Modeling diurnal Temperature-Rainfall relationships under multicollinearity using PLS-SEM: A case study of Ghana.

Area of Science:

• Climatology
• Tropical Meteorology
• Statistical Modeling

Background:

• Rainfall variability in tropical climates is complex, influenced by thermal processes.
• High collinearity between maximum (TMAX) and minimum (TMIN) temperatures complicates understanding their individual impacts on precipitation (RAIN).

Purpose of the Study:

• To investigate the distinct structural associations between TMAX, TMIN, and RAIN in Ghana.
• To explore how these relationships differ between coastal and inland climatic regimes.
• To assess the utility of Partial Least Squares Structural Equation Modeling (PLS-SEM) for analyzing interdependent climate data.

Main Methods:

• Employed PLS-SEM on long-term station-based climate data from Ghana (1981-2020).
• Analyzed 252 station × calendar month observations to model relationships among TMAX, TMIN, and RAIN.
• Utilized multi-group analysis to compare coastal and inland regions.

Main Results:

• A significant negative association was found between TMAX and RAIN (β = -0.454, p < 0.001).
• A significant positive association was observed between TMIN and RAIN (β = 0.166, p < 0.001).
• The model explained 21.1% of rainfall variance, with stronger TMAX-RAIN links inland (β = -0.721) than coastal (β = -0.427).

Conclusions:

• TMAX and TMIN exhibit distinct, statistically significant relationships with rainfall in Ghana.
• These thermal-rainfall associations demonstrate spatial heterogeneity, varying between coastal and inland areas.
• PLS-SEM is effective for analyzing multicollinearity in climate data, though further variables are needed for deeper mechanistic understanding.