Climate impacts and vulnerability assessment

In order to assess the spatial correlation between greenspace changes and the magnitude of climate impacts/vulnerability, we establish climate impact and vulnerability index maps. Using the defined indicators of climate change, we perform the following three steps to calculate the comprehensive weighted value of each pixel.

First, each climate vulnerability factor has different dimensions and orders of magnitude. In order to compare all values uniformly, it is necessary to normalize them so that each value is converted to a range between 0 and 1 by applying the linear Max–Min method (Eq. ²)^12,30.

where, δij is a normalized value for the factor i in pixel j; χij is an original value of the factor, minij and maxij are the minimum and maximum values, respectively.

Second, we use the AHP method, a multiple-criteria decision-making approach introduced by Saaty (1977)⁶⁷, to weigh the normalized factors. This method, which is structured using sets of pair-wise comparisons, is widely used in natural disaster risk studies^30,68. We used Saaty’s original comparative scale between 1 and 9, in which “1” suggests that two factors hold equal importance, while “9” is assigned when one factor is significantly more important than the other. In order to obtain the AHP weight values, we conducted a survey distributed from February 4–5, 2021 to experts through email/telephone interviews with 14 experts from four entities spanning the public sector, academia, and non-governmental research institutes: the Guangzhou Urban Planning Institute, Korean Ministry of Environment, Massachusetts Institute of Technology China City Lab, and Lincoln Institute of Land Policy for China Program. As the first author is familiar with the aforementioned institutes and organizations from previous collaborations and projects, we invited them to select experts with sufficient knowledge and experience in the fields of flood control, climate disaster mitigation, and/or climate change adaptation to participate in the our survey. While the respondents’ work experience ranged from six to 23 years, we chose not to weigh the initial judgements by this. The first-round expert judgements are based on separate individual hazards (typhoon, floods, and high temperatures) (Supplementary Tables ^S5–^S7). The second judgement process took place from February 6–8, 2021 and involved nine experts from the same institutes as in the first round, as five of the original participants were unresponsive within the allotted timeframe. The second round was based on one generic set of indicators chosen from the results of the first round. The weights of indicators for all hazard types together are shown in Supplementary Table ^S8, while full survey results are shown in Supplementary Tables ^S13–^S16. The consistency ratios in each AHP matrix are all less than 0.1, meaning that the criterion matrices are satisfactorily consistent.

Finally, we calculate a comprehensive weighted value (Supplementary Table ^S10) using the following equation:

where ζj is a comprehensive weighted value in pixel j; WE, WS, and WC are the weights of exposure, sensitivity, and adaptive capacity, respectively; ωf, ωg, and ωh are the weights of different indicators in the three vulnerability criteria; Ef, Sg, and Ch is the f^th, g^th, and h^th indicator within the exposure, sensitivity, and adaptive capacity criteria in pixel j, respectively; and δfj, δgj, and δhj are the normalized values for different indicators in pixel j.