CHAPTER 3: RESULTS

3.1. Land cover maps at scale 1:50 000
3.2. Land cover map at scale 1:5 000
3.3. State of vineyards/orchards for part of Sandanski region
3.4. Estimation of surface coverage of winter crops and grassland
3.5. Creation of DEM and slope class map
3.6. Statistics by administrative unit

The main activity of the project was the mapping of the present land cover at 1:50 000 scale for the three large test areas by using Earth resources satellite data and LCCS methodology. By adding to each mapped polygon the relevant soil type and erosion feature, a unique database was created.

Further, the project used very high resolution IKONOS satellite data for mapping land cover/land use in the Sandanski test area at 1:5 000 scale, to update linear features of the existing topographical map of the area and assess the state of vineyards there.

Other applications were also undertaken using Landsat TM data, such as the estimation of the winter crops and grassland surface coverage in the Plovdiv test area, the creation of a DEM and slope map for part of the Sofia test area and the extraction of statistic by administrative unit for a village of the Montana region.

All the above applications were carried out to demonstrate the flexibility and accuracy of remote sensing and GIS technologies in providing essential, updated information for resources mapping and management, mainly for agricultural purposes.

3.1. Land cover maps at scale 1:50 000

The 14 land cover maps prepared by the project cover a total area of 5 600 sq km, that is a fraction of the surface of the territory of Bulgaria, which encompasses 111 000 sq km. The three areas studied cover the following surface: Montana 2 400 sq km, Sofia 1 600 sq km and Plovdiv 1 600 sq km.

Using LCCS, 49 land cover classes were defined for the three study areas mapped at 1:50 000 scale. These classes can be subdivided into six major groups: Cultivated and Managed Land, Natural Vegetation, Artificial Surfaces, Bare Areas, Water Bodies and Mixed Units (Table 3).

In this study the agricultural land is divided into three classes according to the size of the land units:

- Very large herbaceous fields, more than 30 ha

This class has been subdivided into two separate classes, active and not active. ‘Active’ is defined as Maize, Sunflowers or other crops, which are actively growing during the period of image acquisition (early August 1998 for Montana and Sofia and early August 1999 for the Plovdiv area). ‘Not Active’ are fields after harvest, abandoned fields, grassland or weeds.

- Large herbaceous fields, 5 - 30 ha

- Medium herbaceous fields, 2 - 5 ha

Where it was not possible to separate a pure land cover class, 14 mixed classes were created. These units are defined as combinations of not more than three pure LCCS classes, for example - Grassland/Shrubs/Trees, Gardens/Low density urban areas etc.

Additional user defined attributes (Z attributes) have been used to accommodate specific local conditions and aspects of landuse.

There are two phases for defining a legend for a map of a particular area. Firstly a set of ‘classifiers’ must be developed and secondly environmental attributes must be added to these classifiers. In our study the additional attributes were soil types and erosion features.

Figure 7 shows an example of land cover map at 1:50 000 scale.

Table 4 indicates the distribution of the land cover classes in the three test areas.

3.2. Land cover map at scale 1:5 000

The Sandanski area was mapped using very high resolution (1 m - pan sharp) IKONOS satellite data and existing large scale topographic map at 1:5 000 scale.

A test area of 6.25 sq km northwest from the town of Sandanski was selected. Eight land cover/land use classes were interpreted. The area includes the new dam constructed ten years ago, part of the city, including the cemetery, part of the Rekichkata River and the little hill, called “Babunata”. The relief is hilly.

The very high-resolution satellite data were also used for checking and updating the drainage system (channels, riverbed, water bodies) and road network. Roads are unpaved.

The overlaying of the new digitized line features detected from satellite image over the existing large-scale topographic map at scale 1:5 000 showed differences mainly in the road network. New roads were detected; some of the old ones have changed their direction and shape, some did not exist anymore. The network of channels detected from satellite data is not so different from the respective network on the topographic map. During the field checking it was found that there is water in the channels, but most of them are filled with vegetation and have to be cleaned. In the places where roads were widened, parts of the channels have been destroyed.

Table 3: LCCS legend

Major Classes in the map’s Legend	Second Level Classes according LCCS	LCCS Classes
		“1. Very large herbaceous fields”
		“1.A. Very large herbaceous fields (active crop)”
		“1.B. Very large herbaceous fields (not active crop)”
CULTIVATED AND MANAGED LAND		“2. Large size herbaceous fields”
	A11. Cultivated and Managed Terrestrial Areas	“3. Medium size herbaceous fields”
		“4. Horticultures”
		“5. Gardens”
		“6. Vineyards”
		“7. Orchards”
	A23. Cultivated Aquatic Or Regularly Flooded Areas	“8. Rice fields”
		“9. Deciduous forest”
		“10. Coniferous forest”
		“11. Mixed forest”
NATURAL VEGETATION	A12. Natural and Semi-Natural Terrestrial Vegetation	“12. Patches of trees”
		“13. Grassland”
		“14. Riverine forest”
		“15. Mountain grassland”
	A24. Natural and Semi-Natural Aquatic Or Regularly Flooded Vegetation	“16. Flooded forest”
		“17. Wetlands”
		“18. Peat bogs”
	A11. Cultivated and Managed Terrestrial Areas	“30. City parks”
		“31. Park \”Vrana\””
		“19. High density urban area”
		“20. Medium density urban areas”
		“21. Residential complexes”
		“22. Industrial and other associated areas”
		“23. Airports”
	B15. Artificial Surfaces and Associated Areas	“24. Cemeteries”
		“25. Built-up areas”
		“26. Dump sites”
		“27. Extraction sites”
		“28. Tailing ponds”
		“29. Sport facilities”
		“32. Greenhouses”
BARE AREAS	B16. Bare Areas	“33. Rocky river bed”
WATER BODIES	B27. Artificial Waterbodies, Snow and Ice	“34. Artificial lake”
		“35. Ponds”
		“36. Large/Medium size herbaceous fields”
		“37.Medium/Large size herbaceous fields”
MIXED UNITS		“38. Low density residential areas/Gardens”
		“39.Gardens/Low density residential areas”
		“40. Vineyards/Grassland”
		“41. Orchards/Grassland”
		“42. Gardens/Grassland”
		“43. Grassland/Gardens”
		“44. Grassland/Shrubs/Trees”
		“45. Grassland/Shrubs”
		“46. Trees/Shrubs/Grassland”
		“47. Bare areas/Grassland”
		“48. Bare rocks/Grassland”
		“49. Woodland/Grassland”

Table 4: Distribution of land cover classes

Legend No.	Montana area - 240 000 ha		Plovdiv area - 160 000 ha		Sofia area - 160 000 ha
	Area (ha)	Percent (%)	Area (ha)	Percent (%)	Area (ha)	Percent (%)
1A	36 132.08	15.05	4 798.31	3.00	1 690.07	1.06
1B	38 369.94	15.99	10 850.76	6.78	9 892.07	6.18
2	33 478.25	13.95	17 431.23	10.89	9 241.41	5.77
3	19 009.92	7.92	30 606.90	19.13	3 966.79	2.48
4	540.57	0.22	*	*	21.05	0.01
5	1 976.59	0.82	8 672.35	5.42	3 287.38	2.05
6	2 748.42	1.14	1 064.03	0.66	*	*
7	239.40	0.10	1 969.03	1.23	181.18	0.11
8	*	*	743.79	0.46	*	*
9	13 776.43	5.74	4 747.22	2.97	31 155.05	19.47
10	205.23	0.08	1 542.56	0.96	10 084.59	6.30
11	172.56	0.07	2 056.88	1.28	4 489.10	2.80
12	190.00	0.08	70.86	0.04	112.58	0.07
13	22 546.07	9.39	10 209.29	6.38	15 250.87	9.53
14	6 129.27	2.55	3 986.95	2.49	1 769.68	1.10
15	*	*	*	*	5 473.72	3.42
16	52.55	0.02	*	*	*	*
17	424.98	0.18	99.47	0.06	1 041.65	0.65
18	*	*			123.50	0.08
19	*	*	224.60	0.14	681.76	0.43
20	22.98	0.01	1 697.46	1.06	5 102.96	3.19
21	67.52	0.03	622.55	0.39	2 895.21	1.81
22	147.81	0.06	1 431.30	0.89	5 076.97	3.17
23	310.89	0.13	483.21	0.30	473.32	0.29
24	*	*	24.35	0.01	133.06	0.08
25	442.67	0.18	742.27	0.46	268.95	0.17
26	*	*	29.99	0.02	86.39	0.05
27	71.59	0.03	69.12	0.04	1 834.66	1.15
28	*	*	*	*	305.04	0.19
29	*	*	63.22	0.04	122.83	0.08
30	*	*	232.29	0.14	1 132.43	0.71
31	*	*	*	*	111.09	0.07
32	*	*	69.02	0.04	*	*
33	*	*	*	*	66.31	0.04
34	555.29	0.23	1 798.49	1.12	976.58	0.61
35	195.35	0.08	384.85	0.24	226.81	0.14
36	18 147.05	7.56	25 101.36	15.69	6 572.03	4.11
37	14 298.88	5.96	9 059.43	5.66	1 428.05	0.89
38	9 946.52	4.14	8 096.18	5.06	10 239.86	6.40
39	1 089.07	0.45	247.59	0.15	2 211.06	1.38
40	242.29	0.10	63.15	0.04
41	*	*	521.21	0.32	96.79	0.06
42	604.92	0.25	348.35	0.22	784.84	0.49
43	*	*	*	*	669.45	0.42
44	6 206.02	2.58	2 922.56	1.83	6 935.56	4.33
45	6 574.22	2.74	2 463.31	1.54	7 277.91	4.55
46	3 814.02	1.59	731.79	0.46	3 704.31	2.31
47	1 270.68	0.53	*	*	1 984.14	1.24
48	*	*	*	*	820.90	0.51
49	*	*	3 722.72	2.33	*	*

Fig. 7. Example of land cover map 1:50 000 scale

Fig. 8. Linear features update for a large-scale topographic map by IKONOS data

The shape of the riverbed of the Rekichkata River has also changed. The new dam, situated northwest from the town is not reported in the existing topographic and cadastrial maps. The existing water impoundment, located in the eastern part of the study area has almost dried up.

The above shows that the existing large-scale topographic map needs to be updated and that very high-resolution satellite data could be successfully used for this purpose (Figure 8).

3.3. State of vineyards/orchards for part of Sandanski region

The IKONOS data were also used for checking and updating the state of permanent crops such as vineyards and orchards. Two types of cultivation technology for the vineyards were observed, namely:

• vineyards, covering large field sizes, planted on rows at a distance of 2-2.5 m from one another. The plants are tied on wires, which are supported by small concrete pillars. This is typical mainly of the former cooperative farms;
• vineyards covering medium size fields, planted in rows at a distance of 1-1.5 m from one another. Small wooden or concrete pillars support each plant. The plants are cut very short and only a part of the stem is left. This is typical mainly for the private farms.

The texture of the first type of vineyards and of the orchards is very evident on the images and a differentiation between cultivated and abandoned ones is thus possible. This was verified during the field checking. Analysis for a part of the first type of vineyards situated on the left bank of the river shows that approximately one third of them have been abandoned. Some of them have been completely removed and the land is now used for cultivation of other crops. On the right bank of the river, some of the vineyards were also removed. Vineyards remaining are of the second type; they appear dark red on the image without clear texture.

Most of the orchards, especially those along the Rekichkata River, are abandoned. Some of them were completely cut down. There are a few new ones, covering small fields (Figure 9).

The above clearly shows that very high-resolution satellite data provide important inputs in the updating of large scale land cover/land use inventory and monitoring.

3.4. Estimation of surface coverage of winter crops and grassland

The Landsat TM 5 acquired on April 1999 over the Plovdiv region was also used for estimation of the surface coverage of winter crops and grassland. At this time of the year winter crops and grassland are green and fully cover the fields. In a FCC of band 4, 5 and 3 (RGB) winter crops appear bright red and grassland orange, however it is difficult to discriminate pure training sets for a supervised classification.

An unsupervised classification was thus performed using TM bands 2,3,4,5 and 7, but the results were considered unsatisfactory as a high percentage of pixels were obviously incorrectly classified.

Therefore an NDVI image was produced from TM bands 3 and 4. This was added to the previous image as an additional layer using the layer stack function of ERDAS Imagine. The unsupervised classification was then applied again. The results showed a significant improvement on the previous classification. Fifty classes were created, then each class was analyzed visually, using as reference the FCC 4, 5 and 3 (RGB) and the resulting polygons were then assigned to one of the three following land cover types:

1. Winter Crops
2. Grassland
3. Others (all other land cover types)

From this final classification, surface coverage for each of the classes (winter crops and grassland) was obtained, by multiplying the number of pixels for each class by 900 m2, which is the surface covered by 1 pixel on the ground. The main purpose of this approach was the estimation of the area covered by winter crops and the establishment of a methodology for the assessment of winter crops coverage in the future. Figure 10 shows this application for a portion of the Plovdiv area.

The results of this exercise were also used to assist in the visual interpretation of the August 1999 image, covering the same area. Actually some field boundaries between different crops, are not clearly visible in the August image and consequently field sizes appear very large. However in the April classification, differences between winter and other crops are clearly evident and reveal much smaller field sizes. This information was used in the August 1999 image interpretation.

Fig. 9. State of vineyards for part of the Sandanski region, using IKONOS very high-resolution data acquired in August 2000.

Type	Area/ha/	Percent/%/
Old vineyards	68,12	65,31
Abandoned vineyards	33,26	31,89
Removed vineyards	2,92	2,80
Total	104,30	100,00

Fig. 10. Assessment of winter crops and grassland areas in the Plovdiv region

Type	Percent/%/	Area/ha/
Others	68.01	108830.70
Winter crops	18.82	30117.44
Grassland	13.15	21051.85
Total	100,00	160000.00

Fig. 11. Actual and potential erosion features of a portion of the Sofia region

Erosion	Slope (degrees)	Area (ha)	Percent (%)
	0-2	5077,98	50,78
None (n)	2-10	268,53	2,69
	10-15	0,25	0,00
	15-30	13,91	0,14
	>30	-	-
Total		5360,67	53,61
	0-2	183,086	1,83
None (n)	2-10	1003,327	10,03
	10-15	747,961	7,48
	15-30	1199,254	11,99
	>30	66,824	0,67
Total		3200,452	32,00
	0,2	319,041	3,19
None (n)	2-10	682,111	6,82
	10-15	88,572	0,89
	15-30	57,926	0,58
	>30	1,12	0,01
Total		1148,77	11,49
	0-2	3,629	0,04
None (n)	2-10	16,535	0,17
	10-15	11,657	0,12
	15-30	7,354	0,07
	>30	-	-
Total		39,175	0,39
	0-2	193,651	1,94
None (n)	2-10	31,54	0,32
	10-15	0,827	0,01
	15-30	-	-
	>30	-	-
Total		226,018	2,26

3.5. Creation of DEM and slope class map

For a selected test area in the eastern Sofia region, isolines at 10 m intervals were digitized in AutoCAD from a topographic map at 1:25 000 scale. They were assigned an attribute value according to their height in meters above sea level. The resulting dataset was then used to produce a DEM (Digital Elevation Model) using ERDAS Imagine software.

From the DEM, a slope map was produced, based on the following five slope classes namely 0 to 2 degrees slope, 2 to 10, 10 to 15, 15 to 30 and more than 30.

By overlaying the slope map and the LCCS land cover map, including the attributes of soil types and erosion features, the correlation between slope and erosion, both actual and potential, is evident. The resulting information definitely plays a very important role in land management mainly for soil conservation (Figure 11).

3.6. Statistics by administrative unit

By adding another layer to the existing database, that is the municipal and district boundaries, it will be possible to extract statistic (surface coverage, percentages) on all the classes occurring in a particular area. However our main study covered only three test areas and thus this further application was not always possible.

To demonstrate this very important aspect of the new database created by the project, statistics were extracted from the land cover GIS database for the village of Rassovo in the Montana test area. The village boundary was overlaid with the land cover layer and the surface for each land cover class occurring in the village territory was then calculated.

It was found that the legend used for cadastrial purposes does not correspond with the LCCS legend for all classes, however some classes are the same, for example vineyards. The area of vineyards calculated from our land cover map is of 37.7 ha, while the area reported by the cadastrial data is of 39 ha, which indicates very good correlation between the two data sources.