*2.5. Generalized Additive Model (GAM)*

Health outcomes can exhibit linear or nonlinear variation, as can urbanization, landuse expansion, and industrial development. The GAM model provides primary functions (linear, polynomial, or spline) to fit the variation of dependent variables [40], such as health outcomes, and independent variables. Given the complexity of the relationship between urbanization and environmental health, we constructed the following GAM model:

$$\mathcal{Y} = \beta\_0 + \sum\_{j=1} f\_j(\mathbf{X}\_j) + \epsilon\_j$$

where *Y* is a dependent variable (such as total mortality), *fj*(·) is a random univariate function for independent variable *Xj* (such as the urbanization rate), and *<sup>j</sup>* is a normal random error term.

The model determines the relationships between the dependent variable and independent variable according to the degree of freedom (DF): if DF = 1, the relationship is linear; if DF > 1, the relationship is nonlinear; and a higher DF indicates a more significant nonlinear relationship.

We included the urbanization rate (UR, the percentage of the urban population in the total population), the construction land area (CLA, km2), and the proportion of heavy industrial output to total industrial output (PHI, %) as independent variables. We included total mortality (2001–2015), the number of cancer cases (2002–2011), and the mortality from cancer from 2009–2015 in Changzhou as dependent variables into the GAM, to explore the relationship between urbanization and health risks. Mortality from cancer in 2009 and 2010 was estimated according to its linear relationship with that of Wujin district, with a R<sup>2</sup> of 0.98.

#### **3. Results**
