OF THE STATE OF SAARLAND, GERMANY

                                Helmut Lohmann
                            von der Heydt, Haus 11
                         D-66041 Saarbrucken, Germany


   In February 1990 hurricanes damaged large parts of the forests in southwest
Germany.  In the public and private owned forests of the state of Saarland,
Germany, with a total forest area of about 90,000 ha, more than 2.5 million
cubic meters of wood of different kinds and sizes was thrown down or broken by
the storm events and had to be cut, skidded and transported out of the forest.
The damaged amount of wood was about 8 times as much as the normally yearly
harvesting rate.  Additionally in the surrounding states, more than 60 million
cubic meters of wood was wind thrown.

   To solve management problems related to wind throws, different forest
inventory techniques were carried out by the forest planning department of the
state forest administration.  Especially measurement and data analysis of
systematically distributed permanent sampling points (1x1 km-grid of points of
inventory across the whole state) was very effective.  In a few weeks time a
complete overview could be provided about the impacts of the hurricane events
on the public and private forest.  The results could be used directly for
forest management of the ensuing problems.

   The paper briefly describes different applied ground-based and aerial
inventory techniques as well as some examples of the results of the analysis
of the inventory data.  The results presented show that calculations of
ecological impact of particular and global problems in forestry and of
financial impacts of the hurricane damage can be carried out, and that the
inventory schemes applied are of direct support for current and future
practical forest management of the catastrophic events.


   In February 1990, three hurricanes destroyed more than 10% of the forest
stands in Saarland and damaged nearly 50% of the rest of the woods and changed
the silvicultural structure.  The consequences for current and future forest
management have been considerable.  The wood market has been disrupted for
more than three years.  Therefore, not only local forest management and
technical problems needed direct solutions, but the emerging global forest
policy problems as well.

   One of the main problems in all situations of catastrophic events is lack
of information.  Management decisions can only be as good as the available
information.  Information is necessary about what has happened and the current
conditions.  Therefore management of catastrophic events is not only a problem
of available manpower and technical equipment and its use.  It surely is also
a problem of immediate provision of proper and qualified information and of
effective exchange and flow of information.  It is very important to get the
right information immediately at the right place.

   Therefore, one of the first steps in mangement of catastrophic events
should be gathering as much useful information as possible.  It is very
important to inform oneself first before acting.  The state forest
administration was quite confused after the hurricane events.  All the
responsible units at different administrative levels, especially the forest
districts, had to organize themselves and there was little coordination
initially.  For that reason, the forest planning department could also only
try to get an idea of what had happened.  Therefore, we organized some studies
in conjunction with the local university on the relationships between forest
site characteristics and storm damage to different tree species.

   After about two weeks, when qualified information became available from the
forest districts giving a broad view about the hurricanes impacts, the
planning department could prepare different forest inventory schemes to
collect more exact data, e.g. about the areal extent of the forests thrown by
the storms and the quantity of thrown and broken wood.  This information was
immediately processed to support the short-term crisis management.  But there
was also information necessary about the stands remaining and their structure.
This data was needed for later forest planning concerning future forest
management of the differently damaged stands, for applying methods and
preparing installations of protection against diseases, insects, deer, etc.
Three years after the events, we also know that the ensuing damage, e.g. by
insects (bark beetles), can be as serious as the original catastrophic damage.

                              INVENTORY METHODS

   We used different inventory techniques.  The main inventory was carried out
as a permanent inventory with fixed circular plots of about 30 to 1250 sq. m.
(circular plots of about 3 to 20 meters in radius) depending on DBH-classes of
the trees to be measured.  This ground-based inventory was applied on a
regularly distributed net of a 1x1 km-grid of inventory points across the
whole country (altogether about 900 inventory plots, from which about 700 are
in public forest).  For ownership reasons, we first have gathered data only
from the inventory points in public forest.  In addition to the standing
trees, we had to measure and estimate effects of the hurricanes on the forest
inside and around the inventory point and its plot area.  All fallen down or
broken trees had to be measured.  In cases of total forest loss, the area of
the wind thrown forest was estimated as well as other global and specific
parameters of the former forest around the inventory point.  In the following
list are some of the individual tree parameters measured on the trees still
standing as well as those which were broken or fallen:

   - species
   - age
   - DBH and/or specific other diameter
   - height
   - length/number of pieces
   - social order
   - tree top classification
   - timber quality
   - kind of damage

   Ths inventory was carried out in about one month by three two-person
inventory groups.  Descriptive statistical results were provided to the
foresters immediately after the data were recorded into the computer.  Already
in May 1990, less than three months after the catastrophic events, we had a
complete and statistically representative picture of the economical and
ecological consequences of the hurricane damage and of the subsequent problems
for current and future forest management.

   By that permanent plot inventory, we obtained only specific data about the
present state of the woods and the consequences for forest management.  But
this did not provide particular standwise information which was necessary for
management in the field.  Therefore, we also prepared a second ground-based
inventory which was not conducted by specific inventory personnel as was the
case in the plot inventory, but rather by the forest ranger himself in each
district.  This inventory included data on each individual destroyed or
influenced forest stand.  The forest officers had to estimate the impact of
the hurricanes on the stand and its different tree species concerning
quantity, quality and sizes (BHD-classes) of the broken and fallen wood and
regarding each size of each damaged area.  By this method we could obtain an
overview of the local situation and a quantitative idea of what had to be
organized and managed by the responsible foresters at the forest district
level.  They have to look for solutions to problems like how to get the
destroyed trees cut and removed from the forest stand; what kind and extent of
further transportation would be required (skidders as well as trucks); what
kind and amount of stock capacity had to be installed; what kind of
reforestation had to be prepared with what kind of tree species, etc.

   The major part of the damaged wood could not be sold directly because the
timber market had already broken down.  Therefore, large storage capacities
and sprinkler installations were necessary.  Because of limitations in
finances and manpower as well as shortage of available harvesting and
transporation machinery, an optimal planning of forest management would have
direct as well as indirect impact on the current and future possibilites of
forestry and financial implications for the forest service.  Good planning
would also optimize later marketing of stored timber because only the best
trunks should be stored and watered.  Some of the types of data gathered by
this second inventory scheme are as follows:

   - total area and affected area of each individual forest stand (ha)
   - affected species
   - mean age of stand and affected species
   - total amount of damaged wood (cubic meters)
   - diameter classification of damaged wood
   - quality classification of damaged wood
   - estimates of percent of stems damaged per hectare
   - kind of harvesting methods possible in the particular case (depending
     on site conditions, tree species and mean trunk size).

For conducting the second ground-based inventory the responsible forester of
a forest district (about 1000 hectares) needed only two weeks to obtain the
required data.  The data were directly recorded into the computer and the
results for each individual stand by tree species were evaluated immediately
as well as total results for the whole forest district.  Thus, in about three
weeks time these results could be prepared for different applications in
forest management at the district level.

   Because the second ground-based inventory was not applied in all forest
districts and because the method itself was only producing roughly estimaged
data (for instance the damaged area estimate of the individual stand is not
very accurate as well as the estimated volume data) there was a third
inventory method necessary.  This inventory does not depend on field
measurement but rather on measurement and interpretation of aerial
photographs.  One month after the hurricane events, aerial photos at scale of
1:32,000 were made of the whole country.  On these photos one can identify
nearly each individual fallen tree in stands of deciduous species (trees were
not in leaf) and obtain quite exact area information of each individual wind
thrown stand of coniferous trees, provided that they have the necessary
photogrammetric equipment.  Unfortunately, we are not well equipped, but there
were orthophotos at scale of 1:10,000 available from 1989.  Therefore, by use
of these map-like orthophotos we were nevertheless able to project the
information on the damaged stands from the new aerial photographs into the
orthophotos and then into our existing 1:10,000 scale forest maps.  By that
procedure, we delineated quite accurately the differently damaged or destroyed
parts of the forest stands.  In Table 1 are listed some of the characteristics
used for differentiating the damaged and undamaged forest stand areas.

Table 1.  Area of damaged forest by degree of damage.

Degree of damage        Area           % of public forest
----------------        ----           ------------------
cut down               3,700            6.0 (like clear cut)
very heavy damage      4,600            7.5 (two tree length)
heavy damage           4,000            6.5 (one tree length)
light damage          23,400           38.0 (single trees)
no damage             26,000           42.0
                      ------          -----
  Total               61,700          100.0

   The areas of damaged forested delineated were digitized and maps of the
differentiated hurricane damaged parts of the forest stands were produced.
Statistical analyses, for instance on area size distribution of destroyed or
damaged stands, have also been conducted.

                              INVENTORY RESULTS

   In the following, some of the results of the different inventories are
presented.  These results are mostly produced by combination of data from the
different schemes.  Only examples of results are presented.  The way the
particular figures are calculated will not be explained.  The examples are
concentrated on data gathered in the public forest of the Saarland.  The
public forest comprises nearly 70,000 hectares.  By the permanent inventory,
about 61,700 hectares of that could be identified as timber forest.
Therefore, the first step in data analysis was conducted on these 61,700

   Table 1 gives an overview of the different degrees of stand damage.  A
classical foresty type differentiation was applied.  About 13.5% of the forest
was very heavily damaged or totally destroyed.  More than 60% of the forest
stands were influenced by the hurricanes and nearly 20% (first three classes)
were heavily changed in their stand structure.

   Table 2 gives an impression of how the main two tree species (spruce and
birch) were damaged.  Naturally the woods in the Saarland area would be
dominated by beech and oak, nearly without any coniferous species execpt fir
in small parts.  Before the hurricane events we actually had about 25% spruce
and 25% beech.  The damaged stands of spruce were mostly wind thrown like a
clear cut.  Instead, beech is mostly broken or fallen down as single trees.
Because beech needs about 200 years in rotation, its trees have risk of storm
damage in older ages than spruce which has a rotation time of about 80 to 120

Table 2.  Estimated distribution of hurricane clear cut areas in relation
          to the main species (spruce and beech) and to age classes.

Age class (years)       Spruce (ha/%)        Beech (ha/%)
-----------------       -------------        ------------
    21- 40                 500/13.5
    41- 60                1000/27.0
    61- 80                1000/27.0
    81-100                 500/13.5
   101-120                 200/ 5.4             100/ 2.7
   121-140                                      200/ 5.4
   141-160                                      100/ 2.7
   161-180                                      100/ 2.7
                          ---------            ---------
     Totals               3200/86.4             500/13.5

   In the coniferous group the weaker diameter classes are dominating, in
the deciduous group the stronger ones.  The complete data analyses were much
more complex than presented here.  They also show inside information of storm
events, about what is happening with different tree species under different
site conditions, stand structures, etc.  One of the results, for example,
showed that spruce was affected relative to tree growth area only 1.7 times
as much as beech, and relative to wood volume 2.5 times as much.  These
figures are much smaller than the damage ratios sometimes claimed for spruce
in relation to beech of 10 or 15.  The differences in damage between the two
species and in damage probability are not so much a problem of simple tree
growth areas or volume ratio, but of the probability of the different possible
kinds of damage.  Because of predominant single tree damage to beech and oak,
stands of these species are mostly affected in a kind of wind caused thinning
instead of the wind clearcut-like effects on stands of spruce.

   We estimated that about two million cubic meters of thrown wood in public
forest had to be harvested (private forest included, about 2.5 million).  This
was more than eight times the normal annual harvest amount of timber in the
Saarland.  Actual harvesting after the hurricane events was much more
extensive.  Therefore, about 1/3 of the useable timber remained in forests and
only 2/3 was taken out.

   The results are helpful in answering questions like how many different
stands are affected by wind throw and what kind of quantitative problem will
arise if all clearcut-like damaged areas are equipped with fences against deer
(estimated total number of fences = 3630; estimated length of fences = 1644
kilometers).  This is only one example of many other calculations we have
carried out about what kind of forestry management problems would need
solutions, if specified forestry concepts like soft silviculture should be
applied in cases of specific mixtures of artificial reforestation and natural
seeding on the clear cut forest areas.


   To improve estimates based on inventories, we still need more knowledge
about the probability structure and the statistical behavior of the data and
the accounted and estimated variables of interest.  Here some remarks on
variances and standard deviations of the variables of interest should be made.
At each inventory point one can estimate the variable of interest, for
instance the growing stock or the amount of broken and thrown wood.  In such
cases the variable of interest is some kind of probability variable with
standard deviation as one of its statistical parameters.  If there are n
inventory points falling into the selected stratum of points from which we are
calculating the mean value of the variable of interest, the variance of this
mean value only depends on n and on the variance of the single point variable.
As calculations have shown, many variables of interest in the permanent
sampling plot inventory design have a relative single-point variation in
relation to the mean value of about 50% to 100%.  If one knew the variance of
the variable of interest of a single point, one could carry out by standard
procedures calculations about how many inventory points (n) would be needed to
reach the necessary or desired precision of the particular resultant variable
in the data analysis.

   But there are still other reasons for variances in the data, which do not
depend on the variation of the single inventory point alone but also on the
probabilities originating from the distribution of the forest stratum in
space.  In cases of totally independent distribution of the stratum in space,
its probability in cases of large populations of possible inventory points
will nearly have the structure of a binomial distribution.  The relative
deviation of the space dependent part of the whole deviation of the variable
of interest of a single inventory point then depends directly on the
probability of the stratum in space.  If the stratum is about 25% of the whole
space area, the space dependent part of the relative deviation is nearly 175%
of the 25% mean value.  For small probabilities of the stratum in space (like
1% or less) this relative deviation will be a probability variable of a
Poisson-like distribution.  In the case of a mean value of 1%, it will have a
deviation value of about 1000% of the 1%.

   In the above case of a 25% space area probability, the space dependent part
of the deviation nearly doubles the original deviation part (see example 2
below) already calculated from the variable's direct deviation between the
points falling into the stratum (25% is the percentage part of public forest
area on the whole land area of the Saarland).

Example 1:  Estimation of the mean growing stock per hectare of the forest
            stands only requires the average value of all singe inventory
point values.  As relative deviation of the mean value we obtained 2.7%.  In
this calculation of variation, no components of variances originate from any
space distribution but only from the deviation between the single point

Example 2:  Estimation of the total growing stock requires further information
            about the total forest area.  This variable of interest is
calculated by multiplying the mean growing stock per hectare by the related
forest space area.  The total variance of this product variable is equal to
the sum of the variance of the per-hectare variable of example 1 plus the
variance of the space area.  We can estimate the variances originating from
the space distribution by dividing the total number of inventory points into
subpopulations.  In our example, we obtained as relative deviation a value of
4.1%.  Altogether we then obtained a deviation value for the total growing
stock of 4.9%.

   We could reduce the variance component originating from space area
estimation to zero, for instance by additionally applying a less expensive
specialized inventory scheme for area measurement only, like aerial photograph
analysis.  The results could then become considerably more accurate or less
expensive by putting all inventory resources into one single expensive ground-
based inventory scheme only.

                          DISCUSSION AND CONCLUSION

   Concerning the second ground-based inventory presented above, we are
convinced that it is very helpful for direct management in the forest
district:  for organizing the employment of manpower, the use of harvesting
processors and skidding machines, the transportation, the storage of damaged
wood, its selling, etc.  This method can also be used to maintain control over
all these activities.  Because the database is fixed on single forest stands
or damaged parts of them, it can be expanded for instance by data of actually
harvested volume of wood of different tree species, size and quality, by cost
data, by sales data, by reforestation data, etc.  Therefore, this inventory
scheme represents some kind of starting data scheme for a stand area fixed
information system of the wind damaged forest management in a district.  Later
it can also be used as a basic information system for correction or
continuation of the medium-term and long-term forest planning.  In combination
with the aerial photo analysis, this method also will reach the necessary
precision in determining area size of the new stand structure of the forest
district caused by the hurricanes.

   For general management of forest damage due to hurricane events, the first
inventory provides more accurate data than the second.  With this data,
processor machinery, storage facilities and marketing of damaged wood can be
prepared and organized at the general level of the forest administration.  The
inventory can be carried out quite rapidly, so that in a short time the
desired information from the data analysis is available.

   As a main result from the statistical point of view, we can conclude that
for many applications in forest inventory, especially if we are interested in
variables like how much wood is in stock or has to be harvested, it is of
great interest for practial and financial reasons to apply different schemes
for different inventory objectives.  For instance, as in the two examples
presented in this paper, it can be helpful and cost reducing to distinguish
between complex schemes for estimating the actual mean value per hectare of
the variable of interest and less expensive inventory schemes for estimation
of the actual space area of the selected stratum.

   The estimates based on inventories are quite accurate, delivering results
very quickly and economically, with proper and reliable information for
practical use.  The inventory schemes applied are of direct help for forest
management after catastrophic events.


Akca, Sangen-Emden R.  1987.  Zur Optimierung von reprasentativen
   Bestandesaufnahmen.  Allg. Forst- und J.Ztg., pp. 43-47.

Lohmann, H. and M. Heupel.  1990.  Ergebnisse einer Stichprobenerhebung
   uber Windwurfe im Saarland und Vorbereitung der Wiederaufforstung.
   Forst und Holz, Nr. 22, pp. 662-664.

Sabrowski, J.  1990.  Schatzung von Varianzen und Kunfidenzintervallen aus
   mehrstufigen Stichproben.  Schriftenreihe der Forstlichen der Universitat
   Gottingen 99, J.D. Sauerlander Verlag, Frankfurt.

Wolfe, B.  1990.  Verbesserung der Effizienz terrestrisher Kontrollstichproben
   durch zweiphasige Stichprobenverfahren.  Jahrestagung Deutscher Forstlicher
   Versuchsantalten - Sektion Biometrie, Gottingen.

Zohrer, F.  1980.  Forstinventuren.  Verlay Paul Parey, Hamburg.