Geographically Weighted Flow Cross K-Function for Network-Constrained Flow Data
Round 1
Reviewer 1 Report
The paper is a nice extension of the flow Cross K-function by considering road network and distance decay. Overall, the paper is well-written and easy to follow. There are some places in the paper need clarification or justification.
1. As the title suggests, this method is novel on two aspects: network constrained and geographically weighted. However, the paper did not make a quantitative comparison between the original flow Cross K-function (Tao and Thill 2019) and their upgraded version. This is the biggest problem in my opinion. As a methodological paper, how could you convince that your extension is meaningful and effective without comparison? I’d highly suggest the authors use either the synthetic data or taxi data or both to cross compare: 1. Flow Cross-K; 2. Network-constrained flow cross-K; 3. Geographically weighted flow cross-K; 3. Geographically-weighted & network constrained flow cross-K (this method). Such comparison can justify the need of this new method, and make it clear which is more important: network or geographic weight.
2. Gaussian kernel density function was chosen to calculate weights. But it lacks specifications: i.e., how it works and how the parameters (e.g., bandwidth) are selected.
3. Only 99-times permutation? Why not 999-times as most other papers do?
4. Please add some discussion of the impact of significance level. How much would the results change if choosing 0.01 or 0.001 instead of 0.05?
5. What are the grey flows in Fig 1a and 1b?
6. Fig 2e is counter-intuitive. Why is it dispersed? For every type-A flow, there is at least one type-B flow nearby. My guess is clustered. Please explain it.
7. By looking at the flow data map in Fig. 5, my guess was that GWFKtaxi,ride-hailing must be clustered because for each taxi flow, there are many ride-hailing flows nearby. I also guess that GWFKride-hailing,taxi can be clustered or dispersed because some ride-hailing flows are surrounded by taxi flows but some are not. However, the results in Fig. 6 are opposite to my observation. Any explanation?
Author Response
Dear Reviewer,
We deeply appreciate the effort and time you and the reviewers have spent in reviewing our manuscript (ID: applsci-2017033). Indeed, these detailed comments from the reviewer are really helpful for further improving the quality of the manuscript. Our detailed responses to the comments are listed in the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
This an interesting paper that introduces new, updated methodology for network constrained spatial flows, and upgrades the cross K-function. It is well-organized paper with illustrative examples. I would suggest the following changes:
- line 27: delete "cross k-function" (already in the title)
- Method section: even though you first state that Euclidean distance is not suitable as measurement, you still use it to some extent. Please do explain the reason more clearly.
- 2.2. This subsection is the essence of your research and you should clearly state what you have updated compared to the original version. The explanation should encompass both global and local scale, and be more elaborate than the current version.
- 2.3. Subsection: are there some other options apart from Monte Carlo simulation (and why are not these used)? What was the reason for setting the threshold on number of simulations on 99.
Author Response
Dear Reviewer,
We deeply appreciate the effort and time you and the reviewers have spent in reviewing our manuscript (ID: applsci-2017033). Indeed, these detailed comments from the reviewer are really helpful for further improving the quality of the manuscript. Our detailed responses to the comments are listed in the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The revision directly addressed my earlier concerns. I'm satisfied with the current form.
