Results and Discussion

Results and discussion

Assessment of historical LU transitions and simulation

Temporal LU analyses were carried out using RS data of 1985, 2005, and 2019 through a supervised classifier based on the Gaussian maximum likelihood algorithm. Fig. 6 depicts LU in Karnataka, indicating that the Western Ghats districts have a higher forest cover. The state has witnessed large-scale transitions with the decline of forests from 21% (1985) to 15% (2019). LU transitions gained impetus with globalization and consequent industrialization, urbanization, and infrastructure development during the post-1990s. Unplanned developmental activities have resulted in the loss of forest cover in the Western Ghats districts of Belgaum, Uttara Kannada, Dakshina Kannada, Shimoga, Kodagu, and Chikmagalur. Large tracts of forests were lost due to developmental activities such as constructing dams and reservoirs, land conversion for built-up areas, creating SEZs and townships. The abrupt LU changes are evident from the decline of productive agricultural lands near cities such as Bengaluru, Mysore, Hubli-Dharwad, and Shimoga. The forest cover is restricted to major conservation reserves such as protected areas, national parks, and wildlife sanctuaries. The category-wise LU dynamics are presented in Table 4, highlighting an increase in the built-up cover from 0.47% to 3% and horticulture area from 8.8% (1985) to 11.1% (2019). The decline in the spatial extent of agricultural lands, forests, and lakes highlights the need for sustainable LU policies to arrest deforestation and abrupt land conversions.

Table 4. Ecosystem Extent – Karnataka state (extent in km² and percentages) – based on temporal RS data analyses

The natural forests show a decline, evident from the decrease in evergreen forests from 7.5% (1985) to 5.7% (2019), moist deciduous forests from 5.7% (1985) to 4.1% (2019), and dry deciduous forests from 4.0% (1985) to 2.2 % (2019). An accuracy assessment of the classified RS data was done through the computation of category-wise accuracies and Kappa statistics. Table 5 lists the category-wise accuracies, which indicate that the overall accuracy ranges from 86.84% (1985), 90.08% (2005) to 89.39% (2019), and Kappa ranges from 0.85 (1985) to 0.88 (2019). The LU transitions across ecosystems for Karnataka are provided in Table 6. New urban agglomerations were noticed across cities and major towns such as Bengaluru, Mangalore, Hubli, Hassan, and Mysuru. Large-scale monoculture plantations of eucalyptus, rubber, acacia, teak, and areca nut have increased and now cover 12% of the state. These abrupt changes result in an imbalance of ecosystem services, affecting the hydrologic regime and availability of natural resources. Tier-1 cities such as Bengaluru and Mangalore, and tier-2 cities such as Mysuru, Hubli-Dharwad, and Belgaum are experiencing loss of agricultural areas in the sub-urban regions, with new layouts and satellite towns.

Table 5. Accuracy assessment

Table 6. Transitions across LU categories during 1985 to 2019 – Karnataka state (extent in km² and percentages)

Table 7 summarises the ecosystem extent account for LULC in Karnataka state. The LU transition from 1985 to 2009 was accounted to understand the area of change and probability of change in each LU type from time t₁ to time t₂ using the Markovian process. The accuracy for the simulated 2019 LU map was assessed by accounting for the consistency of land quantity and spatial position similarity at the pixel level (Table 8). Kappa statistics were computed to evaluate the consistency of land quantity simulated in comparison with the actual LU. The similarity of the spatial position of simulated LU with actual with respect to a certain pixel was also assessed. Fig. 7 illustrates the simulated LUs of 2019 with area details. The simulated LU depicts the expansion of existing built-up cover in peri-urban regions, through LU conversion from agriculture, based on the transition accounted for from 2005 (as per CA neighbourhood influence). The accuracy of the model was found to be 95%; the simulated LU and actual LU appear visually consistent, but spatial differences exist.

Table 7. Ecosystem extent account for LULC in Karnataka state, India

Table 8. Accuracy of the simulation

Fig. 6. LU changes from 1985 to 2019 in Karnataka

Fig. 7. Simulated land uses in 2019

Scenario-based prediction and validation

The distance maps for various urban growth factors (agents) were prepared (Fig. 8), including slope maps, distance to road, proposed railway projects, and industries. Five different constraints maps were prepared for the chosen scenarios. The site suitability maps were generated to define the suitability of each pixel for transition to any LU type. Each pixel in the suitability maps has a value ranging from 0 to 255, with 0 representing unsuitable and 255 representing the highly suitable area for a particular LU type (Pontius and Malanson 2005).

The distance influence of each agent on specific LUs was evaluated; this helped in effectively deriving the weights, unlike traditional weighing. Fig. 4 shows the influence of factors such as roads and industries on the built-up area. Industries have influence up to 2000 m from urban areas and moderate influence up to 10,000 m; the influence gradually decreases after that. Roads have shown significant influence up to 500 m, with a reduction in influence after 1000 m. Fig. 9 shows the factors that influence forest cover LU, which helped in understanding the distance influence of each factor. Industries show high conversion of forests to other LU up to 7000 m, and after that, the influence reduced; roads showed stronger influence up to 2000 m.

The AHP was used to assign weights and evaluated its consistency by computing the consistency ratio. Scenario-wise weights were assigned for each factor and are given in Annexure 1 (Tables a–e). A consistency ratio of <0.1 was achieved for each scenario, and the eigenvector weights for individual factors considered for the scenario are listed in Tables a-e (Annexure). The site suitability of LU transition was accounted for using MCE and was provided as an input for CA-based prediction. The CA has been used for computing projected LUs under five different scenarios based on varied inputs (Fig. 10).

BAU can be considered as a base case that helps in comparing or differentiating with alternative growth strategies. It depicts the likely increase in built-up area to 11.5% (2033) from 3% in 2019 (Table 9, Fig. 11). The forest cover and the agricultural regions would lose significant tract in case of likely expansion of built-up cover. The dry and moist deciduous forest cover has witnessed a higher change with a loss of 6.5% forest cover. The existing infrastructure and built-up cover will transform deciduous forests and agriculture areas along highways across the State. The cities such as Bengaluru, Mangalore, and Dharwad (tier-1) depicted compact growth, whereas Mysuru, Belgaum, Hassan, and Tumkur (tier-2) indicated peri-urban development.

Table 9. Likely LU in 2033 for various scenarios

Upgrading urban infrastructure (roads, commercial establishments, information networks, industrial units, and sports/recreation centers) will result in a built-up cover of 15% compared to BAU. The ALT scenario will increase paved surfaces by 2033 due to the development of towns and compact urban centers (consisting of high-density residential and industrial layouts). The BAU and ALT scenario predictions indicate a significant reduction in forest area and agricultural LU. The ALT scenario shows that forest cover would likely be only 9% compared to 15% in 2019 (Table 9, Fig. 11). The agricultural area showed a loss of 2%, which might affect food availability in the region. The regions where dry crops are grown will witness higher transformation, which signifies the external driver’s role in LU alterations. The ALT scenario indicates higher LU conversion due to transportation and other projects with compact, peri-urban, and mixed LU developments; this necessitates effective LU policies.

The RFP scenario visualizes protecting reserve forests without any alterations for unplanned developmental activities and expansion of earlier LUs. The regions covered under reserve forest protection do not favour development and other LU conversions. The districts that are part of the Western Ghats region might not affect the BAU and ALT scenarios. The regions with dry deciduous forest cover and not falling under reserve forest area protection will experience abrupt growth due to transportation corridors and urban clusters. Unlike other scenarios, urban areas of 2019 are considered as a factor influencing LU changes. The RFP scenario shows a likely built-up cover of 11%, retaining a good amount of evergreen forest cover of 5% (Table 9). As depicted in Fig. 11, the AF scenario shows likely improvement in forest cover due to plantation activities and protection of degraded forests from uncontrolled LU conversion. Here, the forest cover is 9.3%, a slight increase as compared to BAU and ALT. The built-up cover will limit existing cities and towns by over 11%. The new plantation activities will compensate for the abrupt LU conversion, especially in moist, dry deciduous cover regions. Districts such as Bidar, Kolar, and Koppala did not show any improvement due to the availability of the least forest land area.

The SDP scenario signifies stringent policy implementation initiatives with minor disturbances, and the simulated LU using this scenario is depicted in Fig. 11. The forest cover will remain 11% compared to 17% in 2019 (Table 9), the least probable loss among all the scenarios. The existing built-up cover will remain in the tier-1 and tier-2 cities; the least increase was shown in peri-urban areas. Districts such as Bidar, Gulbarga, Bagalkot, and Kolar might experience higher LU conversion rates due to existing infrastructure and be devoid of any forest policy. Overall, the RFP, AF, and SDP scenarios favour the protection of evergreen and moist deciduous forest categories. The BAU and ALT scenarios show a merging of urban clusters, indicating that smaller clusters of paved areas are no longer available and tend to agglomerate with bigger ones to form urban centers, satellite towns, and dispersed growth at the periphery of tier-2 city limits. There might be new constructions in the rural hinterland, resulting in loss of evergreen, moist deciduous cover and agricultural areas. Categories such as plantation, water, and open field had similar values across all scenarios. The scenarios will directly help decision-makers and planners to mitigate environmental impact with strict growth rules; otherwise, the conversion of primary, secondary forest/ moist/dry deciduous forest into urban areas will be favoured (BAU and ALT) can induce ecological imbalances. Another important observation in terms of statistics was that slope steepness restriction played an important role in restricting urban settlements from further infiltrating into forested areas. Collectively, sustainable development seems feasible with scenarios 3–5, without altering the existing edges of agriculture, forest, water bodies, and other protected areas or conservation zones.

The limitations of the current study lie in the factors considered, which vary based on developmental issues and regional dynamics. The model results may vary depending on the assumptions, modeling strategies, computational approaches, and data. The driving factors are generally region-specific, and the likely LU changes are expressed based on local variations in geographic and socio-economic conditions. The modeled scenarios applied at the micro (e.g., cities) level might show variations in results than the entire study region (state). The challenges to be addressed in LU transition models are: (a) lack of framework for systematic assessment of LULC changes and hence, its validation considering the decision-making process and authenticity of interactions between various agents in a system; (b) resolution of data, availability, and trade-offs between different resolutions. LU change is dependent on the land management policy framed by local, regional, national, and global agencies. Therefore, there is a need to integrate these factors along with human-environment interactions to achieve desirable results.

Fig. 8. Distance influence on LU change-built-up

Fig. 9. Distance influence on LU change-forest

Fig. 10. Projected LU under BAU, ALT, RFP, AF, and SDP scenarios

Fig. 11. Projected LU (2033) of policy scenarios

The modeled scenarios aid as a base for the decision-making process, to balance LU changes associated with developmental activities and conservation. The abrupt changes induced by developmental activities such as creating industries, upgradation of infrastructure, and establishing new linear corridors in regions dominated by forests will enhance deforestation and degradation. The unscientific LU changes in forest landscapes result in ecosystem imbalance and increased carbon emissions in the atmosphere (Ramachandra and Bharath 2019a). Landscapes with increased paved surfaces tend to have considerably greater air temperature than their peri-urban/rural surroundings, altered microclimatic conditions, increased anthropogenic heat release, and increased land surface temperature, which further contributes to global warming (Yao et al., 2017). The buildings and road surfaces act as high heat storages, exposing humans to health-threatening heat. The increase in paved surfaces due to uncontrolled urbanization in Karnataka state can result in an urban heat island effect due to the conversion of latent heat flux into sensible heat flux, thereby threatening human well-being (Ramachandra and Uttam, 2009). Other consequences of unregulated paved surfaces in a landscape are: (a) increase in the energy consumption for cooling with enhanced land surface temperature (Ramachandra et al., 2017b), (ii) escalation in the carbon footprint (Ramachandra and Shwetmala, 2009), (iii) reduction in water availability (Ramachandra et al., 2020), (iv) alteration in the seasonal rainfall pattern (Buyantuyev and Wu, 2012), (v) increase in flood instances (Kumar et al., 2021), (vi) effect on air quality (Feizizadeh and Blaschke, 2013; Fuladlu and Altan, 2021), (vii) phenological changes (Allen et al., 2015), (ix) impact on biodiversity (Ramachandra et al., 2018), (x) reduction in the net primary productivity of vegetation (Jackson and Baker 2010; Bharath et al. 2013; Alavipanah et al. 2015), and finally (xi) vegetation die-off (Breshears et al. 2005; Zhou et al. 2016). In this regard, the Karnataka state authorities should focus on sustainable developmental planning with judicious resource usage and existing vegetation cover improvement. Banning highly-polluting industries, illegal mining, and unscientific land conversion, and regulating urbanization should be prioritized to mitigate impending climate change.