nbul 2004
dispersion
s It cannot
'rscanning
"s no way
ie number
mpetition
loor, setup
nark them
ist several
recovering
long-term
represent
, to allow
ing areas,
is different
man-made
40 x 40 m
one, plant
il samples
International Archives of the Photogrammetry, Remote Sensing and Spatial [Information Sciences, Vol XXXV, Part B4. [Istanbul 2004
From 1998 to 2000 lots of data on those monitoring areas were
gathered to find answers to the question whether there are
enough seedlings to assure reforestation after the devastating
bark beetle infestation and dying of lots of forest stands.
Besides seedlings all other kinds of vegetation in the test areas
were included into data acquisition to allow an estimation of
growing conditions and vegetative competition (Bauer, 2002).
The data was statistically evaluated using proved methods,
modells and hypotheses. E.g. the spatial distribution of events
was calculated using the CLARK & EVANS-Index, that gives
the relation between measured and expected distance to the
nearest neighbour. It is a measure for regular or clustered
horizontal distribution. A clustered distribution will give values
below 1 whereas a complete hexagonal distribution produces
the maximum value of 2,15.
So far no analysis of this rejuvenation data has been performed
using GIS and database technology, particularly in regard to
new possibilities provided through object-relational features
and data mining. Since database models get closer to real world
objects and events by allowing the setup of an integrated and
adjusted representation based on user-defined types, methods
binded to database objects, and further declarative information
that resides in XML, a complete description of spatial objects in
database systems can be provided for the first time. Therefore it
is obvious to make use of object-relational structures provided
by the underlying DBMS to develop a new approach to examine
bark beetle spread and forest rejuvenation.
In the following a brief introduction of object-relational
modeling concepts is given. As the implementation is based on
an ORACLE 9i database, the corresponding SQL -dialeet is
followed.
2. OBJECT-RELATIONAL FEATURES
Within the classic relational database model there are only
scalar but no complex data types. With the introduction of
object types the definition and composition of abstract data
types is possible. An abstract data type can be comprised of a
multitude of scalar types and again of user-defined complex
types and enhances consistency when creating database models.
create or replace type POSITION TY as object (
x | NUMBER (9,2),
y NUMBER (9,2)
create or replace type TREE. TY as object (
position POSITION. TY,
species VARCHAR?2 (30),
plantation_date DATE
I
Data types only represent descriptions of data structures. To
ensure persistency a table must be bound to the data type. A
relation with column objects can be created that is a set of
young trees with tree ids as:
create table TREE. TAB (
tree id INTEGER,
tree TREE TY
In addition to representing column objects object types can also
be used to define row objects that can be managed using object
tables:
create table TREE OT of TREE TY:
Object views allow users to treat relationally data as objects as
they allow to synthesize objects from data that continues to be
stored in relational tables. Object Views are therefore often
refered to as “natural bridges” between both paradigms. Object
views have similar functionality like object tables. They can
have methods. belong to collections. reference one another,
have object identity and can be accessed from SQL. In addition
tables that are assigned to object views can be updated by using
special instead of triggers.
Any instance (row) of an object class contains a unique ID
called object-ID (OID). Gernerally OIDs are system-generated
but can also be derived from a primary key column or can be
user-defined.
Relationships between objects can be defined using reference
types. A reference column stores OIDs of associated (row)
objects since column objects do not have inherent OIDs and
therefore cannot be referenced. Row objects that belong to a
reference can be selected and dereferenced using the DEREF
rsp. VALUE operator. Modeling object relationships with OIDs
and REFs is often compared with foreign key relationships
inside the relational model but implicates some benefits like the
ability to distinguish between equal and identical objects.
Objects are identical if they have one common OID. They are
equal if they have different OIDs but coincide with their
attributes and values.
In the classical Entity-Relationship-Model aggregations and
compositions are modelled through master-detail-relationships.
Object-relational dbms provide collection types that contain
multiple elements and thus are suitable to express l:n
relationships directly. Each element or value for a collection has
the same substitutable data type. The most popular collection
types are varrays and nested tables.
A varray contains a variable number of ordered elements.
Varray data types can be used as a column of a table or as an
attribute of an abstract type.
Named table types can be created in an Oracle database using
SQL. These table types can be used as nested tables to provide
the semantics of an unordered collection. As with varray a
nested table type can be used as a column of a table or as an
attribute of an object type.
create type TREE_NT as table of TREE_TY;
Multi-Level-Collections that lead to multiple nested tables can
be realized if useful for applications but it's up to the user to
balance — a more intuitive representation of data vs. higher
complexity of accessing the data.
An object type declaration can also include methods that are
defined on values of that type. When using these object types in
tables their methods are also applied to the data of these tables.
The method is declared in the create type statement and the
code for the function itself (the definition of the method) is in a
separate create type body statement.
|
|
i
_,. non .anr
