LAND COVER CHANGE KNOWLEDGE REPRESENTATION USING TEMPORAL LOGIC AND OPERATION RELATIONS

The knowledge representation model plays an important role in the dynamic extracting and web service for Land Cover Change. While former researches were mainly focus on the represent of Spatio-temporal data itself, fewer of them attended the relations of Land Cover Information associated with change. This paper focused on the Land Cover Representation Model, and presented a knowledge representation model for LCC using temporal logic and operation relations. Firstly, we analyses the new demanding of change information service for web publishing, secondly, we discuss the temporal logic of snapshots and increments; then we introduce a method to represent the operation relations between snapshots and increments; finally we construct a knowledge representation for Land Cover Change based on the temporal logic and operation relations.


INTRODUCTION
Land cover refers to the natural properties and characteristics of a variety of material types covering the surface of global.Since the Land Cover Change (LCC) is one of the significant resources for scientific research (Herold et al., 2008;Xian et al., 2009;Chen et al., 2011), many institutions have built their own land cover websites to fulfill the needs of researchers and public (e.g.USGS, UMD et.al).However, former websites are mainly focused on static information, for example the accessing service, index service, and functional service, which are isolated and incompatible.While the distributed invocations of LCC service require not only static information, but also to provide change services and assemble the services dynamically, which need us to establish knowledge representation model of the Land Cover Change for the dynamic extracting and web service for Land Cover Change.

Despite
the technical advances towards spatial-temporal data model, issues related to the knowledge representation of it only recently have been to draw attention (Song et al., 2009;Batsakis et al., 2010).But fewer approaches take, directly or indirectly, the relations of land cover ingredients into account (Turner et al., 1995;Nunes, 1999;Lambin et al, 2003).Indeed the temporal and operation relations However, their approach was taken with the computation of spatial data into a more general level and didn't take the temporal relation into consider.The present paper addresses a more specific problem, the knowledge representation focusing on snapshots and increments using temporal logic and operation relations.LCC information poses a knowledge representation problem when extracting implicit information dynamically.To the authors' knowledge, the temporal and operation relations are inherent among snapshots and increments.For example, we can get a snapshot of land cover by a snapshot in other time overlapped with some increments, or get increments by multi-snapshot using the erase operation.
The remainder of the paper is organized as follows.
Section 2 discusses the temporal logic among snapshots and increments using matrix arithmetic.
Section 3 introduces set operation algorithms into the spatial operations of LCC, and analyses the basic 5 operation relations among snapshots and increments, while section 4 describes the knowledge representation for LCC based on the proposed methodology.

TEMPORAL LOGIC AMONG SNAPSHOTS AND INCREMENTS
Temporal logic refers to the logical relations among temporal objects.According to the provisions of ISO19100 ( 2002), temporal object can be divided into "time slice" and "time interval", while "time slice" represents one point in time space, and "time interval" denotes a line with start point, end point and length attributes in time space.Presently, the most popular temporal relations are the 13 relations proposed by Allen (1983) which are shown in Table 1.
Table 1 Allen's 13 temporal relations But they are mainly for the relational expression among time intervals.However, LCC information is recorded and expressed primarily through a series of snapshots or increments; while a series of snapshots record the change of a set of discrete "time slice" and increments convey the change within a certain "time interval".Therefore, in the expression of temporal logic for LCC, the temporal relations between "time slice" and "time slice", "time slice" and "time interval", "time interval" and "time interval" must be considered.
In this paper, the snapshot of Land Cover is represented as and the increment of Land Cover is represented as .For two snapshots , , their temporal relations have 3 cases as follows: ， .
Hence, the temporal relation between and can be defined as which can be abbreviated as .
The temporal relation between and can be obtained through comparing with and respectively.Therefore the temporal relation of and , which could be represented as , can also be shown as below when the relational matrices is adopted to the temporal relations : Because we have defined , so in matrices (1) the does not exist when and , which can be marked as .Therefore, the formula can be simplified as follow: When , and , the value of is inexistence ,which can be marked as .
Hence, the formula can be abbreviated as follow: According to Allen's 13 kinds of temporal relations, we could also acquire Rule 1: Every in formula (2), when or , the value of it is .
So the can be further simplified as follows: That is to say, the can be written as . When we substitute formula (5) into formula (4), we can also get the simplified temporal relation of as follow:

INFORMATION
In the previous section, the temporal logic of LCC information is mainly analyzed.But, besides temporal relations there are also spatial operation relationships between snapshots and increments.For instance, we can get the change of Land Cover by the comparing of tow snapshots of the same area, and we can also obtain a newly snapshot by adding some increments to an intrinsic snapshot.Therefore, the spatial operation relations that may exist between snapshots and increments in the expression of change information knowledge should be described.
And we can also acquire B by the combinatorial operations of formula and , which is show as follow: Table 2 shows the semantic description of above five operation relations among snapshots and increments Table 2 The semantic description of operation relation among LCC (snapshots and increments) and the computation relations between land cover change and static information are the main points for constructing LCCSP, which ensures the dynamic computation and accessing of implicit information.In previous work, Welty et.al (2006) added temporal relations to TimeSlice and TimeInterval for the representation of temporal objects, which may support the service composition and dynamic extraction of implicit information, while their approach aimed at incontinuous change information and didn't contain the operation or computation relation between spatial data.Di and Ramapriyan (2010) in George Mason built the computation relation of geo-information and functions, and provided a series of computation models to meet the needs of research institutions.
and .Consequently the temporal relation among increments can be written as , which can be simplified as .While because the change of Land Cover information is continuous, and we can't divide the change into a minim time element, so the snapshots participate in dynamic computation must be started or finished by the increment.So we can add rule 1 to formula (2) and get the simplified relations.
Figure 2-(b) shows the snapshot data of grass cover in 2010 and Figure 2-(c) represents the reduced grass data from 2000 to 2010.If 2-(c) did not exist in the database, it could be obtained through the operation of 2-(a) and 2-(b).

Fig 2 .
Fig 2. Change Information of Land Cover 2-(a) shows the snapshot of grass cover in 2000 2-(b) shows the snapshot of grass cover in 2010 2-(c) shows the decreased grass from 2000 to 2010 Fig 3. Representation Model of Land Cover Change The model contains two main subclasses: SnapShot and Increment, while SnapShot includes hasGeometry and hasCategory property, which separately link GeometrySnapShot class and CategorySnapShot class.Besides, SnapShot can aslo contain other static properties.Through SnapShotOf property, GeometrySnapShot and CategorySnapShot can link Geometry class and Category class respectively.Increment class is defined by start and finish SnapShot separately, which are marked as StartSnapShot and FinishSnapShot properties.Increment has independent Geometry property.Furthermore, Increment class aslo stores spatial operation relationships between start and finish SnapShot.The Spatial Operations are marked as Increment's SpatialOperationOfSnapShots property, which is defined by SpatialOperations class.SpatialOperations class includes OperationName property specifically including erase, erased, intersection, union and symmetric difference etc 4 CONCLULSIONS