Results and discussion

5.1. Land cover analysis:

This analysis helps in delineating regions under vegetation and nonvegetation. Vegetation cover analysis was done through NDVI. Figure 3, 4 and table 6 indicates the land cover changes of different time periods for Chennai and Hyderabad regions respectively. In Chennai, vegetation cover has dramatically decreased from 70.47% (1991), to 35.53% in 2013, whereas the non- vegetation i.e. built up, paved areas, bare soil etc. have increased 29.53% in 1991 to 64.47% in 2013. Hyderabad also shows similar trend with decrease in vegetation from 95.64% (1989), 93.28% (1999), 82.67% (2009) and 61.15% (2014). Land use analyses was performed to understand the transitions across land use categories like built up, forests, water bodies, etc., Vegetation cover and water bodies aids in moderating local climate and also help in mitigating floods, etc.

Figure 3: Land cover changes 1991-2013, Chennai region

Figure 4: Land cover changes during 1989-2014 in Hyderabad region

Table 6: Temporal land cover details for Chennai and Hyderabad

5.2. Land use analysis and Accuracy assessment

Figure 5 and 6 represents land use dynamics for the study regions during the past four decades. Results revealed the steep increase of 72% in built up areas in Chennai at Ponneri, Pattabiram, Sriperumbudur, Tambaram, etc. during 1991-2000 and 646% during 2000-2013. Areas such as Malakpet, Madapur, Bollaram, Kukkatpally, Cherlapally, etc. Hyderabad showed an increase in built-up area by 93% during1989-1999, 319% (during 1999-2009) and 56% (2009-2014). It is important to notice that both the study regions show significant increase during the years 2000-2010 with emergence of various industrial sectors such as automobile, hardware manufacturing as well as information technology parks. Others category (mainly open spaces) has consistently reduced from 69.5% to 50.5% (1991-2013) in Chennai and 90.5% to 72.6% (1989-2014) in Hyderabad region indicating a large scale conversion to urban land use type. Water bodies of Hyderabad shows a very critical decrease which indicates either these land uses are converted or they have been dried up. Decline from 3.75% to 1.84% during 1989-2014, highlight the grave situation in the region and the need to restore and rejuvenate water bodies which aid as a lifeline of the society. Table 7 summarizes the land use details for Chennai and Hyderabad respectively.

Figure 5: Land use dynamics during 1991 to 2013 in Chennai metropolitan area

Figure 6: Land use dynamics in Hyderabad

Table 7: Land use dynamics in Chennai and Hyderabad

Table 8 lists overall accuracy and Kappa statistic for land use classified information for Chennai and Hyderabad. Overall accuracy for Chennai varied from 86% to 97% and for Hyderabad 87% to 94% highlights the agreement of classified information with the field data.

Table 8: Accuracy assessment of Chennai and Hyderabad regions

5.3. Urban sprawl analysis:

Shannon’s entropy is calculated direction wise considering the proportion of built-up/paved urban area in the gradient and results are listed in table 9. Shannon’s entropy values ranges from 0 (concentrated growth) and log n (dispersed growth or sprawl). ‘n’ indicates the number of circles /gradients in the respective direction. Analysis highlights the tendency of urban sprawl during 2000 and 2012 for Chennai and 2009 and 2014 for Hyderabad. Higher entropy values of 0.444 (NE), 0.396 (NW), 0.409 (SE) for Chennai and 0.442 (SW); 0.352(NE), 0.422 (NW), 0.444 (SE) and 0.355 (SW) in for Hyderabad shows of dispersed growth. The sprawl phenomenon is evident in figures 5 and 7.

Figure 7: Number of patch metrics for Chennai

Table 9: Year wise Shannon’s entropy values for the two cities

5.4. Quantifying spatial patterns of urbanization through metrics:

Six metrics were computed usingFRAGSTATS for each gradient, zone wise tounderstand the spatial patterns of urban growth.

• Number of patch (NP): Figures 7 and 8 gives NP for Chennai and Hyderabad. Patches shows rapid growth in all the directions pointing out fragmentation during 2009, 2012, 2013 and 2014. During 2013 and 2014, core city area (circles 1-9 in Chennai and circles 1-11 in Hyderabad), each patch has agglomerated into a single large urban patch i.e. there is a saturated urban landscape with no other land uses (Egmore, Nugambakkam in Chennai and Abids, Secunderabad, Narayanaguda, Somajiguda, etc. in Hyderabad). Sprawl is evident with higher number of patches in NW, SW directions (Chennai) and NE, SE, SW directions (Hyderabad).
  Figure 8: Number of patch metrics for Hyderabad
• Normalized landscape shape index (NLSI): This metrics provides measure of class aggregation. All four zones show lesser value of NLSI in 2013 and 2014 compared to 1991 and 1989 (reflected in figures 9 and 10 for Chennai and Hyderabad). The minimum values (NLSI <0.5) especially in CBD areas (such as Ambattur, Nungambakkam, Sowcarpet, Egmore, in Chennai region and also Nampally, Secunderabad, Medhipatnam etc. in Hyderabad region) indicates that the landscape consists of a single square urban patch or it is maximally compact (i.e., almost square) in contrast with the higher values in 1991 and 1989 (NLSI ≈ 1, specifying maximally disaggregated urban patches with complex shapes.
  Figure 9: NLSI metrics for Chennai Figure 10: NLSI metrics for Hyderabad
• Clumpiness: This metric indicates the aggregation and disaggregation for adjacent urban patches. Figures 11 and 12 show the values closer to 0 for the regions corresponding to circles 25-35, in NE direction (Manali new town, Ennore and SW direction includes Irungattukottai, Kondavakkam) of Chennai and the regions corresponding to circles 23-31, in NE, SE and SW directions (Keesara Mandal, Rampally, Manneguda, Shankarpally) in Hyderabad, highlighting less compact growth or maximal disaggregation. In 2012 and 2014 values, approaching +1 in core city areas (circles 1-15) of all directions indicate of very complex growth with all maximally aggregated patches forming large urban monotype patch.
  Figure 11: Clumpiness metrics for Chennai Figure 12: Clumpiness metrics for Hyderabad
• Interspersion and juxtaposition index (IJI): This metric show how well an urban patch is associated or interspersed with other adjacent patch types. Lower values as observed (figures 13 and 14) in 1990’s indicates an urban patch is associated only with one other adjacent patch type. This phenomenon does not hold well at the outskirts since urban patch is equally adjacent to all other patch types (i.e., maximally interspersed and juxtaposed to other patch types) showing sprawl in these areas.
  Figure 13: Interspersion and juxtaposition metrics for Chennai Figure 14: Interspersion and juxtaposition metrics for Hyderabad

5.5.Urban growth modelling:

Land uses in 2025 were predicted considering various agents (amenities, road and railway network) and constraints (protected areas, drainage lines and slope) for Chennai and Hyderabad regions to visualize and understand the likely urban growth i. Pair wise comparison between two factors were done to obtain weights for these factors using AHP. Consistency ratio of 0.05 and 0.07 were achieved for Chennai and Hyderabad respectively, which is considered satisfactory to continue with further analysis. Land use changes for the year 2013 for Chennai region and year 2014 for Hyderabad region were simulated. This helped in validation for comparing simulated land use with the actual land use based on the classification of respective remote sensing data. Satisfactory kappa values with greater accuracy achieved, indicate of higher agreement between the actual and predicted land uses (table 10).

Table 10: Validation statistics (Simulated and classified image)

Prediction for the year 2025 was performed using Markovian transition estimator tool considering
(i) constraint of CDP wherein water bodies, forest areas, catchment areas and coastal regulation areas (only in Chennai region) as no development areas. and
(ii) without considering constraint of CDP.
Table 11 lists percentage changes in land use categories, especially two fold increase in built-up areas with the decrease in vegetation and other categories. Two scenarios i.e. with CDP and without CDP showed similar statistics, but it is very essential to note that with the constraint of implementation of CDP, urban growth would be at the outskirts or at the periphery of the city boundary. However, in absence of CDP, distressing trend of large scale land use changes in areas within the CMDA boundary such as Korathur and Cholavaram lake bed, Redhills catchment area, Perungalathur forest area, Sholinganallur wetland area etc. which will either be encroached or completely occupied by built-up category (figure 15). A similar trend is observed in Hyderabad with violations in CDP, vulnerable ecologically sensitive areas such as Musi river bed (Malakpet), Mir Alam and Madeena guda lake bed, Kanchan bagh, Alwal wetland area and Janakinagar wetland gets changed into built-up categories (figure 16).

Zone and gradient wise spatial metrics were computed with 2025 predicted images to understand the spatial patterns of urban growth and sprawl. Figures 9 - 16 depict the metric wise spatial patterns of urbanization. Number of patches and patch density in the core city area (circles 1-9, Chennai and circles 1-12, Hyderabad) in all directions showed almost zero values implies that the entire landscape is completely dominated and saturated by only one single urban patch. For both the regions, in all directions (except Hyderabad, SE, circles 23-35) NLSI values were observed to be lesser than 0.2 indicating the urban patches are more compact, dense and has attained a standard shape. Clumpiness values almost reaching +1 as well as aggregation index values to 100, showed urban landscape maximally aggregated in both regions. Urban shape index values for 2025 are less compared to 2012/2013 for Chennai region and 2014 for Hyderabad region. This decreasing trend in urban landscape shape index further confirm of landscape attaining a standard or regular shape with the decrease in length of the edges. Largest urban patches were observed in circles 7-15, NE, SE (Kolathur and Vadapalani) and 7-11 SW (Poonamalle) of Chennai and circles 9-13 NW (Kukkatpally and Jeedimetla), 11-15, 23-29 SE (Secunderabad, Ghatkesar and Cherlapally IDA) and 9-11 SW (Manikonda and Hitech city) of Hyderabad. These metrics clearly indicate of intensified and concentrated urban growth in the core city and fragmented or dispersed growth in peri-urban regions.

Figure 15: Predicted land use categories for the year 2025 – Chennai region

Figure 16: Predicted land use categories for the year 2025 – Hyderabad region

Table 11: Predicted land use statistics for the year 2025 for Chennai and Hyderabad regions