Land surface temperature estimating in urbanized landscapes using artificial neural networks

Area of Science:

• Environmental Science
• Urban Planning
• Remote Sensing

Background:

• Urban land surface temperature (LST) is a critical factor influenced by land use and spatial patterns.
• Scenario-based modeling aids in developing effective urban land use planning policies.
• Artificial neural networks (ANN) can model complex, non-linear relationships in environmental data.

Purpose of the Study:

• To model and compare the impact of different urban development scenarios on LST in Greater Isfahan.
• To explore the relationship between urban LST and green cover using remote sensing data.
• To evaluate the effectiveness of the Green Development Scenario (GDS) versus the Compact Development Scenario (CDS) in mitigating urban heat.

Main Methods:

• Utilized Landsat 8 OLI imagery to derive vegetation and built-up indices.
• Employed an artificial neural network (ANN) model calibrated with Normalized Difference Vegetation Index (NDVI) and Normalized Difference Built Index (NDBI).
• Defined and simulated two distinct urban development scenarios: Compact Development Scenario (CDS) and Green Development Scenario (GDS).

Main Results:

• The Green Development Scenario (GDS) demonstrated a lower mean LST (40.93°C) compared to the Compact Development Scenario (CDS) (44.88°C).
• Urban LST retrieved from the CDS showed higher statistical significance (p=0.043) than that from the GDS (p=0.010) in the ANOVA analysis.
• ANN model successfully captured the non-linear relationships between LST and spatial patterns under different development scenarios.

Conclusions:

• Developing urban green spaces is a crucial strategy for mitigating risks associated with elevated land surface temperatures.
• The GDS is more effective in reducing urban LST, indicating the significant role of vegetation in urban heat island mitigation.
• Scenario-based LST modeling provides valuable insights for sustainable urban planning and policy-making.