The methodology to execute above mentioned concepts and to achieve the requested tasks can be split into two major parts:

• Methodology to conduct land use surveys digitally
• Methodology to interpret these land use maps and to come up with land allocation maps and monitoring maps

If aerial photos are the primary data source, aerial photo interpretation will be undertaken with delineation of homogenous ‘API units’ with stereoscopes. This is followed by fieldwork, where for each defined mapping unit (not for each polygon) 1-10 samples are taken and their land use is described. Such a description both in terms of land use indicators (e.g., distance of houses, height of trees, etc. for the functionality: 'land use', not only 'land cover') as well as in terms of predefined land use classes is important because these data will be used for medium and large scale survey. Additionally, land use indicators (descriptions) can be used to check vs. the classification.

The delineation of API will be digitized (using software module ILUDArc). Photo codes, which do not directly indicate the land use, are entered as polygon identifiers. This is called ‘labeling’ and is handled by module ILUDLab.

Field data are entered in the DBMS with module ILUDEntr, processed with ILUDProc to assess the relationship between the photo interpretation codes and the actual land use (from the field descriptions). The result indicates, which land use class(es) occur(s) in each photo unit. This is then linked to the GIS data (with ILUDLab).

The linked maps contain information about primary and secondary land use with their ratio, their numerical code, and their character code (abbreviation). It includes additional parameters such as the number of the province, the year of the survey, a quality control figure, and the software version (see Annex for a sample of the attribute data of the land use maps).

All this information is automatically added during the linkage procedure with module ILUDLab. Topology is complete. Due to the size of Indonesia stretching over 9 UTM zones (90º - 144º E), and to keep all georeferences unique, data are always stored in geographical units (latitude/longitude).

The map can then be plotted with module ILUDPlt, which includes a formal data integrity check (data completeness), for complete topology, for correct georeference etc. This is part of the quality assurance of the digital data. The plot can be done on a plotter or printer in any size. Alternatively, final maps can be transferred to other graphic formats (see also Part 7.2 and Annex).

The plotted hardcopy maps follows a standard map layout based on the national grid system. Optionally, maps can be plotted based on administrative boundaries, e.g. for a District or a Subdistrict (with preparation through ILUDAdm, which ‘clips’ and aggregates map sheets).

Care is taken for these cases if two substantially different land uses occur in one mapping unit, i.e. sparsely settled villages between fruit trees (in Indonesia called 'mixed gardens', or fish ponds in a mixed pattern with paddy rice fields). These cases can not be differentiated by photo interpretation, but they are often obvious in the field. These are 'complex units' and are identified both in the databank (with their ratio) and on plotted land use maps.

Various alternative approaches have been practiced and are installed in the system, such as digitization from aerial photographs, digitization of observation points from the tablet, transfer of observation points from DBMS to GIS and v.v., different generalization levels, selection between two land use classification systems, conversion from digitizer units to lat/long or UTM, manual editing of labels at map plot, etc.

(1) dB: dBase - Cl5: Clipper5 - AI: ArcInfo - AE: ArcEdit - AP: ArcPlot

If the software modules are being applied, there is no concern for deviation from the standard and the application of the data dictionaries: Automatically, the programs (ILUD modules) apply the standard codes and the data become formally integer and clean.

Sample for Land Use Map: West-Jawa (Karawang)
(Click map for large size display)

Sample for Land Use Map: Kalimantan (1614-522)
(Click map for large size display)

5.1b Land use mapping from satellite images

If satellite images are available, image processing of the scenes will be done in Erdas (7.5 or Imagine) with ILUD image processing menu and spectrally classified.

The image processing includes data import, enhancement, geometric correction, unsupervised classification with 27 classes, check of GIS file signature with image LAN file, optionally merging of signatures of obvious identical land use classes, assignment of new colors, and finally filtering of 3x3 for TM images.

The approach of an unsupervised classification was chosen because of the inexperience of the staff, the complex land use patterns, which can not be differentiated with spectral classification signatures, and the stress of large scale inventory and mapping for land use planning and land allocation permits. Collection of field data is not a major problem at the institution and results in more reliable data about land use definitions.

After conversion to vector format in the GIS (with module ILUDR2v), this ‘spectral classification map’ is taken to the field and serves as a basis for the field data collection. Further processing is identical to above described procedure from the aerial photo interpretation, including the ‘labeling’ for land use classification, field work, and field data entry.

 

Tab.3: ILUD Cycle for Land Use Mapping from Satellite Image

Step	Task	Com-ponent	Module	Result	Time (1)
1	Image processing	IP	ILUD image processing menu system	Classified image file	½ - 2 days
2	Transfer to GIS (vector)	GIS	ILUDR2v	Spectral classifi- cation coverage	½ - 4 hrs
3	Plotting of remote sensing result	GIS	ILUDPlt	Hardcopy map of spectral image	20-40 min
4	Field work	-	(manually)	Field data sheets	1-4 days (2)
5	Entry of field data	DBMS	ILUDEntr	Data file with all land use descriptions	½-2 days (2)
6	Processing of field data	DBMS	ILUDProc	Data file with info about photo codes	10-30 min
7	Labeling of polygons (per spectral class)	GIS	ILUDLab	Map coverage with labels and topology	15 min
8	Link DBMS - GIS	GIS	ILUDLab	Digital land use map	5-20 min
9	Plotting	GIS	ILUDPlt	Hardcopy map	20-40 min
10	Backup	GIS	ILUDUtl	Backup copy	1-3 min

(1) Per map sheet

(2) Data collected/processed per ‘study area’, time given on pro-rata basis

 

Tab.4: Individual Processes of Land Use Mapping from Satellite Images

 

	Step	Process	Processes	Commands
	1	Image processing	Data input Enhancement Geometric correction Unsupervised classification Check of GIS file signature Filtering	E: bstat E: gcp - coordn - nrectify E: isodata E: colormod E: scan - majority filter 3x3
	2	Transfer to GIS (vector)	Convert to vector format Clip in two steps (+10%, precise) Standardize fuzzy tolerance Standardize file structure Read color assignments Eliminate small units Unique user ids Merge lines Smoothen lines   Convert to lat/long	AI: gridpoly AI: clip - build delete tol. AI: additem Pascal program AI: eliminat AI: tables.calculat - idedit - build AI: renode AE: unsplit AI: generaliz AE: grain (3 times) AI: project cover
	3	Plotting of remote sensing result	Convert to UTM Creation of plot file with all features Plot	AI: project cover AP: displ 1039 hpgl2 - x.hp2
	4	Field work	(same as API)
	5	Entry of field data	(same as API)
	6	Processing of field data	(same as API)
	7	Labeling of polygons (per spectral class)	Build topology Assign labels	AI: build AE: ef lables - sel for ... moveitem
	8	Link DBMS - GIS	(same as API)
	9	Plotting	(same as API)
	10	Backup	(same as API)

(1) dB: dBase - Cl5: Clipper5 - AI: ArcInfo - AE: ArcEdit - AP: ArcPlot - E:Erdas 7.5

Sample for Spectral Classification Map: (2305-423)
(Click map for large size display)

---

  Next chapter

---