**4. Results**

#### *4.1. Accuracy Assessment of Land Cover Products*

According to the confusion matrix analysis, 90% overall accuracy and kappa coefficient value of 0.85 was attained for the 2016 classified map. Similarly, overall classification accuracy levels of 88% with kappa coefficient of 0.84 for 2010, 86% with kappa coefficient of 0.82 for 2000, 85% with kappa coefficient of 0.81 for 1990, and 83% with kappa coefficient of 0.79 for 1976 image classifications were achieved (Table A1 in Appendix A).

#### *4.2. Temporal Land Cover Assessment*

The overall settlement in the ICT increased while tree cover classes (greater and less than 40% tree canopies) decreased in forty years (1976 to 2016). The estimated land cover area for each image is summarized in Table 4 and spatial distributions are presented in Figure 3.


**Table 4.** Land cover 1976, 1990, 2000, 2010, and 2016 assessment based on satellite images.

**Figure 3.** Land cover maps for 1976, 1990, 2000, 2010, and 2016.

The settlement class in ICT increased from 29.99 km<sup>2</sup> (3.31%) in 1976 to 170.40 km<sup>2</sup> (18.80%) in 2016. The tree cover >40% canopy class showed a slight increase, 182.87 km<sup>2</sup> (20.18%) in 1976 to 192.19 km<sup>2</sup> (21.21%) in 1990 and then declined to 132.27 km<sup>2</sup> (14.59%) in 2016. Tree cover <40% canopy land cover class faced overall decline from 1976 to 2016: 417.03 km<sup>2</sup> (46.02%) in 1976 to 342.94 km<sup>2</sup> (37.85%) in 1990, increase from 1990 to 2000 to 399.31 km<sup>2</sup> (44.07%), while again decreased from 2000 to 2016 to 306.53 km<sup>2</sup> (35.72%). The other two land cover classes, soil and water, observed fluctuating trends in the forty year period. The soil area of 270.27 km<sup>2</sup> (29.83%) in 1976 increased to 317.91 km<sup>2</sup> (35.08%) in 1990 and then reduced to 257 km<sup>2</sup> (28.37%) in 2000, then further decreased to 254.34 (28.02%) in 2010, and again increased to 289.45 km<sup>2</sup> (33.82%) in 2016. The change in water area was not significant though it has shown fluctuations during the study period; it increased from 5.86 km<sup>2</sup> (0.64%) in 1976 to 8.44 km<sup>2</sup> (0.94%) in 1990, and to 10.51km<sup>2</sup> (1.16%) in 2000, then decreased to 6.29 km<sup>2</sup> (0.69%) in 2010, and then increased slightly to 6.51 km<sup>2</sup> (0.76%) in 2016 (Table 4).

Based on the temporal land cover data in the ICT, urban landscape settlement increased at an annual rate of 4.34% since 1976, with the highest annual rate of 8.79% during 2000–2010. The tree cover >40% canopy decreased at an annual rate of 0.81% between 1976 to 2016 and tree cover >40% canopy declined at an annual rate of 0.77% in forty years (1976–2016). The tree cover >40% canopy witnessed the highest loss rate at 2.38% per annum between 2000–2010, and tree cover <40% canopy experienced tree cover loss at 2.72% during 2010–2016 (Table 5).


**Table 5.** Land cover annual rate of change (% change per year) from 1976–1990, 1990–2000, 2000–2010, 2010–2016, and 1976–2016.

#### *4.3. Urban Landscape Matrix Analysis*

Landscape matrix analysis was performed at class level, and the results for each class are reported in Figure 4 and Table A2 in Appendix A. The parameters Number of patches (NP), Largest Patch Index (LPI), Edge Density (ED), and Mean Proximity Index (MPI) for the settlement class overall increased from 1976 to 2016. Largest Patch Index (LPI) increased to 16% in 1990 from 13% in 1976, and then reduced to less than 10% in 2000, 2010, and 2016. For tree cover greater than and less than 40% canopy, Number of patches (NP) and Edge Density (ED) initially increased from 1976 to 2000, and then declined from 2000–2016. For tree cover greater than and less than 40% canopy, soil, and settlement classes, the Mean Patch Size (MPS) and Mean Euclidean Nearest Neighbor Distance (MED) sharply declined from 1976 to 1990 and remained lower. In the Mean Euclidean Nearest Neighbor Distance (MED), tree cover greater than and less than 40% canopy, soil and settlement classes remained less than 250 m throughout from 1976 to 2016. From 1976 to 2010, the Mean Radius of Gyration (MRG) decreased for tree cover greater than and less than 40% canopy, and then slightly increased during the years 2010 to 2016. For tree cover >40%, the canopy Mean Proximity Index (MPI) slightly increased between 1976 to 2016 while for tree cover >40%, the canopy MPI decreased from 1976 to 1990, grew between 1990 to 2010, and then again sharply reduced from 2010 to 2016. In all the land cover classes, the Mean Shape Index (MSI) value remained greater than 1 with slight fluctuations. In the Mean Perimeter to Area Ratio (MPAR), the settlement class increased from 640 in 1976 to 1000 in 1990 with slight increase till 2016.

**Figure 4.** Land cover class level landscape matrix parameters graphs for 1976, 1990, 2000, 2010, and 2016.

#### *4.4. Urban Forest Fragmentation Analysis*

The forest fragmentation analysis results are presented in Figures 5 and 6, and Table A3 in Appendix A. The analysis shows that the overall core forests of >500 acres decreased from 391.98 km<sup>2</sup> (65.41%) in 1976 to 241.44 km<sup>2</sup> (40.29%) in 2016. In the forty years' time span, the patch class increased to 20 km<sup>2</sup> (4.54%) in 2016 from 14.74 km<sup>2</sup> (2.46%) in 1976. The perforated forest fragmentation class was 35.21 km<sup>2</sup> (5.88%) in 1976, which increased 41.79 km<sup>2</sup> (7.23%) in 2000, and then again declined to 22.74 km<sup>2</sup> (5.18%) in 2016. For the edge class, a minor increase was observed from 1976 to 2000 and then a drop in 2016.

**Figure 5.** Forest fragmentation maps for 1976, 1990, 2000, 2010, and 2016.

**Figure 6.** Forest fragmentation graphs for 1976, 1990, 2000, 2010, and 2016.

#### *4.5. Ratio of Land Consumption Rate to the Population Growth Rate (LCRPGR)*

For the years 1976–2000, the LCRPGR ratio (SDG indicator 11.3.1) was obtained to be 0.62, which highlights the simultaneous increase of Population Growth Rate (PGR) and Land Consumption Rate (LCR), but the LCR is much slower than the PGR. For the years 2000–2016, the LCRPGR ratio was obtained to be 1.36, which reflects the simultaneous increase of PGR and LCR, with a faster LCR than PGR.
