Modeling Spatio-Temporal Dynamics of BMPs Adoption for Stormwater Management in Urban Areas

Round 1

Reviewer 1 Report

The authors have prepared a paper that should be of interest to many readers of Water.  However, the authors may wish to address the following:

1 - The terminology used is US focused while the issue and approach being addressed is an international issue.  Mention of SUDS and WSUD would mitigate the US bias.

2 - The discussion of the catchments does not include a description of brown-field development being undertaken in the catchments.  As the implementation of BMPs is enhanced by brown-field development, differences in brown-field development between the catchments are an integral component of the discussion.

Author Response

Thank you for your review of this manuscript. We have incorporated SUDS, WSUD, and SCMs into the introduction, and brown-field development in the conclusion.

Line 51-57, In view of their spatially-extended nature, NPS processes are best addressed via control measures that are similarly distributed spatially throughout the landscape. These measures are referred to as best management practices (BMPs), green infrastructure (GI), low-impact development (LID), sustainable urban drainage systems (SUDS), water sensitive urban design (WSUD), or stormwater control measures (SCMs), and their implementation seeks to control runoff, urban stormwater, and NPS pollutants, either individually or jointly

Line 687-697, Overall, this research quantitatively demonstrates spatio-temporal patterns of BMP adoption in urban areas. This research confirms that a “healthy” watershed will treat the NPS pollutants effectively and emphasizes the importance of social intervention in the “unhealthy” watershed or census tracts. More work is needed in the pollutant hotspots to promote BMP adoption and achieve NPS-control goals. Beyond the scope of residential areas, the cleanup and transformation of non-residential zones such as industrial regions, termed “brown-field development”, offers significant potential benefits. However, this research did not include an assessment of the “brown-field” areas within the study watershed. Future research should explore the economic incentives and environmental remediation possibilities associated with such initiatives.

Reviewer 2 Report

1-     The abstract provides a clear and concise summary of the research conducted, highlighting the importance of nonpoint source pollution and the need for targeted social interventions to improve BMP adoption. However, there are a few areas that could be further clarified or expanded upon. Firstly, it would be helpful to provide more specific information on the socio-economic factors examined in the study. What variables were considered and how were they measured? Similarly, it would be useful to know more about the physical attributes of the two watersheds studied and how they influenced BMP adoption.

2-     Secondly, it would be beneficial to provide more detail on the methods used to analyze the data and build the regression models. For example, what specific variables were included in the models and how were they selected?

3-     It would be useful to highlight any implications or recommendations that can be drawn from the findings. For example, what specific social interventions may be effective in improving BMP adoption in different socio-economic contexts?

4-     Introduction provides a detailed overview of the challenges associated with water quality and quantity control in natural and constructed landscapes, and the importance of Best Management Practices (BMPs) in addressing these challenges. The objective of developing a predictive model of BMP adoption likelihood to guide social intervention efforts is also an interesting and valuable contribution to the field. However, I would suggest that the introduction could benefit from some clarification in terms of the organization and structure of the manuscript.

Specifically, it may be helpful to provide a clear roadmap of what the reader can expect in terms of the different sections of the manuscript and how they are structured. Additionally, while the literature review provides a comprehensive overview of previous research on BMP adoption, it may be helpful to more explicitly state how this study builds on and contributes to existing knowledge in the field.

- The study area section is too long and can be summarized.

- There are no distinct materials and methods and results and discussion sections.

- It would be useful to know more about how the physical and demographic features were selected for inclusion in the model. The authors briefly mention that these factors have been proposed as drivers of BMP adoption, but it is not clear how they determined which specific features to include. Did they conduct a literature review or consult with experts in the field? Did they consider other potential factors that may influence BMP adoption but were not included in the final model?

- It would be helpful to have more information on how the models were evaluated and compared. The authors mention that "common diagnostic statistics" were used, but it is not clear what specific metrics were used and how they were interpreted. How did the authors determine which model was the "best" approach?

- It would be valuable to have more information on how the sensitivity analysis was conducted and what insights were gained from it. Which specific features were found to have the greatest impact on BMP adoption likelihood, and how did this inform the development of the reduced model?

-  Additionally, were any interactions between different features investigated, and if so, what were the results? More details on these aspects of the analysis would help readers better understand the implications of the study's findings.

-  More discussion about the results is needed.

Author Response

1-     The abstract provides a clear and concise summary of the research conducted, highlighting the importance of nonpoint source pollution and the need for targeted social interventions to improve BMP adoption. However, there are a few areas that could be further clarified or expanded upon. Firstly, it would be helpful to provide more specific information on the socio-economic factors examined in the study. What variables were considered and how were they measured? Similarly, it would be useful to know more about the physical attributes of the two watersheds studied and how they influenced BMP adoption.

Thank you for your insightful comments. We have incorporated more detailed information about physical and socio-economic attributes into the model to better elucidate their influence on BMP adoption behaviors within the two watersheds. We revised the manuscript to reflect these points as follows clearly:

Line 16-25, The best performance model (random forest regression, R2=0.67, PBIAS=7.2) was used to simulate spatio-temporal patterns of household BMP adoption in two nearby watersheds (Watts Brach between Washington, D.C., and Maryland and Watershed 263 in Baltimore), each characterized by different socio-economics (population density, median household income, renter rate, average area per household, et al. Table 3) and physical attributes (total area, percentage of canopy in residential area, average distance to nearest BMPs et al. Table 3). BMP adoption rate was considerably higher in Watts Brach watershed (14 BMPs per 1000 housing units) than the Watershed 263 due (4 per 1000 housing units) due to distinct differences in the watershed characteristics (lower renter rate and poverty rate, higher median household income, education level, and canopy rate in residential areas).

2-     Secondly, it would be beneficial to provide more detail on the methods used to analyze the data and build the regression models. For example, what specific variables were included in the models and how were they selected?

- It would be valuable to have more information on how the sensitivity analysis was conducted and what insights were gained from it. Which specific features were found to have the greatest impact on BMP adoption likelihood, and how did this inform the development of the reduced model?

Variables that are included in this study are based on the literature review and based on the availability of proper data for those variables. We initially choose 4 physical features and 17 demographic features (Table 3). Most of these features are obtained from the 2010 U.S. Census and 2010 ACS 5-year survey. For the linear regression model, we employ a statistical t-test to determine the significance of each variable. This test provides a p-value for each variable, reflecting the probability that the variable's actual effect is null (insignificant). By selecting only those variables with p-values less than 5%, we adhere to a common convention in statistical analysis: accepting as significant only those variables that have less than a 5% probability of having a null effect (i.e., there is at least 95% probability that they have significant effect). This technique aids in preventing the model from overfitting by incorporating irrelevant variables. For Lasso regression, Ridge regression, Support Vector Regression (SVR), and random forest regression, we included all available variables in census tracts as these techniques have built-in mechanisms for handling a large number of predictors. For example, feature importance of all variables in the random forest regression analysis is displayed in figure 4.

3-     It would be useful to highlight any implications or recommendations that can be drawn from the findings. For example, what specific social interventions may be effective in improving BMP adoption in different socio-economic contexts?

-  Additionally, were any interactions between different features investigated, and if so, what were the results? More details on these aspects of the analysis would help readers better understand the implications of the study's findings.

-  More discussion about the results is needed.

We have added more detailed discussions about social interventions based on different socio-economic factors in Conclusions (lines 666-687).

4-     Introduction provides a detailed overview of the challenges associated with water quality and quantity control in natural and constructed landscapes, and the importance of Best Management Practices (BMPs) in addressing these challenges. The objective of developing a predictive model of BMP adoption likelihood to guide social intervention efforts is also an interesting and valuable contribution to the field. However, I would suggest that the introduction could benefit from some clarification in terms of the organization and structure of the manuscript.

Specifically, it may be helpful to provide a clear roadmap of what the reader can expect in terms of the different sections of the manuscript and how they are structured. Additionally, while the literature review provides a comprehensive overview of previous research on BMP adoption, it may be helpful to more explicitly state how this study builds on and contributes to existing knowledge in the field.

There are no distinct materials and methods and results and discussion sections.

We have restructured the manuscript following the conventional academic structure in order to make it easy for the readers as follows:

1. Introduction
2. Study Areas
3. Materials and Methods
4. Results and Discussion
5. Conclusion
6. References

5- The study area section is too long and can be summarized.

The study area section has been shortened and summarized.

6- It would be helpful to have more information on how the models were evaluated and compared. The authors mention that "common diagnostic statistics" were used, but it is not clear what specific metrics were used and how they were interpreted. How did the authors determine which model was the "best" approach?

The accuracy of all models were evaluated using 3 commonly used metrics: the coefficient of determination (R²), percentage of bias (PBIAS), and mean square error (MSE).

Coefficient of Determination (R²): This metric measures the proportion of the variance in the dependent variable that is predictable from the independent variables. In other words, it quantifies the degree to which your model explains the variation in your data. An R² value of 1 indicates a perfect fit, meaning the model explains all the variability of the response data around its mean.

Percentage Bias (PBIAS): PBIAS measures the average tendency of the predicted data to be larger or smaller than their observed counterparts. It's a particularly useful measure when you want to see if your model is consistently overestimating or underestimating the actual values. A PBIAS of zero indicates accurate model prediction.

Mean Square Error (MSE): MSE is a common measure of prediction error, which quantifies the difference between the values predicted by the model and the values actually observed. It accomplishes this by squaring the difference between predicted and observed values (to eliminate any negative values) and then averaging those squared differences over all data points. A lower MSE indicates a better fit to the data.

We used all these metrics and found that random forest regression was the best model, with R² (0.67), PBIAS (-7.24), and MSE (19.11).  Please note lines 301 through 316 for details.