Lauri Kurvonen

Retrieval of biomass in boreal forests from multitemporal ERS-1 and JERS-1 SAR images

The response of JERS-1 and ERS-1 synthetic aperture radar (SAR) to the forest stem volume (biomass) was investigated by employing a digital stem volume map and weather information. The stem volume map was produced from the National Forest Inventory sample plot data together with a LANDSAT thematic mapper (TM) image. A new indirect inversion method was developed and tested to estimate the forest blockwise stem volume from JERS-1 and/or ERS-1 SAR images. The method is based on using a semiempirical backscatter model for inversion. The model presumes that backscatter from a forest canopy is determined by the stem volume, soil moisture, and vegetation moisture. The area of interest is divided into a training and test area. In this study, the training area was 10% of the test site, while the remaining 90% was used for testing the method. The inversion algorithm is carried out in the following three steps. 1) For the training area, the soil and vegetation moisture parameters are estimated from the backscattering coefficients and stem volume (must be known for training areas) with the semiempirical backscatter model. 2) For the area of interest, the stem volume is estimated from the moisture parameters and backscattering coefficients with the semiempirical backscattering model. 3) If several SAR images are used, the stem volume estimates are combined with a multiple linear regression. The regression equation is defined using the stem volume estimates for the training area. The results for the stem volume estimation using L-band and/or C-band SAR data showed promising accuracies: the relative retrieval rms error varied from 30 to 5% as the size of the forest area varied from 5 to 30000 ha.

Multitemporal behavior of L- and C-band SAR observations of boreal forests

An analysis of L- and C-band boreal forest backscattering properties with respect to various temporally changing parameters is presented. The seasonal and weather dependent parameters considered include the depth of soil frost, topsoil moisture, snow water equivalent, air temperature and precipitation. The effect of these parameters on /spl sigma//spl deg/ are studied for various stem volume (biomass) classes by comparing the results against a cloud model-based semi-empirical modeling approach. Semi-empirical modeling is also used for a forest biomass retrieval experiment. The SAR data set includes 4 JERS-1 (L-band, HH-polarization) and 19 ERS-1 (C-band, VV-polarization) images for a test area in southern Finland. Additionally, a set of 2 JERS-1 and 3 ERS-1 images for another test area in northern Finland is employed. The results show that radar response to forest biomass is more sensitive to changes in temporally varying parameters at C-band than at L-band. The semi-empirical modeling approach describes well the behavior of /spl sigma//spl deg/ at both frequency bands when large forest areas are considered. Moreover, the modeling approach appears to be applicable for different conifer-dominated boreal forest types. Since the modeling approach explains satisfactorily the average backscattering behavior, the results in biomass retrieval show high accuracies (25-30% relative RMSE) when areas under investigation are large enough, i.e. about 20 ha.

Influence of land-cover category on brightness temperature of snow

A helicopter-borne multifrequency radiometer (24, 34, 48 and 94 GHz vertical polarization) was used to investigate the behavior of the brightness temperature of snow in Sodankyla (latitude: 67.41 N, longitude: 26.58 E), Northern Finland. The measurements were carried out during dry snow, wet snow, and snow-free conditions. The angle of incidence was 45/spl deg/ in all measurements. The measurements and the main results are presented. The analysis is focused on the effect of vegetation and land type on the brightness temperature of snow. The main topics of this paper are: (a) the general behavior of the brightness temperature of snow for different land types, (b) the effect of forest vegetation on the brightness temperature of snow, and (c) the capability of the radiometer system to monitor snow extent in forests during the melting period.

Textural information of multitemporal ERS-1 and JERS-1 SAR images with applications to land and forest type classification in boreal zone

The textural information of a multitemporal set of ERS-1 and JERS-1 synthetic aperture radar (SAR) images was studied with the first- and second-order statistical measures. These measures had a higher information value for the land-cover and forest type classification than the SAR image intensity. The multitemporal approach was beneficial for the application of the textural measures; the textural parameters significantly improved the classification of land-cover and forest types. Based on the SAR image texture, the overall classification accuracy for seven land-cover types was 65%, while with the SAR image intensity, the classification accuracy was 50%, respectively. In the forest type classification based on the SAR image texture and intensity, the overall classification accuracy for four forest types was 66%, while with the intensity, the accuracy was 40%, respectively. The weather and seasonal conditions had a significant effect on the textural information of SAR images. The best separability of the signatures and the best land-cover and forest type classification accuracy was achieved under summer conditions. The snow cover and arid conditions decreased the textural information of the SAR images.

Capability of radar and microwave radiometer to classify snow types in forested areas

The authors have organized several airborne campaigns with their dual-frequency scatterometer and multifrequency radiometer under various snow conditions. These data are complemented with ERS-1 SAR images of the same test site. The effect of land use categories (especially forests) to the snow mapping capability of radar and radiometer has been investigated.<<ETX>>

Monitoring of boreal forests with multitemporal special sensor microwave imager data

The feasibility of multitemporal special sensor microwave imager (SSM/I) data for monitoring boreal forests was evaluated. The parameters of interest were forest coverage fraction and forest stem volume (biomass). The employed test sites covered almost the whole of Finland. Two measurement periods were used: July through September in 1993 and January through February in 1994. The apparent emissivities of various land cover types were determined with the mixed pixel approach under summer and winter conditions. The aim was to define the dominating factors on emissivity under winter and summer conditions. The mixed pixel approach was tested for the estimation of forest coverage fraction. The results with multitemporal SSM/I data show that the pixel-wise fractions of water, nonforested, and forested area can be estimated with rms errors of around 10 percent units. The correlation between the estimates and the ground truth was over 0.85. A new inversion method for stem volume estimation was presented. At SSM/I frequencies the forest canopy and the snow-covered ground dominate the emissivity behavior under winter conditions in the boreal forest zone. The method is based on the fact that the emissivity of forest canopy is close to 1, while that of dry snow cover is relatively low. The results with the method showed promising accuracies when wintertime SSM/I data were employed. The rms error was from 13 to 19 m3/ha per pixel (25 km by 25 km), which was 15–16% of the mean stem volume. In the test area the stem volume ranged from 40 to 160 m3/ha per pixel.

Classification of Baltic Sea ice types by airborne multifrequency microwave radiometer

An airborne multifrequency radiometer (24, 34, 48, and 94 GHz, vertical polarization) was used to investigate the behavior of the brightness temperature of different sea ice types in the Gulf of Bothnia (Baltic Sea). The measurements and the main results of the analysis are presented. The measurements were made in dry and wet conditions (air temperature above and below 0/spl deg/C). The angle of incidence was 45/spl deg/ in all measurements. The following topics are evaluated: a) frequency dependency of the brightness temperature of different ice types, b) the capability of the multifrequency radiometer to classify ice types for winter navigation purposes, and c) the optimum measurement frequencies for mapping sea ice. The weather conditions had a significant impact on the radiometric signatures of some ice types (snow-covered compact pack ice and frost-covered new ice); the impact was the highest at 94 GHz. In all cases the overall classification accuracy was around 90% (the kappa coefficient was from 0.86 to 0.96) when the optimum channel combination (24/34 GHz and 94 GHz) was used.

Active and passive microwave remote sensing of boreal forests

Abstract Novel inversion methods are presented for active and passive satelliteborne microwave remote sensing. The objectives are biomass estimation, forest and land-cover-type recognition in boreal forests. A new adaptive inversion method for active sensors was developed for the forest block-wise stem volume estimation from satelliteborne radar images (e.g. JERS-1, ERS-1 SAR and RADARSAT). The inversion results with L-band and/or C-band synthetic aperture radar (SAR) images showed promising accuracies: the relative retrieval rms error varied from 25% to 5% as the size of the forest area varied from 5 to 30 000 h (the forest stem volume varied from 0 to 300 m 3 / ha ) . The textural information of a seasonal set of satelliteborne radar images was studied with the first- and second-order statistical measures. The multitemporal approach was beneficial for the textural measures in forest and land-cover-type recognition. Based on the SAR image texture, the overall classification accuracy for seven land-cover types was 65%, while with the SAR image intensity, the classification accuracy was 50%, respectively. In the forest-type classification based on the SAR image texture and intensity, the overall classification accuracy for four forest types was 66%, while with the intensity alone the accuracy was 40%, respectively. With the passive microwave sensor (e.g. satelliteborne SSM/I radiometer), the mixed pixel approach was employed for stem volume (biomass) and forest coverage fraction estimation. The results obtained, show that the pixel-wise fractions of water, non-forested, and forested area can be estimated with a rms errors of around 10% units. A new stem volume inversion method for wintertime SSM/I data achieved promising accuracies, the rms error was from 13 to 19 m 3 / ha / pixel (25 km ×25 km ) which was 15–16% of the mean stem volume. In the test area, the stem volume ranged from 40 to 160 m 3 / ha / pixel .

Retrieval of forest parameters from multitemporal spaceborne SAR data

The aim of the study is to develop inversion methods for spaceborne SAR images of the boreal forests. The parameters of interest are stem volume (biomass), forest and land-cover types, soil and vegetation moisture. The previously developed forest stem volume/soil moisture retrieval algorithm was tested by producing soil moisture maps for large test areas. The textural information of a seasonal set of ERS-1 SAR images was studied with the first and second order statistical measures. These measures had a higher information value than intensity values (or their principal components) for the forest and land type classification. The multitemporal approach was beneficial for the application of textural measures and the textural parameters significantly improved the classification of land-use and forest types.

Effect of temporally varying parameters on L- and C-band SAR observations of boreal forests

An analysis of L- and C-band boreal forest backscattering properties with respect to various temporally changing parameters is presented. The seasonal or weather dependent parameters investigated include the depth of soil frost, top-soil moisture, snow water equivalent, air temperature and precipitation. The effect of these parameters on /spl sigma//spl deg/ are studied for various stem volume (biomass) classes by comparing the results against a semi-empirical modelling approach. Semi-empirical modelling is also used for a forest biomass retrieval experiment. The SAR data set includes 4 JERS-1 (L-band, HH-polarization) and 19 ERS-1 (C-band, VV-polarization) images for a single test area in southern Finland. The results show that at C-band response to forest biomass is more sensitive to changes in temporally varying parameters than at L-band. The semi-empirical modelling approach was found to describe well the behavior of /spl sigma//spl deg/ when large areas are considered. Hence, the results in biomass retrieval show high accuracies (25-30% relative accuracy) when areas under investigation are larger than 20 ha.