IntechOpen

Home > Books > Vegetation

Open access peer-reviewed chapter

Comparing NDVI and Corine Land Cover as Tools for Improving National Forest Inventory Updates and Preventing Illegal Logging in Serbia

Written By

Jovanović M. Miomir, Miško M. Milanović and Bojan R. Vračarević

Submitted: 11 May 2017 Reviewed: 23 October 2017 Published: 20 December 2017

DOI: 10.5772/intechopen.71845

Vegetation IntechOpen
Vegetation Edited by Allan Sebata

From the Edited Volume

Vegetation

Edited by Allan Sebata

Book Details Order Print

Chapter metrics overview

1,269 Chapter Downloads

View Full Metrics
Advertisement

Abstract

National forest inventories (NFIs) in Serbia have been carried out very rarely (every 20 years), while the last two official estimates of forest areas (for 2011 and 2014) are very imprecise, because they are based on the cadastral data (and Serbia is well known for the lack of cadastre updating). Although forest conservation policymakers in Serbia still have limited financial, human, and political resources, over the past two decades, publicly available, remotely sensed satellite data on deforestation and degradation have dramatically reduced evaluation costs. Since municipalities in Southern Serbia experienced a 15% loss of forest area in the 2006–2014 period, as the obvious result of forceful, rapid process of illegal logging, this study evaluates the possible use of two remote sensing techniques: normalized difference vegetation index (NDVI) and CORINE land cover (CLC) databases for preventing illegal logging in Serbia. It clearly shows that NDVI is very promising for Serbia and also for other post-socialist countries that very rarely carry out national forest inventories (NFIs), and where unrecorded, illegal logging can exceed the legal harvest by a factor of 10.

Keywords

  • CORINE land cover
  • illegal logging
  • national forest inventory
  • normalized difference vegetation index
  • Serbia

1. Introduction

Like many other developing and post-socialist countries where the basic information on the current state of forests and other ecosystems is often inadequate, fragmentary, or outdated [1], in Serbia national forest inventories (NFIs) have been carried out very rarely, at roughly 20-year intervals: in 1961, 1979, and 2003–2006. Since 2006, official estimates of forest areas have been made only for 2011 and 2014, but they are very imprecise, since they are based on the cadastral data (and Serbia is well known for the lack of cadastre updating) [2, 3].

While Serbian state-owned forests (48% of the nation’s forest resources) are managed mainly by the state forest enterprises, according to the management plans prepared on the 10-year basis, the basic information on the current state of private forests is in a much worst condition. Previous forest census completed in 1979 covered only state forests and national parks, without even including private forests (new inventory completed in 2006, finally included private forests). Private forests constitute 52% of the nation’s forest resources, and are characterized by very small plots (average size: 0.3 ha). Nevertheless, Serbia adopted following official definition of forest area: “The forest includes all the inventory unit areas larger than 0.5 ha covered with forest trees…” [2], which is very similar to FAO definition [4]. Actually, NFI features for individual countries have been developed over time to accommodate their unique topographies, climates, forest types, commercial interests, etc. [5, 6, 7]. Although FAO’s reference definition is used as the basis for the national definitions of forest by many countries, national definitions can vary considerably: minimum forest area in Czech Republic is 0.04 ha, Austria (0.05 ha), China (0.0667 ha), Germany (0.1 ha), Estonia (0.1 ha), Ireland (0.1 ha), Latvia (0.1 ha), Lithuania (0.1 ha), Luxembourg (0.1 ha), Slovenia (0.25 ha), Slovak Republic (0.3 ha), USA (0.4 ha), etc. [8]. Unfortunately, 0.5 ha does not fit Serbia well, since privately owned forest parcels, which account for half of the total forest area of the country, cover much smaller areas—average private holding is only 0.3 ha [9].

Since illegal logging contributes up to 30% of the global market, in excess of US $20 billion a year, we cannot rely on official production statistics to capture deforestation. For example, even the latest forest resource data from official statistics in Central and Eastern European countries often do not consider forest degradation and illegal logging [10]. World Bank estimates that unrecorded, illegal logging in some of Central and Eastern European countries, like Albania, exceeded the legal harvest by a factor of 10 [3]. Also, in Serbia there are no unique records about illegal activities in forestry [11]. Illegal logging in Serbia is most intense exactly in the areas adjacent to the territories of Kosovo (Kosovo is ranked as the one of the worst illegal logging offenders in the world with Indonesia [12, 13]) which are formally under Serbian forest estates Vranje, Kursumlija, Leskovac, Raska, and Leposavic), territories to which Serbian authorities have limited access [3, 14, 15].

In order to prevent illegal logging, it is obviously most relevant to concentrate on areas where the illegal logging is most intense, and it is exactly in the southern Serbian municipalities, like Kursumlija, that are adjacent to the territories of Kosovo.

Although forest conservation policymakers in the most post-socialist European countries still have limited financial, human, and political resources, over the past two decades, publicly available, remotely sensed satellite data on deforestation and degradation have dramatically reduced evaluation costs [16, 17]. Such an approach could be much cheaper and the series of data would be quite useful for monitoring the forest cover [12].

Remote sensing is the detection, recognition, or evaluation of objects by means of distant sensing or recording devices [18]. NDVI and CLC belong to these very promising remote sensing techniques that allow monitoring forest changes over time [19, 20]. One of the main differences between NDVI and CLC is that when NDVI focuses on the vegetation cover and its status, CLC has a much broader scope and distinguishes agricultural areas, forests and semi-natural areas, artificial surfaces, urban fabric, industrial, commercial, and transport units, bodies of water, wetlands, glaciers and perpetual snow, and other features [21]. Normalized difference vegetation index (NDVI) [22] is one of the most widely used vegetation indices (VIs) which focuses on the vegetation cover and its status [21, 22, 23, 24, 25, 26, 27, 28]. NDVI, like all VIs, relates the spectral absorption of chlorophyll in the red with a reflection phenomenon in the near-infrared, influenced by the leaf structure type [29]. NDVI also has the advantage of allowing comparisons between images acquired at different times [30]. On the other hand, CLC is a European program launched in 1985 by the European Commission, aimed at obtaining a comparable dataset of land cover for Europe. The aim of CLC is to gather information related to the environment on certain priority topics for the European Union: air, water, soil, land cover, coastal erosion, biotopes, etc. The main goal of the CLC program is to unify heterogeneous thematic cartographies of Europe at various levels (international, national, regional, local), and to update data at regular intervals, every 5–10 years [31, 32]. CLC is a map of the European environmental landscape based on interpretation of satellite images. The data have been validated using local cartography and ground surveys [33, 34, 35]. CLC also has an NDVI module for creating vegetation maps.

Advances in remotely sensed data availability have, obviously, created significant new opportunities to map changes in land cover and forest patterns and enhance understanding of the effectiveness and efficiency of forest conservation policy in Serbia [36, 37, 38].

Advertisement

2. Materials and methods

In order to evaluate the possible use of Corine land cover (CLC) and NDVI for preventing abrupt illegal logging in southern Serbian municipalities [39], our study was carried out for the municipality of Topola, located in central Serbia, and the Municipality of Kursumlija that lies in southern Serbia, bordering the territories of Kosovo (Figure 1).

Figure 1.

Locations of the municipalities of Kursumlija and Topola.

NDVI data for both municipalities are based on Landsat 5 Thematic Mapper (TM) satellite images for 2006, 2011, 2012 and 2014, which were created during spring/summer (August), with minimum clouds (10–20%; [40]).

NDVI was used and necessary corrections/transformations were applied for visible red in constellation with the infrared spectrum of satellite images using the following procedure: GIS Analysis/Mathematical Operation/Image Calculator, and then the equation NDVI = (NIR − RED)/(NIR + RED), in which NIR is the near-infrared channel and RED is the red channel from the visible part of the spectrum [41, 42]. Basic tasks included analysis and photo interpretation of elements, occurrences, and processes detected on images using specialized GIS software (Idrisi 15-Andes) for processing remotely sensed images through application of NDVI. Since shadow areas were less than 5% in the Municipality of Kursumlija and less than 3% in the Municipality of Topola, no topographic corrections were made (Table 1).

LANDSAT 5 (TM sensor)Wavelength (pm)Resolution (m)
Band 10.45–0.5230
Band 20.52–0.6030
Band 30.63–0.6930
Band 40.76–0.9030
Band 51.55–1.7530

Table 1.

Different spectral channels of images (produced by LANDSAT satellite) used in this paper.

Prior to change detection, a normalized difference vegetation index (NDVI) images were generated and a threshold classification technique was applied. Using the NDVI, threshold technique involves less time and data, but selecting an accurate NDVI threshold can be difficult and still remains a challenge. For example, some surveys show that soils have a highly variable NDVI and the mean value (0.20–0.21) that is much larger than the NDVI commonly used (< 0.05). This problem is most severe in areas with sparse vegetation cover (e.g. grassland and shrubland) where typical seasonal NDVI values are in the sensitive range (0.2 < NDVI pixel < 0.4) [26]. Also, although the most common minimum NDVI value for broad-leaved forests is 0.4, even the 0.25 NDVI threshold was appropriate for some deciduous forest types (for northeastern China) [43] and also for temperate forests [44]. Previously, AVHRR-derived NDVI data advanced very high resolution radiometers (AVHRR) have been used to define the length of temperate forest growing seasons, using NDVI threshold values ranging from 0.25 [44] to 0.45 [45]. One survey of Northern Ghana suggests that appropriate NDVI value for forests is 0.32–0.4 [46], while study of Bangladesh it is 0.25 [47]. Also, Xiao et al. suggest that the 0.25 NDVI threshold is appropriate for the onset of greenness development of deciduous forest types in northeastern China [43]. Any generalization, as well as creation of the class itself, is often an arbitrary process. Also, threshold parameters, produce arbitrary and artificial differences in values in the real world [48]. Hence, the underlying logic of the reliable use of the threshold technique should be derived experimentally by defining classes step-by-step to know if they are correct [49]. As Jansen points out, different perspectives, or so-called ‘scapes’, to categorization can be taken that are all equally valid and valuable [49]. The threshold used for classification here has been chosen based on using sites of known change and stability in order to define the most suitable and reliable threshold for Serbia. Extensive field surveys were then guided with a global positioning system (GPS) receiver. The data collected during these field surveys were used to determine the major types of land use in the study areas, which helped design a land cover classification and to associate the ground “truth” of a specific type of land cover with its imaging characteristics, and which helped in making a reliable threshold. In this phase, we had to use supervised classification to correct intensity, hue and saturation of pixels, and also to correct the obtained shades (that were showing vegetation), in order to precisely distinguish forests from vineyards and orchards and to produce precise distinction between shrub and broad-leaved forest for Serbia. For broad-leaved forest in Serbia, appropriate range was between 0.292 and 0.438. The NDVI images were classified into eight classes: (1) grassy areas, meadows, and pastures, (2) sparsely vegetated areas, (3) shrub vegetation, (4) vineyard, (5) orchards, (6) broad-leaved forest, (7) mixed forest, and (8) coniferous forest (Figures 2–4). Vegetation areas are presented with values between 0 and 1. Grassy areas, meadows, and pastures have values that range from zero up to 0.122. Sparsely vegetated areas has an NDVI value from 0.123 to 0.174, shrub vegetation between 0.175 and 0.230, vineyard 0.231 and 0.262, orchards 0.263 and 0.291, broad-leaved forest between 0.292 and 0.438, mixed forest between 0.439 and 0.525, and coniferous forest has an NDVI value above 0.526 [28, 50].

Figure 2.

Vegetation cover of municipality Kursumlija for 2011 obtained by NDVI.

Figure 3.

Vegetation cover of municipality Kursumlija for 2014 obtained by NDVI.

Figure 4.

Vegetation cover of municipality Topola for 2014 obtained by NDVI.

For creating CLC maps for the municipalities of Kursumlija and Topola, image processing was carried out and a digital elevation model was made based on the municipalities’ boundaries and Landsat satellite color composites, and a pseudo-color composite with bands 4, 5, 1 and adequate contrast was applied. Datasets and maps for Serbia, mainly CLC2006 and CLC2012, were extracted from the European Environmental Agency (EEA) website, with a transfer data scale of 1:100,000 [51]. Although in the original CLC project the smallest unit is 25 ha, a recent approach yields more precise results because changes <25 and >5 ha are mapped [52]. Nevertheless, even the smallest 5 ha areas, which are highly appropriate at the EU scale, do not properly reflect the land use situation at the local scale in a country where landscapes and land use change across very short distances [53, 54]. Obviously CLC is an example of a top-down process of European standardization that follows a common system of nomenclature and focuses on common definitions and methods, and does not properly take into account unique features of individual countries of Central and Eastern Europe, like Serbia.

Finally, we performed the map accuracy assessment in order to evaluate the quality of our maps by comparing the mapped or predicted value to the observed or true condition. The basis of any accuracy assessment is a location-specific comparison of a map prediction and a ground observation. Accuracy assessment requires three primary components [55]: (1) the sampling design—choosing which locations are visited; (2) the response design, used to decide whether the map predicted and observed condition match; and (3) estimating accuracy parameters and summarizing the results of the assessment [55, 56, 57, 58, 59, 60]. As Foody [56] points out, accuracy assessment has been a topic of considerable debate and research in remote sensing for many years. This is in part because the promoted standard methods such as the Kappa coefficient are not always appropriate [56]. Instead of Kappa, we have chosen different approach. The main problem with applying CLC for Serbia is that any (for CLC) unrecognizable spatial area (smaller than 4 ha) is automatically added to the closest recognizable larger area. Since CLC provided 135 of these small spatial areas (smaller than 4 ha) for Topola and 324 for Kursumlija, we concentrated exactly on them, and performed multiple GPS field surveys to compare our results with the “ground truth”.

Hence, for our sampling design (1) we have chosen 64 (out of 135) of these spatial areas smaller than 4 ha (their size varied—from 0.02 to 4 ha), then (2) we compared our NDVI results with GPS field survey results, to decide whether the map predicted and observed condition match; and finally (3) we estimated accuracy parameters and summarized the results of the assessment (Table 11). This way not only that we could: (a) show the level of CLCs imprecision for spatial areas smaller than 4 ha (compared to the NDVI results), but also (b) we could check out the level of NDVI accuracy.

Also, since municipalities Kursumlija and Topola include mountainous areas (up to 1400 and 1800 m, respectively), and elevation, aspect, and slope are the three main topographic factors that control the distribution and patterns of vegetation in mountain areas [61, 62, 63, 64], while among these three factors, elevation is most important [64, 65, 66, 67], we performed statistical analysis to see whether different altitude zones have any impact on our NDVI results.

Advertisement

3. Results

When the CLC results for 2006 and 2012 were compared with official forest area estimates for the same years (Tables 2–5), some inconsistencies became apparent:

Type of vegetationKursumlija (km2)Topola (km2)
Year2006201220062012
Settlements4.604.879.119.30
Green urban areas——0.820.78
Non-irrigated arable land0.420.3536.5136.49
Natural grasslands25.7425.4214.4414.27
Complex cultivation patterns78.7379.61154.94155.12
Land principally occupied by agriculture, with significant areas of natural vegetation102.25101.8173.1772.91
Broad-leaved forest620.68619.7855.6855.54
Coniferous forest3.633.440.190.15
Mixed forest6.146.012.122.00
Pastures24.1824.021.111.13
Transitional woodland-shrub75.7877.070.811.21
Sparsely vegetated areas0.780.55——
Total942.93942.93348.9348.9

Table 2.

Land cover in the municipalities of Kursumlija and Topola obtained from CLC for 2006 and 2012.

MunicipalityMunicipality total area (km2)Official statistics for 2006Official statistics for 2012Calculated on the basis of CLC for 2006Calculated on the basis of CLC for 2012
Topola35752.0052.049457.9957.69
Kursumlija952604.41544.2856630.45629.23

Table 3.

Forest areas according to official statistics and calculated on the basis of CLC for the years 2006 and 2012 (in km2).

CLC—Official statistics difference (km2)CLC—Official statistics difference (%)
Year2006201220062012
Topola5.995.6411.5210.84
Kursumlija26.0484.944.3115.61

Table 4.

Difference between forest areas calculated on the basis of CLC and according to official statistics for the years 2006 and 2012 (in km2 and in %).

MunicipalityMunicipality total area (km2)Official statistics for 2006Official statistics for 2011Official statistics for 2014Calculated on the basis of NDVI for 2006Calculated on the basis of NDVI for 2011Calculated on the basis of NDVI for 2014
Topola35752.0052.049447.5252.1451.2444.34
Kursumlija952604.41544.2856546.474600.97530.64507.85

Table 5.

Forest areas according to official statistics and calculated on the basis of NDVI for the years 2006, 2011, 2014 (in km2).

Forest areas obtained from CLC were up to 15.6% larger than the official forest area estimates. Kursumlija’s forest area obtained from CLC for 2006 is 4.3% larger than the official forest area estimates for this municipality and 15.6% larger than the official forest area estimates for 2012, while Topola’s forest area obtained from CLC for 2006 is 11.5% and for 2012 is 10.8% larger than the official forest area estimates.

On the other hand, when compared with official forest area estimates [68, 69, 70], the NDVI results for the Municipality of Topola’s forest area show a mere −0.27% difference for 2006, −1.55% for 2011, and −6.69% difference for 2014, and for the Municipality of Kursumlija − 0.57% difference for 2006, −2.51% for 2011, and −7.07% difference for 2014 (Table 6). For 2006 and 2011, NDVI results completely fit within the ±5% margin of error allowed for this method [71, 72], while for 2014 (−7% difference) our NDVI results exceed ±5% margin of error. Obviously, since NFI 2006, discrepancies between official forest area updates and our NDVI results are permanently growing, which opens possibilities for further analysis.

NDVI—Official statistics difference (km2)NDVI—Official statistics difference (%)
Year200620112014200620112014
Topola0.14−0.81−3.180.27%−1.55−6.69
Kursumlija−3.44−13.65−38.62−0.57%−2.51−7.07

Table 6.

Difference between forest areas calculated on the basis of NDVI and according to official statistics for the years 2006, 2011 and 2014 (in km2 and in %).

Vegetation cover in the municipalities of Kursumlija and Topola obtained from NDVI for 2006, 2011, and 2014 is presented in Tables 7. Since official updates for 2011 and 2014 do not contain different (coniferous, broad-leaved, and mixed) forest areas, comparison of NDVI values with official statistics was not possible.

Type of vegetationKursumlija (km2)Topola (km2)
Year200620112014200620112014
Coniferous forest9.303.653.610.610.170.13
Mixed forest12.424.954.722.082.021.95
Broad-leaved forest579.25522.04499.5249.4549.0542.26
Orchards25.6333.5435.5538.3343.2551.35
Vineyards7.406.415.1512.117.235.02
Shrub vegetation23.3315.4213.2412.224.203.86
Sparsely vegetated areas5.24.124.018.657.556.24
Pastures34.4537.9538.2213.432.893.04
Other255.02323.92347.98220.12240.64243.15
Total952952952357357357

Table 7.

Vegetation cover in the municipalities of Kursumlija and Topola obtained from NDVI for 2006, 2011 and 2014.

There are huge differences between our CLC and NDVI results for forest area for Kursumlija—even 98.59 km2 of a difference for 2012 (99.1%—or 97.74 km2—of these differences belong to the broad-leaved forest area category (Tables 8–10).

MunicipalityMunicipality total area (km2)Calculated on the basis of CLC for 2006Calculated on the basis of CLC for 2012Calculated on the basis of NDVI for 2006Calculated on the basis of NDVI for 2011
Topola35757.9957.6952.1451.24
Kursumlija952630.45629.23600.97530.64

Table 8.

Forest areas calculated on the basis of NDVI and CLC for the years 2006 and 2012 (in km2).

CLC-NDVI difference (km2)CLC-NDVI/NDVI difference (in %)
Year2006201220062012
Topola5.856.4511.2212.59
Kursumlija29.4898.594.9018.58

Table 9.

Difference between forest areas calculated on the basis of NDVI and CLC for the years 2006 and 2012 (in km2 and in %).

Calculated on the basis of CLCCalculated on the basis of NDVI
MunicipalityTotal forest area (km2)Coniferous forestBroad-leaved forestMixed forestTotal forest area (km2)Coniferous forestBroad-leaved forestMixed forest
Topola57.690.1555.542.0051.240.1749.052.02
Kursumlija629.23.44619.786.01530.643.65522.044.95

Table 10.

Forest areas calculated on the basis of NDVI and CLC for 2012(2011) (in km2).

When we finally performed the map accuracy assessment, for all of our NDVI forest spatial areas, the error did not exceed ±3%. Hence, there was an extremely high degree of conformity between NDVI results and the “ground truth” (Tables 11).

No.Forest area obtained by NDVI (ha)Forest area obtained by GPS (ha)Deviation NDVI/GPS (%)
10.4730.47499.79
20.3450.345100.00
30.7150.714100.14
40.7630.762100.13
50.6990.699100.00
60.5450.545100.00
70.1100.110100.00
80.1120.112100.00
90.1750.175100.00
100.0490.05098.00
110.0610.061100.00
120.0580.05998.30
130.0610.06298.39
140.0710.07298.61
150.0440.044100.00
160.0410.04297.62
170.0550.05698.21
180.0550.055100.00
190.0510.051100.00
200.0320.032100.00
210.0350.035100.00
220.0310.030103.33
230.0240.02596.00
240.0210.021100.00
250.0700.069101.45
260.4510.45299.78
270.0350.034102.94
280.0330.032103.125
290.0200.02195.24
300.0220.022100.00
313.9443.846102.55
323.6253.623100.05
333.2343.232100.06
343.9853.98799.95
354.0034.001100.05
363.2563.25899.94
373.6853.683100.05
383.9423.94499.95
392.5452.54799.92
403.0253.02699.97
411.5891.587100.13
422.4522.46099.67
432.5512.551100.00
442.7032.70499.96
452.5552.55699.96
460.6330.632100.16
470.6520.652100.00
480.5800.58199.83
490.5540.55599.82
500.2810.281100.00
510.3320.332100.00
520.3540.354100.00
530.4200.42299.95
540.4500.449100.22
550.3630.36599.45
560.1590.157101.27
570.4550.45699.78
580.0300.03196.77
590.0290.029100.00
600.0210.02295.45
610.0250.02696.15
620.0580.057101.75
630.0490.048102.08
640.0350.034102.94

Table 11.

Deviation between 64 forest areas/units obtained by NDVI and by GPS (for spatial units smaller than 4 ha) in municipality of Topola for 2014 (in %).

Finally, by using SPSS software we showed completely insignificant value of Pearson’s correlation coefficient (0.106) for different altitude zones (240–280, 330–450, 450–550, 750–850, and 900–1400 m) and deviation between forest areas obtained by NDVI and by GPS, clearly showing that topography does not have any relevant impact on our NDVI results.

Advertisement

4. Discussion

Since, according to the official statistics, Kursumlija experienced a 10% (60 km2) loss of the forest area in the 2006–2011 period, or, according to our NDVI results, 15.5% (93 km2) loss in the 2006–2014 period, this is very clear case of alarmingly rapid process of deforestation, in very sharp contrast with very modest rates of deforestation in Serbia and in municipality of Topola. Numerous studies convincingly showed that this extremely quick process of deforestation in Kursumlija is the obvious result of illegal logging [3, 13, 14, 15, 37].

Although CLC have recently been used in Serbia for spatial planning at the local level, the main problem with CLC data is that (a) although CLC data are produced at various levels (international, national, regional, and local; [31, 32]), CLC is actually a predominant regional database, updated rarely (every 5–10 years), whereas NDVI is available every year and (b) NDVI is much more precise than CLC.

When official statistics were compared with NDVI and CLC forest areas for the same year (2006, 2011, 2012, and 2014), NDVI was more precise than CLC. Actually, the main problem with applying CLC for Serbia is that any (for CLC) unrecognizable spatial area smaller than 4 ha is automatically added to the closest recognizable larger area. This proved to be decisive for Serbia, where privately owned forest parcels, which account for half of the total forest area of the country, usually cover much smaller areas (the average private holding is 0.3 ha; [3]) and it is the main cause for CLCs (up to 15.6% imprecise) larger than official forest area estimates. In short, commonly used EU CORINE land cover (CLC), which serves as a tool for fulfilling pan-European monitoring needs [49] proved not to be very suitable for local forest management in Serbia (questionable results regarding forests were also determined in Slovenia, for example, [54]). In addition, apart from the obvious CLC imprecision for studies at the local level, CLC data are not available for every year.

On the other hand, all our NDVI results for Topola and for Kursumlija for 2006 and 2011 completely fit within the ±5% margin of error, while for 2014 they are much closer (±3% difference) to the “ground truth” (checked by our multiple GPS field surveys) and better fit the forest area trajectory. These discrepancies between our NDVI results and official inventory updates are obviously permanently growing: for Kursumlija, 2006 official forest inventory shows only 3.44 km2 more forest areas; for 2011, 13.65 km2 more; and for 2014, 38.62 km2 more forest areas. Since our NDVI results for forest areas proved to be very close to the “ground truth” for 2014, it clearly indicates that in case of Serbia, official forest area updates are not precise. As Foody [56] points out, although remote sensing has been used successfully in mapping a range of land covers, land cover maps are often judged to be of insufficient quality for operational applications, especially when they are evaluated against some ground or other reference data set (like official forest area updates, in our case). Disagreements between the two data sets are typically interpreted as errors in the land cover map derived from the remotely sensed data [56].

Since our NDVI results for forest areas in 2014 proved to be very close to the “ground truth”, these growing discrepancies between official forest area updates and our NDVI results are obviously the result of insufficient quality of official forest area updates.

Actually, there are several reasons that official forest area updates are very imprecise:

First, private forests are inventoried via assessment methods (cadastral data are used for assessing forest area) and Serbia is well known for the lack of cadaster updating.

Second, while national forest inventory in forests includes only inventory unit areas larger than 0.5 ha, private forests (that constitute 52% of the nation’s forest resources) are characterized by very small plots (average size: 0.3 ha).

Third, official inventory of illegal loggings in private forests in Serbia is notoriously imprecise. The real amount of logged wood in private forests is six time as high as the registered amount [73].

Fourth, Serbian forest authorities have limited access to this extremely sensitive Serbia-Kosovo border area (formally under Serbian forest estates Vranje, Kursumlija, Leskovac, Raska, and Leposavic).

Also, it is important to underline here that none of the number of shortcomings that are usually addressed to the NDVI use [74, 75], proved to be relevant in the case of Kursumlija. Firstly, it is very easy to locate and quantify overall amounts of timber harvested in the case of Kursumlija municipality, because illegal logging produce large canopy gaps that go/extend from the border of Kosovo to approximately 3–4 km into the Kursumlija’s territory. Secondly, from the field survey (hammer marks and size/height of stumps in the field), it is obvious that it is the clear case of illegal cutting, and thirdly, since illegal logging in Kursumlija is organized by groups of individuals, with market-oriented behavior [76] which are part of organized crime and closely tied to other criminal activities such as corruption, violence, and money laundering [77], it is, of course, not any sort of concession allocation issue.

Hence, it is exactly the objectivity of remote sensing that can be of the greatest help in resolving extremely quick and forceful process of illegal logging in this very sensitive southern Serbian area [12, 13, 76, 77], by providing an reliable, up-to-date alternative data source to quantify forest cover and change (independent of often very imprecise official governmental data sources) [78].

Advertisement

5. Conclusions

Through this analysis of NDVI and CLC results, CLC proved not to be a very suitable tool for local forest management in Serbia: (a) CLC is an example of a top-down process of European standardization that does not properly take into account unique individual features of post-socialist countries like Serbia, which resulted in extremely high (up to 15.6%) level of results imprecision and (b) CLC data are not available for every year. On the other hand, it is evident that NDVI, especially in southern Serbian municipalities with prevalent illegal logging can provide local forest managers with much precise annual information about forest area change.

Despite certain shortcomings [18, 24, 74, 75], classification and area estimation of various land cover types based on NDVI, has obviously advanced to a point where it surpasses old wood inventory techniques, especially in the case of post-socialist countries like Serbia. Specifically,

  • It is relatively cheap [17] and quick, and it can provide forest managers with precise, up-to-date, annual information.

  • It is easy to implement, which is of crucial importance for Serbia, where national forest inventories have been carried out very rarely. The last three national forest inventories were carried out at roughly 20-year intervals (in 1961, 1979, and 2007).

The objectivity of the method can significantly help in avoiding abrupt, unrecorded illegal logging that in post-socialist countries can exceed the legal harvest by a factor of 10.

Advertisement

Acknowledgments

This work was supported by the Ministry of Science and Technological Development of the Republic of Serbia under grant no. 37010.

References

  1. 1. World Bank. Forests Sourcebook: Practical guidance for sustaining forests in development cooperation. Washington: World Bank; 2008
  2. 2. Ministry of agriculture forestry and water management of the Republic of Serbia. The National Forest Inventory of Serbia. Belgrade: Planeta Print; 2009
  3. 3. Glavonjic B, Jovic D, Vasiljevic A, Kankaras R. Forest and forest products country profile: Serbia and Montenegro. Geneva: United Nations; 2005
  4. 4. FAO. Global forest resources assessment update 2005. Terms and definitions (final version) forest resources assessment programme working paper 83/E. Rome, 2004
  5. 5. Rominjin E, Ainembabazi JH, Wijaya A, Herold M, Angelsen A, Verchot L, et al. Exploring different forest definitions and their impact on developing REDD+ reference emission levels: A case study for Indonesia. Environmental Science & Policy. 2013;33:246-259
  6. 6. GFOI. Integrating remote-sensing and ground-based observations for estimation of emissions and removals of greenhouse gases in forests. Geneva: Group on Earth Observations (GEO), 2013
  7. 7. Kohl M, Traub B, Paivinen R. Harmonization and standardization in multi-national environmental statistics—mission impossible? Environmental Monitoring and Assessment. 2000;63(2):361-380
  8. 8. Lawrence M, McRoberts RE, Tomppo E, Gschwantner T, Gabler K. Comparisons of national forest inventories. In: Tomppo E, Gschwantner T, REM ML, editors. National Forest Inventories. New York: Springer; 2010
  9. 9. Nonić D, Petrović N, Medarević M, Glavonjić P, Nedeljković J, Stevanov M, Orlović S, Lj R, Djordjević I, NR PZ. Forest Land Ownership Change in Serbia. COST Action FP1201 FACESMAP Country Report. Vienna: European Forest Institute Central-East and South-East European Regional Office; 2015
  10. 10. Taff GN, Müller D, Kuemmerle T, Ozdeneral E, Walsh SJ. Reforestation in Central and Eastern Europe after the breakdown of socialism. In: Nagendra H, Southworth J, editors. Reforesting Landscapes. New York: Springer; 2010. p. 121-147
  11. 11. World Bank. Ensuring Sustainability of forests and livelihoods through improved governance and control of illegal logging for economies in transition. Washington D. C.: World Bank. 2005
  12. 12. Harou P, Hajredini E. Illegal logging in Kosovo. Washington D. C.: USAID; 2009
  13. 13. Jovanovic M, Milanovic M. Remote sensing and Forest conservation: Challenges of illegal logging in Kursumlija municipality (Serbia). In: Shukla G, Chakravarty S, editors. Forest Ecology and Conservation. Rijeka: InTech; 2017
  14. 14. The Regional Environmental Center for Central and Eastern Europe. Illegal logging activities in the Republic of Serbia—B. Diagnostic Audit. Belgrade; 2009
  15. 15. Regional Environmental Center. Illegal logging activities in the Republic of Serbia—A. Fact-finding study. Belgrade; 2009
  16. 16. Blackman A. Evaluating forest conservation policies in developing countries using remote sensing data: An introduction and practical guide. Forest Policy and Economics. 2013;34:1-16
  17. 17. Woodcock CE, Allen R, Anderson M, Belward A, Bindschadler R, Cohen WB, et al. Free access to Landsat data (Letter). Science. 2008;80:320
  18. 18. Franklin SE. Remote sensing for sustainable forest management [Internet]. New York. 2001. 407 p. Available from: http://www.amazon.ca/exec/obidos/redirect?tag=citeulike09-20&path=ASIN/1566703948
  19. 19. Romijn E, Lantican CB, Herold M, Lindquist E, Ochieng R, Wijaya A, et al. Assessing change in national forest monitoring capacities of 99 tropical countries. Forest Ecology and Management. 2015;352:109-123
  20. 20. MacDicken K. Global Forest resources assessment 2015: What, why and how? Forest Ecology and Management. 2015;352:3-8
  21. 21. Jensen JR. Remote Sensing of the Environment: An Earth Resource Perspective. Upper Saddle River, NJ: Prentice Hall; 2007
  22. 22. Weier J, Herring D. Measuring Vegetation (NDVI & EVI) [Internet]. Feature Articles. 2000. pp. 1-4. Available from: http://earthobservatory.nasa.gov/Features/MeasuringVegetation/measuring_vegetation_1.php
  23. 23. Chen CH. In: Image Processing for Remote Sensing. New York: CRC Press; 2008. 400 p
  24. 24. Campbell J, Wynne R. Introduction to Remote Sensing. New York: Guilford Press; 2011
  25. 25. Win RN, Suzuki R, Takeda S. Remote sensing analysis of Forest damage by selection logging in the Kabaung reserved forest, Bago Mountains Myanmar. Journal of Forest Research. 2012;17(2):121-128
  26. 26. Montandon LM, Small EE. The impact of soil reflectance on the quantification of the green vegetation fraction from NDVI. Remote Sensing of Environment. 2008;112(4):1835-1845
  27. 27. Vohland M, Stoffels J, Hau C, Schüler G. Remote sensing techniques for forest parameter assessment: Multispectral classification and linear spectral mixture analysis. Silva Fennica. 2007;41(3):441-456
  28. 28. De Jong, S. M., Van der Meer FD. Remote sensing image analysis: Including the spatial domain. New York: Springer; 2005
  29. 29. Wang Q, Tenhunen JD. Vegetation mapping with multitemporal NDVI in north eastern China transect (NECT). International Journal of Applied Earth Observation and Geoinformation. 2004;6(1):17-31
  30. 30. Lillesand TM, Kiefer RW, Chipman JW. Remote sensing and image interpretation [Internet]. Vol. 3, New York, Chichester, Brisbane, Toronto: 6IS s, 2004, 756 p. Available from: http://www.osti.gov/energycitations/product.biblio.jsp?osti_id=6028047
  31. 31. Bossard M, Feranec J, Otahel J. CORINE Land Cover technical guide—addendum 2000. Technical report 40 [Internet]. 2000 [cited 2016 May 17]. Available from: http://www.eea.europa.eu/publications/tech40add
  32. 32. Neumann K, Herold M, Hartley A, Schmullius C. Comparative assessment of C0RINE2000 and GLC2000: Spatial analysis of land cover data for Europe. International Journal of Applied Earth Observation and Geoinformation. 2007;9(4):425-437
  33. 33. Heymann Y, Steenmans CH, Croissille G, Bossard M. CORINE land cover, Technical guide. Luxembourg: office for official Publications of the European Communities; 1994
  34. 34. Perdigäo V, Annoni A. Technical and methodological guide for updating CORINE land cover data base [Internet]. 1997 [cited 2014 Jan 1]. Available from: http://www.ec-gis.org/docs/F27057/CORINE.PDF
  35. 35. Genovese G, Vignolles C, Negre T, Passera GA. Methodology for a combined use of normalised difference vegetation index and CORINE land cover data for crop yield monitoring and forecasting: A case study on Spain. Agronomie. 2001;21(1)
  36. 36. Potapov PV, Turubanova SA, Tyukavina A, Krylov AM, McCarty JL, Radeloff VC, et al. Eastern Europe’s forest cover dynamics from 1985 to 2012 quantified from the full Landsat archive. Remote Sensing of Environment. 2015;159:28-43
  37. 37. Jovanovic M, Milanovic M, Zorn M. The use of NDVI and CORINE land cover databases for forest management in Serbia. Acta Geographica Slovenica. 2018;58(1):109-123
  38. 38. Jovanović M, Milanović M. Normalized difference vegetation index (NDVI) as the basis for local Forest management. Example of the municipality of Topola. Polish Journal of Environmental Studies. 2015;24(2):529-535
  39. 39. Verstraete MM, Pinty B. Designing optimal spectral indexes for remote sensing applications. IEEE Transactions on Geoscience and Remote Sensing. 1996;34(5):1254-1265
  40. 40. Chávez PSJ. Image-based atmospheric corrections—revisited and improved. Photogrammetric Engineering & Remote Sensing. 1996;62(9):1025-1036
  41. 41. Hájek F. Process-based approach to automated classification of forest structures using medium format digital aerial photos and ancillary GIS information. European Journal of Forest Research. 2008;127(2):115-124
  42. 42. Johnson LF, Trout TJ. Satellite NDVI assisted monitoring of vegetable crop evapotranspiration in california’s san Joaquin Valley. Remote Sensing. 2012;4(2):439-455
  43. 43. Xiao X, Boles S, Liu J, Zhuang D, Liu M. Characterization of forest types in northeastern China, using multi-temporal SPOT-4 VEGETATION sensor data. Remote Sensing of Environment. 2002;82:335-348
  44. 44. Myneni RB, Tucker CJ, Asrar G, Keeling CD. Interannual variations in satellite-sensed vegetation index data from 1981 to 1991. Journal of Geophysical Research. 1998;103:6145-6160
  45. 45. Jenkins JP, Braswell BH, Frolking SE, Aber JA. Predicting spatial and interannual patterns of temperate forest springtime phenology in the eastern U.S. Geophysical Research Letters. 2002;29(24)
  46. 46. Ofosu Anim D, Tiereyangn Kabo-bah A, Nti Nkrumah P, Tackmore Murava R. Evaluation of NDVI using SPOT-5 satellite data for Northern Ghana. Environmental Management and Sustainable Development. 2013;2(1)
  47. 47. Roni R. Surface temperature and NDVI generation and relation between them: Application of remote sensing. Asian Journal of Engineering and Technology Innovation. 2013;1(1):8-13
  48. 48. Di Gregorio A, O’Brien D. Overview of land-cover classifications and their interoperability. In: Giri C, editor. Remote Sensing of Land Use and Land Cover. London, NewYork: CRC Press; 2012
  49. 49. Jansen LJM, Badea A, Milenov P, Moise C, Vassilev V, Milenova L, et al. The use of the land-cover classification system in eastern European countries: Experiences, lessons learnt and the way forward. In: Manakos I, Braun M, editors. Land Use and Land Cover Mapping in Europe. London, NewYork: Springer; 2014
  50. 50. Hord RM. Digital image processing of remotely sensed data [Internet]. Enschede; 1982. Available from: https://books.google.co.in/books?id=e-tRAAAAMAAJ
  51. 51. Büttner G, Kleeschulte S. Implementation of CLC2000 in Serbia-Montenegro—technical annex. 2005
  52. 52. EEA. CLC2006 Technical Guidelines [Internet]. 2007 [cited 2014 Jan 1]. Available from: http://www.eea.europa.eu/publications/technical_report_2007_17
  53. 53. Hočevar M, Kobler A. Uporaba podatkov CORINE land cover za izgradnjo informacijskega sistema o gozdovih na državni ravni. Gozdarski Vestnik. 2001;46(3):353-369
  54. 54. Gabrovec M, Petek F. V Sloveniji vedno manj gozda? (Ne)uporabnost metodologije CORINE Land Cover. Delo. 2004;46(303)
  55. 55. Stehman SV, Czaplewski RL. Design and analysis for thematic map accuracy assessment. Fundamental principles. Remote Sensing of Environment. 1998;64:331-344
  56. 56. Foody GM. Status of land cover classification accuracy assessment. Remote Sensing of Environment. 2002;80:185-201
  57. 57. Czaplewski RL. Accuracy assessment of maps of forest condition: Statistical design and methodological considerations. In: Wulder MA, Franklin SE, editors. Remote Sensing of Forest Environments: Concepts and Case Studies. Boston: Kluwer; 2003. p. 115-140
  58. 58. Strahler AH, Boschetti L, Foody GM, Friedl MA, Hansen MC, Herold M, et al. Global Land Cover Validation: Recommendations for Evaluation and Accuracy Assessment of Global Land Cover Maps. Ispra; 2006
  59. 59. Stehman SV, Foody GM. Accuracy assessment. In: Warner TA, Nellis MD, Foody GM, editors. The SAGE Handbook of Remote Sensing. London: Sage; 2009. p. 297-309
  60. 60. McRoberts R, Cohen WB, Næsset E, Stehman SV, Tomppo EO. Using remotely sensed data to construct and assess forest attribute maps and related spatial products. Scandinavian Journal of Forest Research. 2010;25:340-367
  61. 61. Titshall LW, O’Connor TG, Morris CD. Effect of long-term exclusion of fire and herbivory on the soils and vegetation of sour grassland. African Journal of Range & Forage Science. 2000;17:70-80
  62. 62. Wood SW, Murphy BP, Bowman DM. Firescape ecology: How topography determines the contrasting distribution of fire and rain forest in the south-west of the tasmanian wilderness world heritage area. Journal of Biogeography. 2011;38:1807-1820
  63. 63. Hwang T, Song C, Vose JM, Band LE. Topography-mediated controls on local vegetation phenology estimated from modis vegetation index. Landscape Ecology. 2011:541-556
  64. 64. Kayiranga A, Ndayisaba F, Nahayo L, Karamage F, Nsengiyumva JB, Mupenzi C, et al. Analysis of climate and topography impacts on the spatial distribution of vegetation in the Virunga volcanoes massif of east-Central Africa. Geosciences. 2017;7:17
  65. 65. Day FP, Monk CD. Vegetation patterns on a southern Appalachian watershed. Ecology. 1974;55:1064-1074
  66. 66. Busing RT, White PS, MacKende MD. Gradient analysis of old spruce-fir forest of the great Smokey Mountains circa 1935. Canadian Journal of Botany. 1992;71:951-958
  67. 67. Jin XM, Zhang Y-K, Schaepman ME, Clevers JGPW, Impact SZ. Of elevation and aspect on the spatial distribution of vegetation in the qilian mountain area with remote sensing data. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. 2008;37:1385-1390
  68. 68. Statistical Office of the Republic of Serbia. Municipalities and Regions of the Republic of Serbia 2012. Belgrade: Statistical Office of the Republic of Serbia; 2013
  69. 69. Statistical Office of the Republic of Serbia. Municipalities and Regions of the Republic of Serbia 2007. Belgrade: Statistical Office of the Republic of Serbia; 2008
  70. 70. Statistical Office of the Republic of Serbia. Municipalities and regions of the republic of Serbia 2015. Belgrade: Statistical Office of the Republic of Serbia; 2016
  71. 71. Lunetta RS, Knight JF, Ediriwickrema J, Lyon JG, Worthy LD. Land-cover change detection using multi-temporal MODIS NDVI data. Remote Sensing of Environment. 2006;105(2):142-154
  72. 72. Eastman R. IDRISI32 Release 2—Tutorial. Worcester: Clark Labs; 2001
  73. 73. Markus-Johansson M, Mesquita B, Nemeth A, Dimovski M, Monnier C, Kiss-Parciu P. Illegal Logging in South Eastern Europe. Szentendre: The Regional Environmental Center for Central and Eastern Europe. 2010
  74. 74. Khai TC, Mizoue N, Kajisa T, Ota T, Yoshida S. Stand structure, composition and illegal logging in selectively logged production forests of Myanmar: Comparison of two compartments subject to different cutting frequency. Global Ecology and Conservation. 2016;7:132-140. Available from: http://dx.doi.org/10.1016/j.gecco.2016.06.001
  75. 75. Lawson S, MacFaul L. Illegal logging and related trade [Internet]. Chatham House, 2010, pp. 1-154. Available from: papers2://publication/uuid/C7E061BB-61C3-4A58-A30D-DF4FFDA3CBD2
  76. 76. Bouriaud L, Nichiforel L, Nunes L, Pereira H, Bajraktari A. A property rights-based analysis of the illegal logging for fuelwood in Kosovo. Biomass and Bioenergy. 2014;67:425-434
  77. 77. WWF. Illegal logging & The EU. An analysis of the EU export & import market of illegal wood and related products. Brussel: WWF European Policy Office; 2008
  78. 78. Sundstrom A. Understanding illegality and corruption in forest governance. Journal of Environmental Management. 2016

Written By

Jovanović M. Miomir, Miško M. Milanović and Bojan R. Vračarević

Submitted: 11 May 2017 Reviewed: 23 October 2017 Published: 20 December 2017

© 2017 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution 3.0 License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Continue reading from the same book

View All
Chapter 2 Changes in Land Management of Iberian Rangelands a...

By Manuel Pulido Fernández, Joaquín Francisco Lavado ...

1311 downloads

Chapter 3 Vegetation Changes in the Miombo Woodlands in Nort...

By Albert Chirima, Peter Mundy, Nicholas Ncube and An...

1715 downloads

Chapter 4 Neotropical Forests from their Emergence to the Fu...

By Cristiane Dias e Sarmento and Marcel Giovanni Cost...

1533 downloads