observed in r/o for deciduous plots, but not for coniferous plots. Moreover, it is shown that 1.3Ah is a good estimate of the basal area-weighted (Loreys) height, with a correlation coefficient equal to 0.98 and a root-mean-square error of 0.9 m.

Experiences from Large-Scale Forest Mapping of Sweden Using TanDEM-X Data

This paper report experiences from the processing and mosaicking of 518 TanDEM-X image pairs covering the entirety of Sweden, with two single map products of above-ground biomass (AGB) and forest stem volume (VOL), both with 10 m resolution. The main objective was to explore the possibilities and overcome the challenges related to forest mapping extending a large number of adjacent satellite scenes. Hence, numerous examples are presented to illustrate challenges and possible solutions. To derive the forest maps, the observables backscatter, interferometric phase height and interferometric coherence, obtained from TanDEM-X, were evaluated using empirical robust linear regression models with reference data extracted from 2288 national forest inventory plots with a 10 m radius. The interferometric phase height was the single most important observable, to predict AGB and VOL. The mosaics were evaluated on different datasets with field-inventoried stands across Sweden. The root mean square error (RMSE) was about 21%–25% (27–30 tons/ha and 52–65 m3/ha) at the stand level. It was noted that the most influencing factors on the observables in this study were local temperature and geolocation errors that were challenging to robustly compensate against. Because of this variability at the scene-level, determinations of AGB and VOL for single stands are recommended to be used with care, as an equivalent accuracy is difficult to achieve for all different scenes, with varying acquisition conditions. Still, for the evaluated stands, the mosaics were of sufficient accuracy to be used for forest management at the stand level.

Improved Prediction of Forest Variables Using Data Assimilation of Interferometric Synthetic Aperture Radar Data

ABSTRACT The statistical framework of data assimilation provides methods for utilizing new data for obtaining up-to-date forest data: existing forest data are forecasted and combined with each new remote sensing data set. This new paradigm for updating forest database, well known from other fields of study, will provide a framework for utilizing all available remote sensing data in proportion to their quality to improve prediction. It also solves the problem that not all remote sensing data sets provide information for the entire area of interest, since areas with no remote sensing data can be forecasted until new remote sensing data become available. In this study, extended Kalman filtering was used for assimilating data from 19 TanDEM-X InSAR images on 137 sample plots, each of 10-meter radius at a test site in southern Sweden over a period of 4 years. At almost all time points data assimilation resulted in predictions closer to the reference value than predictions based on data from that single time point. For the study variables Loreys mean height, basal area, and stem volume, the median reduction in root mean square error was 0.4 m, 0.9 m2/ha, and 15.3 m3/ha (2, 3, and 6 percentage points), respectively.