Machine learning and spatial regression approaches to estimating willingness to pay for ecosystem services in Jakarta and Taoyuan

Ecosystem services (ES) are essential to environmental sustainability and human well-being. Among them, hydrological ecosystem services (HES) play a critical role in flood mitigation, climate regulation, and water security. This study examines the socioeconomic drivers influencing individuals' willingness to pay (WTP) for HES in two contrasting urban settings: Jakarta in Indonesia and Taoyuan in Taiwan. We applied and compared three data-driven models (i.e., Logistic Regression, Geographically Weighted Logistic Regression, and Extreme Gradient Boosting) to assess both spatial and non-linear determinants of WTP. A total of 1006 respondents were surveyed using structured face-to-face interviews. Respondents were selected via purposive sampling to ensure representation of relevant sociodemographic and regional characteristics. Results show that WTP, expressed in 2024 USD per person per year, is lower in Jakarta (5.52 USD) compared to in Taoyuan (9.99 USD). Demographic and socio-economic variables, particularly gender and education, are key predictors of WTP, followed by support for environmental initiatives. The influence of these factors varies spatially, suggesting that effective ES protection policies should be tailored to local population characteristics. By integrating spatially explicit and data-driven approaches, Payment for Ecosystem Services (PES) policies can more effectively promote community participation and support sustainable ecosystem preservation, particularly for urban ES valuation across East and Southeast Asia.