Measuring Recovery to Build up Metrics of Flood Resilience Based on Pollutant Discharge Data: A Case Study in East China

Jinglu Song, Bo Huang, Rongrong LiWater 2017 · DOI

TL;DR

This study introduces a novel method to measure urban flood resilience by analyzing pollutant discharge data from industrial and municipal facilities. By tracking the recovery of these essential services after a flood, researchers can create a robust, data-driven validation for resilience metrics, offering valuable insights for urban planners in China.

What Problem Does This Paper Address?

The increasing frequency and intensity of flash floods, exacerbated by climate change, pose a significant threat to urban areas, particularly in China. While traditional disaster resistance measures focus on preventing damage, the concept of "disaster resilience" emphasizes a community's ability to recover and return to normal functioning after an event. However, measuring and validating resilience indicators has been a persistent challenge due to a lack of standardized definitions and reliable data. Existing methods often rely on subjective variable selection or external measurements that may not capture the full picture of post-disaster recovery. This paper aims to bridge this gap by developing a quantitative method to assess recovery capability, using it as an objective external validation for resilience metrics.

Methods

The core of this research lies in a novel approach to measuring recovery capability using time-series analysis of waste-water and waste-gas discharge/emission data. The authors collected hourly data from 42 monitoring points of key enterprises (power plants, sewage treatment plants, and factories) in Changzhou, China.

  1. Data Preprocessing: Time-series data was preprocessed to remove "tendency" and "periodicity" influences, ensuring stationarity for accurate change detection.
  2. Change Detection: Three nonparametric change detection methods (changepoint, cpm, and ecp packages in R) were employed to identify the start and end times of disturbances in the pollutant discharge data, indicating when infrastructure was affected by the flood and when it began to recover.
  3. Optimal Model Selection: The detection results from the three methods were compared against real flood inundation data to determine the most accurate approach. The ECP method was found to be optimal.
  4. Validation and Refinement: Detection results were further refined by comparing them with other methods to minimize misdetection.
  5. Recovery Duration Calculation: The time taken for each monitoring point to return to a stable status was calculated. This "return time" was then aggregated to the sub-district (or town) level to represent the recovery capability of that area.
  6. Resilience Metric Validation: Ordinal logistic regression was used to link the measured recovery capability (dependent variable) with various potential resilience indicators (independent variables) across social, economic, infrastructural, and environmental domains. This allowed the researchers to identify which indicators were statistically significant predictors of recovery capability.

Datasets

  • Pollutant Discharge Data: Hourly waste-water and waste-gas discharge/emission data from 42 monitoring points of local power plants, sewage treatment plants, and main factories in Changzhou, China. This data was sourced from the Jiangsu Province's self-monitored pollutant discharge information platform.
  • Precipitation Data: Records from rain gauges and stream measurement stations in the study area.
  • Flood Inundation Map: Generated using the ArcHydro toolbox in ArcGIS software, based on water level data from gauge stations.
  • Demographic Data: 2010 Population Census of China (PCC) data at the sub-district level.
  • Economic Data: GDP data from the Global Change Research Data Publishing and Repository (GCRD).
  • Geospatial Data:
    • Open Street Map (OSM) data for road density, access to open space, and commercial/manufacturing establishments.
  • GLOBELAND30 dataset for build-up area.
  • ASTER GDEM dataset for elevation and slope.

Key Insights and Ideas

  • Pollutant Discharge as a Proxy for Recovery: The study demonstrates that monitoring waste-water and waste-gas discharge from industrial and municipal facilities can serve as a reliable, real-time indicator of urban infrastructure recovery after a flood. This is a significant advancement, especially in regions where other high-resolution data sources are scarce.
  • Data-Driven Resilience Validation: By using measured recovery capability as an external validation metric, the study provides an objective way to assess the effectiveness of different resilience indicators. This moves beyond subjective selection and strengthens the scientific basis for resilience assessment.
  • Interconnectedness of Resilience Factors: The findings suggest that resilience is a multi-dimensional concept, and improvements in one area can have cascading positive effects. For instance, better health services and higher urbanization levels are linked to faster recovery.
  • Synergistic Benefits of Resilience Investments: The study implies that investments in resilience for one type of natural hazard can have positive spillover effects for others, suggesting a more holistic approach to disaster preparedness.

Key Results

  • The ECP change detection method was identified as the most accurate for analyzing the pollutant discharge data, achieving a correct detection rate of 0.72.
  • The ordinal logistic regression model, which explained 61.4% of the variance in measured recovery capability, identified six significant variables as proxies for flood resilience:
  • Social: Sex Ratio, Health Services
    • Economic: Ratio of Urban to Rural
    • Infrastructural: Road Density, Access to Open Space
    • Environmental: Slope
  • Higher levels of health services and a greater ratio of urban to rural populations were found to be strong predictors of faster recovery.
  • Road density and access to open space were also identified as crucial infrastructural factors for resilience.
  • The control variable, "Return Speed of Water Level," was also significant, highlighting the direct impact of flood recession on overall recovery.

Limitations and Future Work

  • Data Availability: While pollutant discharge data is valuable, its availability and quality can vary. The study acknowledges the constraint of data availability for other empirical measurements like vulnerability and adaptive capacity.
  • Model Explanatory Power: The authors suggest that the explanatory power of the final model could be improved by incorporating additional variables, such as those related to government policy, community capital, and emergency relief services.
  • Change Detection Accuracy: While efforts were made to improve accuracy, misdetection in change detection remains a challenge. Further research is needed to enhance these methods and validation techniques.
  • Generalizability: The study was conducted in a specific case study area in East China. Future work should involve applying this method to different regions and disaster types to evaluate its broader applicability.
  • Other Resilience Dimensions: The study focused on recovery capability. Future research could explore other dimensions of resilience, such as adaptive capacity and vulnerability, to provide a more comprehensive assessment.

Summary updated

Abstract

Building “disaster-resilient” rather than “disaster-resistant” cities/communities requires the development of response capabilities to natural disasters and subsequent recovery. This study devises a new method to measure resilience via recovery capability to validate indicators from social, economic, infrastructural, and environmental domains. The pollutant discharge data (waste-water and waste-gas discharge/emission data) of local power plants, sewage treatment plants and main factories were used to monitor recovery process of both people’s living and local industrial production as the waste water/gas is released irregularly during the short disaster-hit period. A time series analysis of such data was employed to detect the disturbance on these infrastructures from disasters and to assess community recovery capability. A recent record-breaking flash flood in Changzhou, a city in eastern-central China, was selected as a case study. We used ordinal logistic regression to identify leading proxies of flood resilience. A combination of six variables related to socioeconomic factors, infrastructure development and the environment, stood out and explained 61.4% of the variance in measured recovery capability. These findings substantiate the possibility of using recovery measurement based on pollutant discharge to validate resilience metrics, and contribute more solid evidences for policy-makers and urban planners to make corresponding measures for resilience enhancement.
BibTeX
@inproceedings{song2017measuring,
  author = {Jinglu Song and Bo Huang and Rongrong Li},
  title = {{Measuring Recovery to Build up Metrics of Flood Resilience Based on Pollutant Discharge Data: A Case Study in East China}},
  booktitle = {Water},
  year = {2017},
  doi = {10.3390/w9080619},
}
Figure