Multi-source Forest Inventory Research Articles

Multisource forest inventories: A model-based approach using k-NN to reconcile forest attributes statistics and map products

Forest map products are widely used and have taken benefit from progresses in the multisource forest inventory approaches, which are meant to improve the precision of forest inventory estimates at high spatial resolution. However, estimating errors of pixel-wise predictions remains difficult, and reconciling statistical outcomes with map products is still an open and important question.We address this problem using an original approach relying on a model-based inference framework and k-nearest neighbours (k-NN) models to produce pixel-wise estimations and related quality assessment. Our approach takes advantage of the resampling properties of a model-based estimator and combines it with geometrical convex-hull models to measure respectively the precision and accuracy of pixel predictions. A measure of pixel reliability was obtained by combining precision and accuracy.The study was carried out over a 7,694 km2 area dominated by structurally complex broadleaved forests in centre of France. The targeted forest attributes were growing stock volume, basal area and growing stock volume increment. A total of 819 national forest inventory plots were combined with auxiliary data extracted from a forest map, Landsat 8 images, and 3D point clouds from both airborne laser scanning and digital aerial photogrammetry. k-NN models were built independently for both 3D data sources. Both selected models included 5 auxiliary variables, and were generated using 5 neighbours, and most similar neighbours distance measure. The models showed relative root mean square error ranging from 35.7% (basal area, digital aerial photogrammetry) in calibration to 63.4% (growing stock volume increment, airborne laser scanning) in the validation set. At pixel level, we found that a minimum of 86.4% of the predictions were of high precision as their bootstrapped coefficient of variation fall below calibration’s relative root mean square error. The amount of extrapolation varied from 4.3% (digital aerial photogrammetry) to 6.3% (airborne laser scanning). A relationship was found between extrapolation and k-NN distance, opening new opportunities to correct extrapolation errors. At the population level, airborne laser scanning and digital aerial photogrammetry performed similarly, offering the possibility to use digital aerial photogrammetry for monitoring purposes.The proposed method provided consistent estimates of forest attributes and maps, and also provided spatially explicit information about pixel predictions in terms of precision, accuracy and reliability. The method therefore produced high resolution outputs, significant for either decision making or forest management purposes.

Mapping Floods in Lowland Forest Using Sentinel-1 and Sentinel-2 Data and an Object-Based Approach

The impact of floods on forests is immediate, so it is necessary to quickly define the boundaries of flooded areas. Determining the extent of flooding in situ has shortcomings due to the possible limited spatial and temporal resolutions of data and the cost of data collection. Therefore, this research focused on flood mapping using geospatial data and remote sensing. The research area is located in the central part of the Republic of Croatia, an environmentally diverse area of lowland forests of the Sava River and its tributaries. Flood mapping was performed by merging Sentinel-1 (S1) and Sentinel-2 (S2) mission data and applying object-based image analysis (OBIA). For this purpose, synthetic aperture radar (SAR) data (GRD processing level) were primarily used during the flood period due to the possibility of all-day imaging in all weather conditions and flood detection under the density of canopy. The pre-flood S2 imagery, a summer acquisition, was used as a source of additional spectral data. Geographical information system (GIS) layers—a multisource forest inventory, habitat map, and flood hazard map—were used as additional sources of information in assessing the accuracy of and interpreting the obtained results. The spectral signature, geometric and textural features, and vegetation indices were applied in the OBIA process. The result of the work was a developed methodological framework with a high accuracy and speed of production. The overall accuracy of the classification is 94.94%. Based on the conducted research, the usefulness of the C band of the S1 in flood mapping in lowland forests in the leaf-off season was determined. The paper presents previous research and describes the SAR parameters and characteristics of floodplain forest with a significant impact on the accuracy of classification.

Reducing error in small-area estimates of multi-source forest inventory by multi-temporal data fusion

Abstract Since the 1990s, forest resource maps and forest variable estimates for small areas have been produced by combining national forest inventory (NFI) field plot data, optical satellite images and numerical map data. A non-parametric $k$-NN method has frequently been employed. In Finland, such multi-source NFI (MS-NFI) forest variable estimates for municipalities have been produced eight times. A relatively large variation has been observed between subsequent estimates. In this study, a large-scale evaluation of small-area estimates from an MS-NFI conducted in 2013 was carried out in comparison with pure NFI field data-based estimates and error estimates. The proportion of municipalities with significant differences was larger than expected, e.g. over 10% for the mean volume, which indicates systematic error in the small-area estimates. A multi-temporal data fusion combining MS-NFI estimators from three time points—2011, 2013 and 2015—was tested as a means to improve single time point MS-NFI estimates of the mean volumes of growing stock and of tree species groups. A generalized least squares (GLS) technique and unweighted averaging were tested. The improvement was small but consistent when validated against the NFI field data-based estimates for the municipalities. The unweighted averaging worked nearly as well as a GLS estimator.

Ecosystem services of boreal forests – Carbon budget mapping at high resolution

The carbon (C) cycle of forests produces ecosystem services (ES) such as climate regulation and timber production. Mapping these ES using simple land cover -based proxies might add remarkable inaccuracy to the estimates. A framework to map the current status of the C budget of boreal forested landscapes was developed. The C stocks of biomass and soil and the annual change in these stocks were quantified in a 20 × 20 m resolution at the regional level on mineral soils in southern Finland. The fine-scale variation of the estimates was analyzed geo-statistically. The reliability of the estimates was evaluated by comparing them to measurements from the national multi-source forest inventory. The C stocks of forests increased slightly from the south coast to inland whereas the changes in these stocks were more uniform. The spatial patches of C stocks were larger than those of C stock changes. The patch size of the C stocks reflected the spatial variation in the environmental conditions, and that of the C stock changes the typical area of forest management compartments. The simulated estimates agreed well with the measurements indicating a good mapping framework performance. The mapping framework is the basis for evaluating the effects of forest management alternatives on C budget at high resolution across large spatial scales. It will be coupled with the assessment of other ES and biodiversity to study their relationships. The framework integrated a wide suite of simulation models and extensive inventory data. It provided reliable estimates of the human influence on C cycle in forested landscapes.

Utilizing a Multi-Source Forest Inventory Technique, MODIS Data and Landsat TM Images in the Production of Forest Cover and Volume Maps for the Terai Physiographic Zone in Nepal

An approach based on the nearest neighbors techniques is presented for producing thematic maps of forest cover (forest/non-forest) and total stand volume for the Terai region in southern Nepal. To create the forest cover map, we used a combination of Landsat TM satellite data and visual interpretation data, i.e., a sample grid of visual interpretation plots for which we obtained the land use classification according to the FAO standard. These visual interpretation plots together with the field plots for volume mapping originate from an operative forest inventory project, i.e., the Forest Resource Assessment of Nepal (FRA Nepal) project. The field plots were also used in checking the classification accuracy. MODIS satellite data were used as a reference in a local correction approach conducted for the relative calibration of Landsat TM images. This study applied a non-parametric k-nearest neighbor technique (k-NN) to the forest cover and volume mapping. A tree height prediction approach based on a nonlinear, mixed-effects (NLME) modeling procedure is presented in the Appendix. The MODIS image data performed well as reference data for the calibration approach applied to make the Landsat image mosaic. The agreement between the forest cover map and the field observed values of forest cover was substantial in Western Terai (KHAT 0.745) and strong in Eastern Terai (KHAT 0.825). The forest cover and volume maps that were estimated using the k-NN method and the inventory data from the FRA Nepal project are already appropriate and valuable data for research purposes and for the planning of forthcoming forest inventories. Adaptation of the methods and techniques was carried out using Open Source software tools.

Integrating multi-source forest inventory and animal survey data to assess nationwide distribution and habitat correlates of the Siberian flying squirrel

Modern intensive forestry worldwide is posing serious threats to old-growth forest specialist species. Cost-effective conservation can only be achieved when the ecology and distribution of species are well understood. The present distribution of the Siberian flying squirrel (Pteromys volans), a vulnerable species in Finland (and Europe), is poorly known, and its ecology studied only locally. We conducted a nationwide survey of the species to assess its distribution and derived habitat variables from the National Forest Inventory dataset to study its habitat requirements. We performed separate logistic regression models of the species occurrence at two separate spatial scales, small (9ha plot) and landscape (1km radius) extent. At the small scale, occurrence increased with average forest age and volume of spruce trees. At the landscape scale, occurrence was positively related to cover of mature spruce/mixed forests, density of suitable habitat patches and, rather unexpectedly, agricultural fields. Occurrence of flying squirrel was generally higher in restricted regions of central-western Finland, where the forested landscape is more fragmented by agricultural fields, compared to the rest of the country. Our findings indicate that the species can make use of fragmented habitat mosaics including forests and the interface between forests and fields. Conserving spruce dominated mature stands in these landscape mosaics is a key to maintaining adequate habitat for the species. We conclude that conservation efforts should also be directed to the most highly man-modified landscapes, where the preservation of habitat patches of good quality for the species should be integrated within the forest management plans.

Mapping biomass variables with a multi-source forest inventory technique

Map form information on forest biomass is required for estimating bioenergy potentials and monitoring carbon stocks. In Finland, the growing stock of forests is monitored using multi-source forest inventory, where variables are estimated in the form of thematic maps and area statistics by combining information of field measurements, satellite images and other digital map data. In this study, we used the multi-source forest inventory methodology for estimating forest biomass characteristics. The biomass variables were estimated for national forest inventory field plots on the basis of measured tree variables. The plot-level biomass estimates were used as reference data for satellite image interpretation. The estimates produced by satellite image interpretation were tested by cross-validation. The results indicate that the method for producing biomass maps on the basis of biomass models and satellite image interpretation is operationally feasible. Furthermore, the accuracy of the estimates of biomass variables is similar or even higher than that of traditional growing stock volume estimates. The technique presented here can be applied, for example, in estimating biomass resources or in the inventory of greenhouse gases.

Data Combination and Feature Selection for Multi-source Forest Inventory

Both satellite images and aerial photographs are now used operationally in Finland’s forestry for different tasks; satellite images are used for national forest inventory purposes and aerial images for forest management planning. Due to the double coverage, it could be advantageous to utilize the strengths of both image types. The aim of this study was to evaluate the potential of

Smoothing methodology for predicting regional averages in multi-source forest inventory

The paper examines alternative non-parametric estimation methods or smoothing methods in the context of the Finnish multi-source forest inventory. It uses satellite images in addition to field data to produce forest variable predictions for regions ranging from the single pixel level up to the national level. With the help of the bias-variance decomposition, the influence of the smoothing parameters on prediction accuracy is considered when the smoother's pixel-level predictions are averaged in order to produce predictions for larger areas. A novel variation of cross-validation, called region-wise cross-validation, is proposed for selecting the smoothing parameters. Experimental results are presented using local linear ridge regression (LLRR), which is a variant of the better known local linear regression method.

Accuracy of forest mapping based on Landsat TM data and a kNN-based method

A multi-source forest inventory (MSFI) method has been developed for use in the Norwegian National Forest Inventory (NFI). The method is based on a k-nearest neighbour rule and uses field plots from the NFI, land cover maps, and satellite image data from Landsat Thematic Mapper. The inventory method is used to produce maps of selected forest variables and to estimate the selected forest variables for large areas such as municipalities. In this study, focus has been on the qualitative variables ‘dominating species group’ and ‘development class’ because these variables are of central interest to forest managers. A mid-summer Landsat 5 TM scene was used as image data, and all NFI plots inside the scene were used as a reference dataset. The relationship between the spectral bands and the forest variables was analysed, and it was found that the levels of association were low. A leave-one-out method based on the reference dataset was used to estimate the pixel-level accuracies. They were found to be relatively low with 63% agreement for species groups. An independent control survey was available for a municipality and estimates from the MSFI were compared to it. The levels of error were quite high. It was concluded that the large area estimates were biased by the reference dataset.

Model-based prediction error uncertainty estimation for k-nn method

The k-nearest neighbour estimation method is one of the main tools used in multi-source forest inventories. It is a powerful non-parametric method for which estimates are easy to compute and relatively accurate. One downside of this method is that it lacks an uncertainty measure for predicted values and for areas of an arbitrary size. We present a method to estimate the prediction uncertainty based on the variogram model which derives the necessary formula for the k-nn method. A data application is illustrated for multi-source forest inventory data, and the results are compared at pixel level to the conventional RMSE method. We find that the variogram model-based method which is analytic, is competitive with the RMSE method.

Landsat TM imagery and high altitude aerial photographs in estimation of forest characteristics

Satellite sensor data have traditionally been used in multi-source forest inventory for estimating forest characteristics. Their advantages generally are large geographic coverage and large spectral range. Another remote sensing data source for forest inventories offering a large geographic coverage is high altitude aerial photography. In high altitude aerial photographs the spectral range is very narrow but the spatial resolution is high. This allows the extraction of texture features for forest inventory purposes. In this study we utilized a Landsat 7 ETM satellite image, a photo mosaic composed of high altitude panchromatic aerial photographs, and a combination of the aforementioned in estimating forest attributes for an area covering approximately 281 000 ha in Forestry Centre Häme-Uusimaa in Southern Finland. Sample plots of 9th National Forest Inventory (NFI9) were used as field data. In the estimation, 6 Landsat 7 ETM image channels were used. For aerial photographs, 4 image channels were composed from the spectral averages and texture features. In combining both data sources, 6 Landsat channels and 3 aerial image texture channels were selected for the analysis. The accuracy of forest estimates based on the Landsat image was better than that of estimates based on high altitude aerial photographs. On the other hand, using the combination of Landsat ETM spectral features and textural features on high altitude aerial photographs improved the estimation accuracy of most forest attributes.

Performance of different spectral and textural aerial photograph features in multi-source forest inventory

Most multi-source forest inventory (MSFI) applications have thus far been based on the use of medium resolution satellite imagery, such as Landsat TM. The high plot and stand level estimation errors of these applications have, however, restricted their use in forest management planning. One reason suggested for the high estimation errors has been the coarse spatial resolution of the imagery employed. Therefore, very high spatial resolution (VHR) imagery sources provide interesting data for stand-level inventory applications. However, digital interpretation of VHR imagery, such as aerial photographs, is more complicated than the use of traditional satellite imagery. Pixel-by-pixel analysis is not applicable to VHR imagery because a single pixel is small in relation to the object of interest, i.e. a forest stand, and therefore it does not adequately represent the spectral properties of a stand. Additionally in aerial photographs, the spectral properties of the objects are dependent on their location in the image. Therefore, MSFI applications based on aerial imagery must employ features that are less sensitive to their location in the image and that have been derived using the spatial neighborhood of each pixel, e.g. a square-shaped window of pixels. In this experiment several spectral and textural features were extracted from color-infrared aerial photographs and employed in estimation of forest attributes. The features were extracted from original, normalized difference vegetation index and channel ratio images. The correlations between the extracted image features and forest attributes measured from sample plots were examined. Additionally, the spectral and textural features were used for estimating the forest attributes of sample plots, applying the k nearest neighbor estimation method. The results show that several spectral and textural image features that are moderately or well correlated with the forest attributes. Furthermore, the accuracy of forest attribute estimation can be significantly improved by a careful selection of image features.

Local radiometric correction of digital aerial photographs for multi source forest inventory

Digitized colour infrared images are a potential information source for multi-source forest inventory applications and particularly for estimating the forest characteristics of relatively small areas. The main problem in computer-aided image analysis of aerial photographs is the presence of bidirectional reflectance, which causes the spectral values of the image pixels to depend on their location in the image. Because of this, and the lack of operational methods for radiometric correction, the full potential of aerial photographs has not been utilised. This paper presents a local radiometric correction method that can be used for reducing the effect of bidirectional reflectance. The method employs satellite images that are less affected by the bidirectional reflectance for the local adjustment of the registered pixel values of aerial photographs. Because of the different spatial resolution of the aerial and satellite imagery, the local correction coefficients are computed for units that are larger than a single aerial photo pixel. In this way, the general level of brightness of the correction units can be determined on the basis of the satellite imagery while retaining the finer spatial resolution of the original aerial photo. The main advantage of the suggested method is that it does not require complex mathematical models for simulating the effect of bidirectional reflectance, neither does it require a priori knowledge of the actual forest attributes in the inventory area, but relies only on the image data.

Combining remote sensing, data from earlier inventories, and geostatistical interpolation in multisource forest inventory

Multisource forest inventory with two-phase sampling offers several advantages in the forest management planning when compared with the traditional visual inventory by stands. For example, by combining data from remote sensing imagery with field measurements, it is possible to estimate the forest characteristics of large areas at a more reasonable cost than by using the traditional visual inventory by stands. In this study, the k-nearest-neighbours estimation (k-nn), stand inventory data, and geostatistical interpolation were combined for estimation of five forest variables (mean diameter, mean height, mean age, basal area, and volume) per sample plot and stand. Digitized aerial photograph features, visually interpreted aerial photograph features, and updated stand inventory data were used as the auxiliary data sources in the estimation of forest variables. The results show that, at the sample plot level, the k-nn estimates based on the auxiliary data sources were more accurate than the updated stand inventory data transferred to the plot level. At the stand level, the updated stand inventory data were more accurate than the k-nn estimates. When the k-nn estimates were combined with the updated stand inventory data, the accuracy of the estimates was significantly improved at both the sample plot and stand level. The geostatistical interpolation, which was tested on the stand level estimation, did not result in any further improvement in the accuracy of the estimates.

Image segment-based spectral features in the estimation of timber volume

Plot- and stand-level errors associated with satellite image-based multisource forest inventory (MSFI) applications have been relatively high. The reasons suggested for that are related to the limited spatial resolution of the image material. The introduction of very high spatial resolution (VHR) images to MSFI applications should, therefore, diminish these errors. The use of VHR images is, however, problematic, because pixel-by-pixel analysis methods are no longer applicable. The paper presents an image segment-based approach to the determination of feature extraction and image analysis units. The study was carried out in Southern Finland and employed a spectrally averaged imaging spectrometer (AISA) image and field data gathered from sample plots. A two-phase segmentation method was applied and a large number of segment-based spectral features was extracted and used as input to a feature selection procedure. Forward selection based on an improvement of RMSE was applied. The performance of segment-based features (SF) was compared to that of reference features (RF) extracted from square-shaped windows. The estimation results revealed that even though the applied segmentation method succeeded well in the determination of units of feature extraction and image analysis, the differences between the performance of SF and RF were small and the plot-level estimation errors remained high. The study suggests that large estimation errors are due to the local nature of the field data and may be diminished using data that is representative at the segment level.

Multi-source inventory of the forests of the Hebei Forestry Bureau, Heilongjiang, China

A multi-source forest inventory method is applied to the estimation of forest resources in the area of the Hebei Forest Bureau in Heilongjiang province in North-East China. A stratified systematic cluster sampling design was utilised in field measurements. The design was constructed on the basis of information from earlier stand-level inventories, aerial orthophotographs, experiences from other sampling inventories and the available budget. Sample tree volumes were estimated by means of existing models. New models were constructed and their parameters estimated for tallied tree volumes and volume increments. The estimates for the area of the Bureau were computed from field measurements, and for the areas of the forest farms estimated from field measurements and satellite images. A k-nearest neighbour method was utilised. This method employing satellite image data makes it possible to estimate all variables, particularly for smaller areas than that possible using field measurements only. The methods presented, or their modifications, could also be applied to the planning and realisation of forest inventories elsewhere in Temperate or Boreal zones. The inventory in question gave an estimate of 114 m3/ha (the multi-source inventory 119 m3/ha) instead of 72 m3/ha as previously estimated from available information. Totally nineteen tree species, genera of species or tree species groups were identified (Appendix 1). The forests were relatively young, 60% of them younger than 40 years and 85% younger than 60 years.

Improving multi-source forest inventory by weighting auxiliary data sources

A two-phase sampling design has been applied to forest inventory. First, a large number of first phase sample plots were defined with a square grid in a geographic coordinate system for two study areas of about 1800 and 4500 ha. The first phase sample plots were supplied by auxiliary data of Landsat TM and IRS-1C with principal component transformation for stratification and drawing the second phase sample (field sample). Proportional allocation was used to draw the second phase sample. The number of field sample plots in the two study areas was 300 and 380. The local estimates of five continuous forest stand variables, mean diameter, mean height, age, basal area, and stem volume, were calculated for each of the first phase sample plots. This was done separately by using one auxiliary data source at a time together with the field sample information. However, if the first phase sample plot for which the stand variables were to be estimated was also a field sample plot, the information of that field sample plot was eliminated according to the cross validation principle. This was because it was then possible to calculate mean square errors of estimates related to a specific auxiliary data source. The procedure produced as many estimates for each first phase sample plot and forest stand variable as was the number of auxiliary data sources, i.e. seven estimates: These were based on Landsat TM, IRS-1C, digitized aerial photos, ocular stereoscopic interpretation from aerial photographs, data from old forest inventory made by compartments, Landsat TM95–TM89 difference image and IRS96–TM95 difference image. The final estimates were calculated as weighted averages where the weights were inversely proportional to mean square errors. The alternative estimates were calculated by applying simple rules based on knowledge and the outliers were defined. The study shows that this kind of system for finding outliers for elimination and a weighting procedure improves the accuracy of stand variable estimation.

Adapting Finnish Multi-Source Forest Inventory Techniques to the New Zealand Preharvest Inventory

To measure many small areas in a forest, each to the required detail and precision, and sufficiently frequently, is too expensive using ground-based inventories alone. The Finnish satellite image-aided National Forest Inventory (NFI) method was modified and tested with a view to developing a new preharvesting forest inventory method for assessing the volumes of potential timber assortments (log products) in New Zealand radiata pine (Pinus radiata D. Don) plantations. Data from 188 ground plots were supplied by a forest company for a 1000 ha block of radiata pine. This material was combined with known stand boundaries from maps and data from a Landsat TM image in order to predict for each pixel and each stand the pruned and unpruned sawlog, pulp and total standing volumes. A k-nearest neighbour algorithm was applied for the estimation. Some stand characteristics, obtained from stand histories, were used as co-variates, in addition to spectral features. Cross-validation tests indicated that estimates of plot variables were unbiased, though with high RMSEs. Stand level estimates were also unbiased on average, but showed high RMSEs. Research into the effects of stratification and methods by which to segment stands into sub-areas uniform with respect to timber assortments is continuing.