http://creativecommons.org/licenses/by/4.0/

Natural Resources Institute Finland (Luke) owns the copyright, data protection, and other immaterial rights to this product. The Topographic Database from the National Land Survey has been utilized when making the product. This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA. When using the material, the owner of the rights to the material must be given as "©Natural Resources Institute Finland, 2015" and the name of the material must be given as "The Multi-source National Forest Inventory Raster Maps of 2013". For research use, the description of the method is in the references in the metadata element Lineage. A scientific citation practice shall be used in research use.

The first country level estimates correspond to years 1990-1994. The most

recent versions are from years 2005, 2007, 2009, 2011 and 2013.

The first set of the products freely available are from year

2009. The new set of the products will be produced annually or biannually in

the future. The map from products are in a raster format with a pixel size of

16m x 16m and in ETRS-TM35FIN coordinate system. The products cover the combined

land categories forest land, poorly productive forest land and unproductive

land. The other land categories as well as water bodies have been delineated

out using the elements of topographic database of the Land Survey of Finland.

For the 2013 products, in total 57563 NFI field plots were used, locating

either on forest land, poorly productive forest land or unproductive

land. The satellite image data consisted of 41 Landsat 8 OLI images and

four IRS-P6/Resourcesat 1 LISS-III images.

The field data in the 2013 products were up-dated to correspond

the situation on 31 July, 2013. The length of the up-dating period was

calculated for each field plot from the date of the field measurement to the

up-dating date 31 July, 2013. The start of the tree growth was supposed to be

on May 1.

The relative increment of the volume of the growing stock in a forest stand

was calculated using the models by Nyyssönen and Mielikäinen (1978) for pine

(Pinus silvestris) and spruce (Picea abies). The models for pine were used for

broad-leaved trees. The volume increments were calculated by stand layers in

case of multi-layer stands. The proportions of volumes by layers were estimated

proportionally to the quantity the basal area of the layer multiplied by the

mean height of the layer.

Regeneration cuttings on the field plots were assessed using satellite images

and, in some cases, with aerial photographs. The stand data of the plots cut

were changed to stand data for open area plots. The final volume increments

were calibrated in such a way that the volumes by tree species on July 31,

2013, was the same as that given by the regression line estimated from field

data alone.

For the relative height increment, diameter increment and basal area

increment, simple fixed parameter regression models were estimated using data

from the permanent sample plots of NFI10.

The models were used in a similar way as the volume models.

The biomass estimates by field plots and biomass compartments

were up-dated proportionally to the volume changes.

For a cover as complete as possible from the entire country, the 2011 product

has been completed by the data estimates from the recent years. The product

thus consists of the following sub-products:

1. The estimates from 2013, the NFI field data from 2003-2013 updated to 31.7.2013 and the satellite images from 2012-2014

2. The estimates from 2011, the NFI field data from 2007-2011 updated to 31.7.2011 and the satellite images from 2009-2012,

3. The estimates from 2009, the NFI field data from 2006-2010 and the satellite images from 2009-2010,

4. The estimates for about the municipalities Inari and Utsjoki in North Lapland, the NFI field data from 2003 and the satellite images from 2009,

5. The estimates for about municipality Enontekiö in North Lapland, the NFI field data from 2003 and the satellite images from 2000,

6. The estimates from 2007, the NFI field data from 2005-2008 and the satellite images from 2005-2007.

Data source index, MS-NFI-2013, has been added to the product to indicate the

source of the estimates.

The map form estimates were made using the improved k-Nearest Neighbour method

(ik-NN method). The value of five for k was used most frequently. The weights

of the features in the ik-NN method are sought using an optimization method

based on genetic algorithm. Coarse scale estimates of forest variables were

used as the supplementary data. The volumes by tree species groups were

selected as the variables. The purpose is to direct the selection of the

neighbours, on the average, to forests similar to the target pixel (see the

references below). The estimation was made separately for mineral soils, mires

and open bogs and fens. The stratification of both the satellite image and the

field plots were made using the topographic map data of Land Survey Finland.

The product consists of 44 theme maps in raster format plus data source

index. These can be grouped as follows:

The volume of growing stock is available as a total for all tree species and

broken down into tree species groups (Scots pine, Norway spruce, Birch, Other

broad leaved trees) and into timber assortments (saw timber, pulpwood). The

group Scots Pine includes other coniferous species than Norway Spruce, and the

group Birch includes Betula pubescens and Betula pendula. The volume of a tree

is defined as the volume of the stem wood above stump until the top of the

tree. The volume of a tree in the field data is estimated using the parameters

measured in the field and the volume models. The unit and class interval of

the volume is 1 m3/ha in the products available for downloading.

The biomass of the growing stock has been estimated and is available by tree

species groups and by seven tree compartments. The biomass of stem and bark of

a tree is defined as the biomass of the stem above bark and above stump until

the top of the tree. The biomass of the living branches includes the biomass

of the living branches without needles or leaves. The biomass of the dead

branches includes the biomass of the dead branches possibly left in a living

tree. The foliage biomass includes the biomass of the living needles or

leaves. The biomass of stumps includes the biomass of the above and below

ground stump parts without roots. The root biomass includes the biomass of the

living roots with a diameter of at least 1 cm. The biomass of stem residual is

defined as that part of the stem biomass that can not be used as timber or

pulpwood due it size or quality.

The biomasses of the sample trees on a NFI field plot are calculated from the

living sample trees belonging to a plot using the wood density models (Repola

et al. 2007) and biomass models (Repola 2008, 2009). The biomasses of the

trees called tally trees are estimated using the estimates of the sample trees

(with more parameters measured) and the parameters of tally trees and

stands. The unit of the biomass in the maps available for downloading is 10 kg/ha.

The basal area of the growing stock on a forest stand is the cross section

area of the tree stems of a stand per hectare and measured at a height of 1.3

m. The basal area is measured in the field for the field plot stands on forest

land and poorly productive forest land in the classes of 1 m2/ha.

The age of the growing stock on a forest stand is the weighted average of the

trees, the basal area of the tree as the weight. The age is assessed in the

field for the field plot stands on forest land and poorly productive forest

land in the classes of one year.

The mean height of the trees on a forest stand is the height of the basal area

median tree for the development classes young thinning stand or more mature

stands. It is about the same as the basal area weighted average height. For

seedling stands, the mean height is the average height of the dominant and

co-dominant seedlings. The mean height is assessed in the field in the classes

of 1 dm.

The mean diameter of trees is assessed at a height of 1.3 metres and is the

the diameter of the basal area median tree. It is about the same as the

weighted average diameter, the basal area of a tree as the weight. It is

assessed for the field plot stands on forest land and poorly productive forest

land in the classes of 1 cm.

The canopy cover of trees is the vertical projection area on the horizontal

plane of the canopies of the individual trees on a field plot (without double

counting the overlapping canopies). In NFI10, it was assessed in the field as

a shares (0-99%) on a fixed radius plot. For the NFI11 plots, it was estimated

using k-NN method and the NFI10 plot data. In North Lapland in NFI9, the

canopy cover was assessed in three categories if the plot was either on forest

land, poorly productive forest land or unproductive land. A regression model

was constructed to estimate the cover in the classes of one percent.

The canopy cover proportion of broad leaved trees is derived from the total

cover using the volume of the growing stock. In the seedling stands the canopy

cover of broad leaved trees is assessed using the shares of the stem numbers.

The theme "Land class" divides forestry land into sub-categories forest land

(pixel value 1), poorly productive forest land (2), unproductive land (3) and

other forestry land: forestry roads, forest depots, etc. Outside forestry land,

the land class describes land use. In the present themes, the combined

mask of forest land, poorly productive forest land and unproductive land is

based on the topographic database from the National Land Survey. One

of three land categories is estimated for the pixels inside the three mask

categories. The "Land Class based on FAO FRA" divides forest into four

categories based on the definition of the United Nations FAO Global

Forest Resource Assessment (FRA): forest (1), other wooded land (2),

other land (3) and other land with tree cover (4)

The main site class divides the forest land, poorly productive land and waste

land into mineral soils (1) and peatlands, and further divides the peatlands

into spruce mires (2), pine mires (3) and treeless mires (4). Both the satellite

images and the NFI field plots are stratified to mineral soils and peatlands

before analysis according to the topograhic database from the National

Land Survey. The most probable of the four NFI main site classes is

predicted for each pixel within these strata. This means that each stratum

may include both mineral soils and peatlands according to the NFI classification.

The site fertility classes are used for grouping the forest by vegetation

zones into uniform classes according to their site fertility and wood

production capacity. In national land-use classification, all stands on

mineral soil with site fertility class in 1 - 6 were classified as forest land

(1 is herb rich sites, 2 is herb rich heath forests, 3 is mesic forests, 4 is

sub-xeric forests, 5 is xeric forests, 6 is barren forests). Class 7 (rocky

and sandy soils and alluvial lands) can be forest land, poorly productive

forest land, or unproductive land, and class 8 (summit and fjeld land with

single coniferous trees) either poorly productive forest land or unproductive

land. Classes 9 (mountain birch dominated fjelds) and 10 (Open fjelds) are

poorly productive forest land or unproductive land. Both natural and drained

peatlands are classified into six site fertility classes independently of the

land class. Class 1 includes euthropic mines and fens, 2 mesothropic mires and

fens, 3 meso-oligothropic mires, 4 oligothropic mires, 5 oligo-ombothropic

mires and 6 Sphagnum fuscum dominated mires. Both the field plots and

satellite images are stratified prior the analyses into three strata, mineral

soil, pine mires and spruce mires, treeless peatland . The site fertility

class is estimated for each pixel as the most likely site fertility

class. Thus in the products, each NFI field data based category can occur

within each map based stratum.

The estimation errors at pixel level are rather high but decrease when the

area in question increases, i.e., when the area of interest consists of

several pixels. The errors vary by the themes and depend also on the actual

value in the field, for example on the volume of growing stock and the site

fertility class.

The following error estimates are based on the MS-NFI 2009 product.

The magnitude of the average errors of the volume estimates at pixel level are

presented below (SF = South Finland, NF = North Finland, min = mineral soil,

peat = peatland):

species group assort. SF/min SF/peat NF/min NF/peat

all all 86 66 47 32

pine all 63 53 40 26

pine saw t. 39 29 19 7

pine pulpw. 40 37 30 22

spruce all 63 37 27 12

spruce saw t. 43 23 12 3

spruce pulpw. 33 21 18 10

birch all 32 30 19 16

birch saw t. 10 7 2 1

birch pulpw. 25 25 17 13

other br. l. all 22 10 8 4

other br. l. saw t. 7 3 2 1

other br. l. pulpw. 16 8 7 2

The magnitude of the average error of the biomass estimates at pixel level are

presented below (SF = South Finland, NF = North Finland, min = mineral soil,

peat = peatland):

tree species compartment SF/min SF/peat NF/min NF/peat

pine stem and bark 2400 2100 1500 980

pine living branches 400 350 350 230

pine dead branches 95 85 71 52

pine foliage 150 140 140 100

pine stump 190 170 140 94

pine roots 590 500 410 250

pine stem residual 200 220 190 220

spruce stem and bark 2300 1400 1000 450

spruce living branches 550 340 340 150

spruce dead branches 100 63 48 23

spruce foliage 360 250 230 110

spruce stump 210 120 110 48

spruce roots 760 470 430 200

spruce stem residual 19 18 16 14

broad leaved stem and bark 9 8 8 6

broad leaved living branches 31 35 50 46

broad leaved dead branches 62 52 40 25

broad leaved foliage 58 47 41 35

broad leaved stump 9 8 6 6

broad leaved roots 85 66 47 31

broad leaved stem residual 39 29 18 7

The magnitude of the average error of the estimates of the other continuous

variables at pixel level are presented below (SF = South Finland, NF = North

Finland, min = mineral soil, peat = peatland):

theme SF/min SF/peat NF/min NF/peat unit

age 32 35 50 47 a

basal area 9 8 6 6 m3/ha

mean height 59 47 42 35 dm

mean diameter 9 8 8 6 cm

canopy cover 20 18 16 14 %

canopy cover of br. l. 15 12 11 10 %

The overall accuracy (OA) at pixel level of the land class is on the average

92% when the classification is compared to that based on the NFI field

data. The user accuracy (UA) of the category forest land is on average 98%

while the producers accuracy (PA) is on average 95%. The corresponding figures

on poorly productive forest land are 38% and 50% and on unproductive land 74%

and 87%.

The pixel level OA of main site class (mineral soil, spruce mire, pine mire,

treeless mire) is 84%. For the category mineral soil, UA is 95% and PA

88%. The corresponding figures for spruce mires are 20% and 45%, for pine

mires 76% and 76% and for treeless mires 71% and 84%.

The assessment of site fertility is very challenging even in the field and the

results vary by the assessors (field crew leaders). For site fertility class,

OA is 50% as compared to the NFI field data. In most cases, the difference was

not more than one class. The differences were most frequent on one hand for

the categories herb rich sites and herb rich heath forests and for the

corresponding peatland sites, and on the other for poor mineral soil sites and

for ombrotrophic peatlands. The accuracies are highest for mesic forest and

for the corresponding peatland sites (meso-oligotrophic peatlands). In this

category, UA is 60-65% and PA 60% when compared to the result based on the NFI

field data.

The errors of estimates at areal level are lower than the errors presented above.

More information about the methods and the accuracies are given in the publications, e.g.:

Tomppo, E., Haakana, M., Katila, M. & Peräsaari, J. 2008. Multi-source

national forest inventory - Methods and applications. Managing Forest

Ecosystems 18. Springer. 374 p. ISBN 978-1-4020-8712-7,

Mäkisara, K., Katila, M., Peräsaari, J. & Tomppo, E. 2016. The Multi-Source

National Forest Inventory of Finland – methods and results 2013. Natural

resources and bioeconomy studies 10/2016, Natural Resources Institute

Finland. 215 s. http://urn.fi/URN:ISBN:978-952-326-186-0

Tomppo, E. & Halme, M. 2004. Using coarse scale forest variables as

ancillary information and weighting of variables in k-NN estimation: a genetic

algorithm approach. Remote Sensing of Environment 92: 1-20.

Tietoaineistosarja

ISO19115

2003/Cor.1:2006