Application of ERS-1 data to forest inventory
Proposal for Application Oriented ERS-1 Pilot Project,
September 11, 1992
Abstract
The main objective of the proposed two and a half year project is to
evaluate the feasibility of ERS-1 SAR images to estimating forest resources in
large areas, including (a) national level (Finland) and (b) multinational level
(Boreal Forests). Four test sites are used, located in southern and
northern Finland, Russia and Canada, respectively.
Two application strategies are used for each site: (a) SAR images are compared
with ground-based data only, and (b) SAR images are used together with
additional
information (optical satellite data, traditional ground-based data and
airborne ranging radar data). Seasonal effects in SAR data applicability are
examined. The optimum SAR parameters for forest inventories are examined
using an airborne multiparameter (frequency, polarization, incidence angle)
scatterometer. The feasibility of an airborne ranging radar for reference
measurements (ground truth) is examined).
Project Objectives
The main objective of the project is to evaluate the feasibility of ERS-1 SAR
images to estimating forest resources in large areas, including (a) national
level (Finland) and (b) multinational level (Boreal Forests). Four test sites
are used, located in southern finland, northern Finland, Russia and Canada,
respectively.
Two application strategies are used for each site: (a) SAR images are compared
with ground-based data only, and (b) SAR images are used together with additional
information (optical satellite data, traditional ground-based data and
airborne ranging radar data). Seasonal effects in SAR data applicability are
examined. The optimum SAR parameters for forest inventories are examined
using an airborne multiparameter (frequency, polarization, incidence angle)
scatterometer. The feasibility of an airborne ranging radar for reference
measurements (ground truth) is examined.
Special emphasis in the evaluation is in (a) accuracy, and (b) the cost-benefit
analysis of the radar-related methods compared with the present methods.
50% of Finland's export net income is based on forest products (paper, timber,
machinery). National Forest Inventories are carried out in order to properly
plan Finland's forest industries and utilization and management of the forests.
The main interest of the Finnish Forest Research Institute (FFRI) in the
proposed project is development of new operational methods for the Finnish
National Forest Inventory (FNFI) and to use the same methodology on a larger scale
as well. FFRI is responsible for the FNFI, performed seven times so far.
In each inventory, the national forest resources (soil, site quality, amount
and quality of growing stock and growth by tree species) have been
estimated using ground-based field measurements. Using the traditional techniques,
one inventory takes about seven to ten years. Recently, new methods for this
task have been developed, including use of spaceborne and ground-based data
from several sources.
FFRI is also involved in international projects. The purpose of one of the
projects(now in planning phase) is to map forests of the whole
circumpolar Subarctica. Remote sensing data is foreseen as a substantial
source of information.
The main interest of the Laboratory of Space Technology, Helsinki University
of Technology (LST/HUT) in the proposed project is to provide microwave remote
sensing expertize to the development of new forest inventory methods.
LST/HUT specializes in microwave remote sensing, including construction of
airborne sensors (radars and radiometers), airborne campaigns and development
of algorithms to retrieve geophysical parameters from spaceborne
microwave data. The main application areas are forests, sea ice, snow, and
atmospheric ozone.
Present Methods
FFRI is responsible for the FNFI, which has produced large-area forest resource
information during the past 70 years. The information has been applied in planning
Finland's forest industries and utilization and management of the forests.
The first seven inventories, conducted in 1921-1984, were based on ground-based
field measurements only. In the first four inventories, linewise survey sampling
was applied, with lines passing from southwest to northeast through the country.
since the fifth inventory (started in 1964), systematic cluster sampling was
employed, including aerial photos and two phased stratified sampling in
Lapland. From those inventories, only large-area estimates were applicable
due to the limited amount of data. The eighth national forest inventory
was started in 1986 and will be completed in 1995.
During the eighth national inventory, FFRI has developed a new inventory
system in order to get geographically localized, up-to-date information,
applicable to smaller areas than earlier. In addition to ground-based
measurements, the method employs satellite imagery, digital map data
(e.g., digital elevation model when relevant) and, in the future, other
geographical data, including soil and meteorological data. Using image
analysis methods, estimates of all variables of the inventory can be computed for
each pixel.
The system is now operative. Using the new method, an inventory has been made
in an area of about 8 million hectares producing theme maps and statistics for
large and small areas (even single stands). Landsat TM images have been
applied so far. In future, FFRI plans to use SPOT images together with
TM images.
The most expensive part of the inventory are ground-based field measurements.
About three to three and a half million Finnish Marks are spent every year for
those measurements, the area being about four million hectares. The total annual
budget of the inventory amounts to six million Finnish Marks (about one million
ECU).
In addition to the Finnish National Forest Inventory, forest industries and other
landowners spend on various forest inventories about 100 million Finnish
Marks (16 million ECU) annually. The most expensive part (60 to 70 million FIM)
is the standwise forest inventory, which is used to small-scale operations
and management planning.
FFRI is also involved in international projects. The purpose of one
of the projects is to map forests of the whole circumpolar Subarctica. This
project is in the planning phase. Another project includes forest damage evaluation
in Central Europe.
The Laboratory of Space Technology, Helsinki University
of Technology has designed and constructed several airborne microwave sensors for
developing various remote sensing applications. One of the sensors, an 8-channel
ranging scatterometer (called HUTSCAT; 5.4/9.8 GHz, 4 linear polarizations
at each frequency) has proved to be an efficient tool for collecting information
on tree characteristics.
So far an accuracy of 0.8 meters (standard deviation) has been achieved with the
HUTSCAT in measuring the dominant and mean tree heights with the ranging
capability. The corresponding accuracy for the stem volume is 40m3/ha (15%
accuracy). These numbers are comparable to the accuracies obtained in time-consuming
field measurements. Hence, airborne ranging radars may offer a new technique for
providing reference data (ground truth) for forest inventories.
HUT has experience in interpreting microwave (radar and radiometer) interaction
with geophysical targets, including forests, sea ice, snow and recently,
atmospheric ozone.
Anticipated Results
A problem related to the use of spaceborne optical remote sensing data is that
due to the cloud cover, imagery that covers the whole country each year, cannot
be obtained. For example, in the summers of 1989 and 1990, the relative coverage
in Finland obtained by using Landsat TM data was 30 to 40 %. To some extent, tree
and stand models and simulation can be applied to update the data between two image
acquisition dates. However, the estimates could be more accurate and informative if,
for instance, clear cutting areas could be estimated using other data sources,
e.g. remote sensing data. Ground measurements are practically impossible for
this purpose because of their cost and airborne data is also fairly expensive.
A new promising source of spaceborne remote sensing data for this purpose is ERS-1
SAR.
The forest resources of circumpolar Subarctica are not yet known precisely.
Some attempts to coordinate projects for inventories are going on. For many
purposes, the distribution of forest land and tree species composition is relevant
information. Remote sensing data should be involved in those inventories because
the areas considered are huge.
In this project, answers to the following problems are sought:
Estimation based on SAR data only
- Which variables can be estimated with SAR data alone, e.g. land use category
(productive forest land, other wooded land, waste land, non-forest land,
water), site quality, tree species proportions, tree stem volume by tree species,
growth, leaf area index and total biomass of trees?
- What is the smallest area for which the estimates can be computed reliably;
examples are (a) the whole country or a part of the country (order of magnitude
10 million hectares), (b)forestry board district(0.5 -1 million hectares), (c)
a municipality (50000 ha), (d) forest holding?
- What kind of changes can be estimated using SAR data, e.g. clear cuttings,
deforestation and defoliation?
- Is there an optimum season(winter, spring, summer, autumn) for using SAR
data for forest inventory? Is a single season enough or should multitemporal
data be used?
- What are the optimum SAR parameters for forest inventory (based on HUTSCAT
scatterometer data obtained at 5 and 10GHz, 4 linear polarizations, incidence
angle 0 to 45 degrees off nadir)?
Estimation based on a combination of ERS-1 SAR data, other satellite
data and data from various informations systems
- Does the SAR data improve the accuracy of estimates of the above
variables if used together with optical satellite images (e.g. SPOT, Landsat),
ground-based reference measurements and data from other information sources?
Reference measurements
Satellite-based forest resource mapping needs a large data set of reference
measurements to characterize each type of forest stands. Typically, reference
measurements are carried out manually. The capability of an airborne ranging
radar system to substitute these reference measurements is examined in this study.
- What characteristic can be obtained from the measurement conducted with an
airborne ranging radar? What are the obtained accuracies?
- If reference measurements are obtained using both the ranging radar and the
aerial photography, what are the accuracies of the reference measurements?
- Is it possible to substitute at least a part of ground measurements in
intensive inventories and all measurements in extensive inventories by airborne
ranging radar (HUTSCAT)?
A realistic outcome of the proposed project is substantial progress in developing
spaceborne SAR methods for operational forest inventories in northernmost areas.
Further progress, possibly reaching the pro-operational/operational state, can
be expected during the lifetime of the ERS-2 satellite.
The results of the project and developed methodology may possibly be applied -
in addition to mapping coniferous forests - also for global monitoring after
further development.
Examples are:
- Estimation of forest resources of Boreal Forests.
- An optimum combination of using microwave, infrared and visible light
in satellite-based forest inventory.
- A new method of reference measurements(the method could be used alone
for forest areas smaller than 100.000 hectares).
- Monitoring of distribution of forest land in circumpolar Subarctica and
its possible changes caused by climate changes.
- Deforestation in the Tropics.
The annual cost of the FNFI is six million Finnish Marks (MFIM) (about one
million
ECU), out of which 3.5 MFIM are used for ground-based reference
measurements. The total annual cost of various forest inventories in Finland
is more that 100 MFIM (16 million ECU). In future, it may be possible to
decrease these costs by using spaceborne SAR data and airborne profiling
radar data.
An especially large potential of satellite technology is in applying SAR
data to the entire Boreal Forests area (Scandinavia, Canada, Russia). The
value of this application cannot be even estimated at this time.
Measurements
It is proposed to use four test sites in the study, located in southern
and northern Finland, Russia and Canada, respectively.
The following measurements/data collection procedures concerning the
test sites are planned:
- Updating of the sample plot measurements from the FNFI, where necessary
(including tree age and height, defoliation, soil, site quality, amount
and quality of growing stock and growth by tree species).
- Helicopter-borne radar ad radiometer measurements in the Finnish
test sites, coincident with selected cases of ERS-1 SAR data acquisition.
- Collection of ground-based data on weather and soil conditions,
coincident with the ERS-1 SAR data acquisition.
- Acquirement of optical satellite data(Landsat TM and SPOT, if possible)
coincident with selected cases of ERS-1 SAR data acquisition.
- Digital map data (including digital elevation model).
If the proposal made by Prof. Martti Hallikainen to ESA (MIMR Expert Group,
June 1992) for an airborne microwave campaign MACSI'95 in northern
Finland will be accepted, our project will take advantage of this possibility
when conducting airborne measurements and collecting other forest
information.
ERS-1 data sets
Time period
Data from the ERS-1 SAR is requested for a period of two years: January 1993
through January 1995. The reason for the early start is that the minimum
needed SAR data coverage is a full year, due to comparing the usefulness
of data from various seasons. Since the lifetime of ERS-1
may come to an end in 1994, the above proposal is realistic. If data
can be only delivered starting December 1993, the period to be covered with
data will be extended until December 1995.
Test sites
SAR data are requested for four test sites, located in southern and
northern Finland, Russia and Canada, respectively; see Table 1. Using
four test sites, the effect of several parameters concerning climate and
trees can be evaluated. This approach thus leads to generalization of
the resultas for large areas in the circumpolar Subarctica.
Table 1. The test sites.
Number Name Latitude Longitude
1 Porvoo 60 30 00 N 25 30 00 E
2 Inari 68 45 00 N 27 30 00 E
3 Sortavala 62 00 00 N 30 45 00 E
4 Whitecourt TBD TBD
In the southern test site in Finland, pine-, spruce-, and birch-dominated
forests occur, with little variation in tree age (due to industrial use).
Most of the forests in Scandinavia are like those in that test site. In
the northern test site in Finland, mainly pine-dominated, unevenly aged and
sparse forests occur. In the Russian test site, the forests are often
spruce-dominated, unevenly aged and sparse forests occur. Thus they differ
substantially from the carefully treated "industrial" forests in Finland.
Untreated forests characterize the vast Boreal Forest belt in Russia.
The location of the Canadian test site is selected in cooperation with the
Petawawa National Forestry Institute, Ontario.
For each of the first three test sites, reference data from the Finnish
National Forest Inventory presently exists. The data will be updated, where
necessary. From the Canadian test site, reference data will be obtained in
cooperation with the local forestry institute.
Frequency and amount of SAR data
In 1993, one SAR image per test site is requested every six weeks in
order to have complete coverage for studying seasonal effects. Thus the
total in 1993 is 36 SAR images. In 1994, we concentrate our efforts on seasons
found to be the most promising for SAR-based forest inventory. Depending on
the results obtained in 1993, a total of SAR images with minimum of
18 and maximum of 36 is requested in 1994. Thus a total of 54 to 72 SAR
images over the two-year period is requested.
Type of Data Products
ERS-1 SAR Precision (SAR.PRI) images are used. The product medium is CCT.
The product characteristics, including localization, resolution, accuracy
etc. as stated in the ERS-1 User Handbook are adequate for our purposes.
The preferred time delay for receiving the images is 3 weeks after
acquisition. The maximum time delay for receiving images is 6 weeks
after acquisition.
Project Outline, Resources and Facilities
Project outline
The project starts in January 1993 and ends in June 1995. SAR data are
requested for a test period of two full years, complemented with airborne
radar and radiometer measurements in the Finnish test sites and ground-based
data collection. We prefer to have SAR data for the period of January 1993
through january 1995. If not possible, the project will be extended
correspondingly. The project outline is the following:
1993:
Ground truth information on the four test sites is updated. Methods to
handle and compare SAR images and optical images with ground based information
are completed. Preparations for airborne radar and radiometer measurements in
the Finnish test sites are made. Monthly SAR images are corrected and
backscatter intensity correlated with tree parameters for each season.
Selected airborne measurements are conducted in the Finnish test sites;
these data are analyzed along with satellite and ground-based data.
Preliminary results on the main research topics are obtained by the end of
1993.
1994:
Basically the same approach is used as in 1993. Main attention is paid
to features/algorithms/seasons found to be the most promising in 1993.
Algorithms are further developed.
1995:
Data analysis and algorithm development is completed and the cost-benefit
analysis is performed. The final report is written.
FFRI Facilities (Image processing Systems):
- Hardware
- VAX/VMS-3100/76-stations with 8-bit colour display
- DEC/ULTRIX(5200)-stations with 24-bit colout display
- Sun Sparcstations II with 8-bit colour display
- Versatec Electrostatic Plotter (paper size A0)
- Software
- DISIMP software (by CSIRO Australia for raster image processing
- own utilities for raster image processing
- ERDAS and ARC/INFO running in Sun Sparcstations for raster and vector
data processing
HUT Facilities:
- Helicopter-borne ranging scatterometer system operating at 5.4 and 9.g GHz,
with VV, HH, HV, VH polarization modes at each frequency (simultaneous data
collection at 8 channels).
- Side-looking airborne radar (SLAR) operating at 9.5 GHz, VV polarization,
spatial resolution (along track x across track) 30 m x 20 m (for range 3km),
with the across track resolution being adjustable from 20 to 400 m.
- Helicopter-borne ranging scatterometer operating at 5.3 GHz with VV, HH,
HV, and VH polarization modes (simultaneous data collection at 4 channels).
- Helicopter-borne profiling microwave radiometer system operating at 24, 35,
48 and 94 GHz, V polarization , incidence angle adjustable from 0 to 45
degrees.
- Helicopter-borne imaging 94GHz microwave radiometer with V and H
polarizations, incidence angle 50 degrees (available for the 1994 field
experiments).
- Image processing system Sun Sparcstation II with ERDAS and Arc-Info
software.
Project Personnel:
1993: FFRI: 2.0 person-years
HUT: 2.5 person-years
1994: FFRI: 2.0 person-years
HUT: 2.5 person-years
1995: FFRI: 1.0 person-years
HUT: 1.5 person-years
Project Schedule
1993:
- Updating of forest parameter information in the test sites
- Monthly SAR image acquisition
- Airborne microwave measurements in the Finnish test sites
- Collection of weather and soil data in the test sites
- Geometrical and radiometrical correction of SAR/optical images
- Comparison of SAR and optical data
- Development of classification algorithms based on ERS-1 SAR data and
data from other sources
- Preliminary estimation of accuracy of algorithms
- Preliminary decision on what kind of variables are possible to estimate
by means of SAR data (land use category, tree species, growing stock etc.)
1994:
- Decision on what seasons/features/algorithms are most promising and
concentration on them
- Selected SAR image acquisition
- Selected airborne microwave measurements in the Finnish test sites
- Collection of weather and soil data in the test sites
- Geometrical and radiometrical correction of SAR/optical images
- Improvement of forest inventory algorithms (multitemporal aspects, etc.)
- Evaluation of feasibility of data fusion using SAR and microwave
radiometer data
- Evaluation of feasibility of airborne ranging radar data for reference
measurements
1995:
- Cost-benefit analysis and comparison with traditional methods when relevant
- Statistics of forest resources and theme maps both in digital and analog
form
- Final report of the project
Funding
A proposal concerning funding of this project will be submitted to the
Technology Development Centre.
A limited amount of manpower and funding will be provided by the two
participating institutes.
Contact information
- Finnish Forest Research Institute METLA
- Unioninkatu 40A, FIN-00170 Helsinki, Finland
- tel. +358-0-857 051, fax +358-0-625 308
- Internet: Erkki.Tomppo@metla.fi, Petri.Mikkela@metla.fi
[METLA] [National Forest Inventory]
[ERS1-SAR Project]
PM, August 30, 1995.