¹ Agriculture and Regional Informations Systems Unit
Space Applications Institute
Joint Research Centre
21020 Ispra
Italy

² ECVAM
Institute for Health and Consumer Protection
Space Applications Institute
Joint Research Centre
21020 Ispra
Italy

OliCount 2000 - Millions of Trees for the Millenium

Abstract

Under European Community legislation, Member States with olive cultivation are required to implement a Geographic Information System for management of crop declarations and aid applications. A specific requirement within the regulations is the establishment of a comprehensive database, including digital orthophotographs, land parcel boundaries and the location of individual olive trees. In support of the Member States who are faced with this task, the Joint Research Centre of the European Commission has previously investigated technical aspects of computer-assisted olive tree identification from digital orthophotographs, resulting in "OliCount" - a tool implemented in C++ and delivered with an ArcView interface. "Olicount 2000" is a project building on this earlier work, aimed at addressing the additional problems related to the requirements of the Member States now that they are approaching the stage of designing their systems. These issues include the calculation of areas under olive cultivation, the integration with existing systems (most importantly the national Parcel Identification System) and the management of between 5 and 300 million trees depending on the Member State. This paper illustrates the problems faced with some real examples, demonstrating in the ArcView environment the integration of the OliCount algorithm with a cadastral and declaration database, the development of a custom interface with Dialog Designer and the OliArea module for calculation of areas.

Introduction

Olives, and olive trees dominate many aspects of the Mediterranean landscape and economy. The evolution of technology and policy over the last twenty years has led to a situation where it is both feasible and necessary to have an accurate, georeferenced register of olive tree cultivation. Currently European Union Member States are involved in varying stages of GIS implementation to support this requirement. The Joint Research Centre (JRC) of the European Commission is actively supporting these activities through applied research and technical assistance. During 1997/8, The Agricultural and Regional Information Systems Unit (ARIS) of JRC developed an algorithm, "OliCount", to support one of the major tasks of creating an olive tree geo-database – the identification and location of the trees. Additional projects have been undertaken since then, namely "OliStat" – a statistical assessment of the total number of trees in each Member State, and "OliArea" – the calculation of areas under olive tree cultivation. OliArea also led to the development of the concept of an "Olive Tree Parcel". These related activities are now coordinated under the title "OliCount2000", which looks more closely at the requirements of implementing an Olive Tree GIS, integrating cadastral, digital orthophoto and declaration data with OliCount and other computer-assisted photo-interpretation tools. An ArcView prototype has been developed and is described here.

The Policy Framework

The first piece of legislation referring to olive tree registers dates back to 1975. European Council Regulation R154/75 states that - "in order to obtain the information needed to determine the Community's potential production of olives and olive oil and to improve the operation of the Community aid system for the latter product, producer Member States should be required to establish a register of olive cultivation". This register is intended to be at the level of an individual holding, and should include:

• total growing area together with the cadastral reference of the parcels
• total number of trees
• agronomical and socio economical characteristics

Subsequent regulation R2276/79 of the European Commission required checking aid requests against the register database, and for the first time mention is made of the use of orthophotographs with cadastral boundaries marked upon them.

However, the subsequent accession of Portugal and Greece, neither of whom had a cadastre, caused problems in the application of these regulations.

Commission Regulation R2366/98 supersedes all previous legislation, clearly stating that MS producing olive oil should implement a 'olive cultivation GIS', with a computerised alphanumerical and graphical database, comprising digital orthophotos, parcel identification system boundaries (cadastral or other) and the location of individual olive trees.

The requirement here is to develop a truly integrated system, where an applicant’s request for aid can be checked directly against the number of trees registered based on their annual declaration.

This requirement appears at first to be reasonably straightforward, however the fact that there are estimated to be in excess of 700 million olive trees to be located, covering a total area of 5.45 million hectares, means this is a significant task.

Technical Solutions

ARIS has developed a prototype called "OliCount2000" which integrates the two main components described above - (i) OliCount and (ii) OliArea, which has been demonstrated to Member States involved in the "Olive Cultivation GIS" ("GIS-OLI"). The aim of OliCount2000 is not to impose a specific technical solution on any Member State, but rather to demonstrate the fundamental GIS operations which should be expected in such a system, and the benefits of a harmonised approach, specifically for the concept of the Olive Tree Parcel.

ARIS developed OliCount from 1997 onwards in the context of the OliStat Project in response to the fundamental problem of olive tree identification and counting. The objectives of the product were to:

• Collect existing know-how on automated or computer-assisted counting of olive trees on digital aerial photography
• To develop an algorithm based on the knowledge collected
• To implement this algorithm and integrate within an existing GIS application environment
• To pass on the knowledge, algorithm and software tools to interested organisations

The resulting OliCount product consisted of the algorithm, implemented in C++ and distributed as a DLL (Dynamic Link Library), plus an ArcView project. ArcView provided the basic GIS functionality, a customisable user interface and the ability to launch the DLL with parameters derived from the ArcView environment. Olive tree identification is based on a set of input parameters, which essentially define the morphology (shape) and radiometry (grey level):

• The minimum and maximum grey values ("thresholds") in the image which distinguish tree from background
• The size range (diameter) of the tree
• The aspect ratio of the tree crown
• The density range within each tree "envelope" (i.e. the bounding rectangle). This is a measure of the size of holes in the tree crown relative to the overall tree area.

ArcView provides the following basic functions:

• Visualising the image(s), parcel(s) and selection of the area of interest (domain).
• Automatic identification of the rectangle which bounds an irregular parcel polygon.
• Dialogue boxes for the entry of the parameters used by the OliCount algorithm.
• Automatic geographic selection of the appropriate image(s) which are in the area of interest.
• Clipping of the result (a rectangle) to the parcel polygon(s).
• Display and interactive editing of the result (adding and deleting trees).
• Connection to external databases for storage of results

A typical result from OliCount is shown in figure 1. Note that the algorithm performs counting according to the bounding rectangle of a polygon, ArcView subsequently performs clipping to the polygon boundary.

OliCount was used directly for the OLISTAT project. OLISTAT was launched by the Commission to assess the total number of olive trees in Europe. A statistical analysis was performed in each of the 5 Member States involved, based on an area-frame sampling of circles of 100m diameter ("placettes"), within which a combination of manual and computer-assisted photointerpretation techniques were applied. ARIS performed the quality control on the results of this project, and developed a custom application in ArcView incorporating OliCount once more and using the "Dialog Designer" extension. OliStat increased the estimate of trees from 450 million to in excess of 700 million.

Because of the twin requirements for identification of tree locations and calculation of area under olive cultivation, and the requirement of olive tree area calculation by Directorate General Agriculture, ARIS developed "OliArea". Taking into account basic definitions stipulated in R2366/98 (Olive Tree Parcel Definition and Scattered Tree Area), and using the results of OliStat, with trees located within individual placettes, a technique was developed which performed the following:

• Creation of a network between trees to create groups
• Definition of an olive tree parcel according to buffers around the generated network
• Calculation of voronoi polygons, giving an area per tree

OliArea has been implemented using ArcView, Dialog Designer and Spatial Analyst. The latter extension was required for calculating voronoi polygons.

OliCount2000 is designed to pull together these related strands of work. The project presents both a prototype system, drawing on our previous experiences and a new methodology for the logical grouping of trees – the "Olive Tree Parcel".

Figure 2 shows the typical interface for the first stage of using OliCount2000. An application for aid is selected, and the relevant cadastral parcels for this declaration are selected. In the case where cadastral parcels are available in vector format, the extent of the relevant polygons can be used to zoom to the appropriate area. Where the cadastre has been scanned, a single point must be digitised within the parcel and recorded to perform a similar orientation function. Note that although the cadastre is used for orientation, it does not define the precise boundary for any of the olive tree identification functions. Of course, an Olive Tree Parcel may correspond to a single cadastral parcel. If the location of trees has not yet been performed, then an interpretation phase follows. Thus the operator defines an area of interest (highlighted in red), and launches OliCount according to a set of parameters. The OliCount result can be edited, by adding or deleting trees, and the operator can optionally display the total number of trees identified and compare to the number declared. Once the photo-interpretation is finished, the two totals are declared and tolerances applied. The tolerances vary according to the total number of trees. Other declared parcels within the same application for aid are then checked, and an application level tolerance applied. Applications with > 3% difference with respect to the interpreted totals will be subject to a field visit.

Figure 3 shows the subsequent phase - identifying the Olive Tree Parcel. Trees within 20 m of each other are connected by a network (shown in red). A buffered polygon (magenta) shows the limits of the Olive Tree Parcel, which becomes the clipping boundary for the voronoi polygons. Since the original OliArea work, the process of voronoi calculation has been converted from raster to vector, and the Spatial Analyst extension is no longer required for this aspect.

This new concept of the Olive Tree Parcel is a response to the problems associated with a system based on cadastre, namely :

• It is quite specifically not the objective of GIS-OLI or IACS to finance the update of national cadastres.
• Cadastre boundaries invariably present problems in terms of position, distortion, etc., and rarely fit the images displayed underneath them.
• Olive Tree Parcel identification - particularly with the use of a tool like Oliarea - is no more difficult than the vector digitisation of an agricultural parcel boundary.
• More problematically, any divergence between the cadastre and reality must be corrected by the cadastre service - and this takes (usually) a lot of time (years).
• The use of a cadastre when a GIS is involved does not save money; it complicates and makes things more expensive.
• The use of the cadastre is effective when used as the identifier in an alphanumeric system, but not when a geographical system becomes available.

Multiple reference layers can be handled, but once a system specific system has been created (e.g. agricultural parcels), then the other "reference" system (e.g. cadastre) has no technical relevance; indeed it is a hindrance.

The Future

We hope that Member States can examine OliCount2000 in detail and adopt (or adapt) any of the elements that benefit them. This could take the form of direct implementation, for example using the OliCount DLL directly as part of their own GIS-OLI, or by adapting our ideas of user interface design for the aspect of checking declarations, and most importantly by adopting the idea of the Olive Tree Parcel.

The results of this work and ongoing status can be examined at http://mars.aris.sai.jrc.it/wine_olive/olicount2000/prototype/proto.html