3.6 Methodology
3.6.5 Change detection techniques
3.6.5.1 Post-classification comparison
The post-classification comparison (PCC) technique is the most straightforward method of change detection. It depends on the comparison of independently produced classified images by properly coding the classification results for time 1 and time 2 (Singh, 1989) and the output produces a change map that indicates a complete matrix of change (Figure 3.4).
Figure 3.4 A flowchart of post-classification change detection technique
Date 1 Date 2 Image classification Classification map of date 1 Classification map of date 2 Change map Output Input
Asubonteng (2007) points out that PCC is a commonly-used quantitative technique for change detection, although challenges can arise when classifying historical image data.
Alphan et al. (2008) apply Post-classification comparison on supervised maximum likelihood decision rule for Landsat imagery from 1989, 2000, and Aster imagery from 2004, in an attempt to assess land cover changes in Kahamanmaras, Turkey. The study found that dramatic land cover conversions took place as a result of urbanization and agricultural expansion, resulting in vegetation degradation in the study area. Yang and Wen (2011) use post-classification comparison on a supervised maximum likelihood classification of Landsat ETM+ imagery from 2000 and 2002 to detect changes in Guangzhou in southern China. This study reveals that the largest change areas are exchanges of building and unused land with grassland. The advantages of post-classification comparisons include: (1) the technique avoids the need for strict radiometric calibration and minimizes the impact of atmospheric, sensor and environmental differences between multi-temporal images; and (2) the technique provides a complete matrix of change directions unlike image differencing. Macleod and Congalton (1998) highlight that post-classification comparison has significant limitations because it combines the errors from both classifications. Additionally, enough ancillary data must be presented to classify both data. Moreover, the change-map output of two classifications often display accuracies similar to the product of multiplying the accuracies of each individual classification (Stow et al., 1980 and Mas, 1999). Further limitations include the need for knowledge, expertise, and time to create classification products (Lu et al., 2004).
3.6.5.1.1 Land-Use/Land-Cover Change Rates
Post-classification comparison provides "from-to" information. Actual change can be obtained by a direct comparison between classified images from one date with that obtained on another date. Temporal changes that have occurred between the two dates can be measured by performing a change matrix (Appendix 1).
Between 1972 and 1984, agricultural land represented the highest percentage of coverage in Edd Al Fursan locality. This indicates that agricultural activities were the main source of
income. The dynamic change of natural vegetation cover shows that forest land significantly decreased from 26 % in 1972 to 18 % in 1984, while grassland comprised the same percentage in the two respective years. This degradation in natural forest is a result of the drought periods in the late seventies and the beginning of the eighties that led people to exploit it as the main source of income. Moreover, the fallow and bare lands increased in 1984. The increase in fallow land resulted from people abandoning their farmlands while the increase in bare land can be attributed to the conversion of forest to bare land as a result of drought conditions as mentioned above.
In the period between 1984 and 1989, there was an extension period of drought in this area. There was no significant change in vegetation cover, but there was a high increase in fallow land as drought conditions push farmers to abandon their agricultural land and looking for other livelihood options. However, the increase in fallow land has led to a decrease in bare land area from 13.5 % in 1984 to 11.5 % in 1989.
The LULC changes for the period between 1989 and 1999, as it indicates an increase in the natural vegetation cover. Grasslands as well as natural forests increased from 16 %, 17 % to 21 % and 19 %, respectively. This is due to a good rainy season during this period as shown by the Ministry of Agriculture’s annual rainfall report and expressed by farmers during interviews. This indicates that the environment had begun to recover from the drought conditions experienced previous to this period.
There was no obvious increase in agricultural land in this period, while there was a decrease in fallow from 30 % in 1989 to 24. % in 1999. This can be interpreted that some of the fallow land were converted to the natural vegetation. However, the bare land shows less decrease in 1999.
Between 1999 and 2008 there was a dramatic change in vegetation cover, as human and animal populations have increased in recent years due to conflict and environmental degradation resulting in high demand on food and fodder, which led to over-cultivation and overgrazing. However, grass and forest lands decreased from 21 %, 19 % in 1999 to 19 %,
and 17 % in 2008, respectively. Moreover, agricultural land has increased from 25 % in 1999 to 30 % in 2008 resulting in a decrease in fallow from 24 % in 1999 to 20 % in 2008. On the other hand, bare land increased in 2008 in comparison to 1999 where constituted 14 % and 11%, respectively. This was the result of intensive agriculture and decrease in natural vegetation cover.
The LULC change rate for the 1972-2008 period indicates a high rate of deforestation, or 9.2 %. Therefore, cultivated land has increased from 26 % in 1972 to 30 % in 2008. This period shows an intensive pressure for land occupation due to land degradation as well as conflict that pushed a high number of IDPs to settle in this area, whereas most of them depend on forest products for their livelihood. Bare land levels have increased from 9 % in 1972 to 14 % in 2008, largely as the result of deforestation and intensive agriculture.