|Below, I provide a short overview of the paper, my recommendation, and then more specific comments. |
Broad overview. This paper addresses a critical problem for water resource management, regulatory, and research communities – how do we quantify the spatial variability in intermittence across broad spatial scales? Authors address this by applying a water balance model across two watersheds in Australia: Southeast Queensland (temperate to subtropical climates) and Tamar River of Tasmania (temperate climate). The authors modeling approach can be divided into three steps: (i) developing the AWRA-L runoff simulations, (ii) comparing AWRA-L runoff simulations with and without routing, and (iii) quantifying flow intermittency using daily and monthly outputs from the AWRA-L runoff model and comparing those outputs to the WaterDyn model. The authors highlight their model fit using flow regime metrics across multiple gages in each region, highlight that [atleast in this exercise] a routing model is was not necessary to simulate streamflow, and finally demonstrate spatial and temporal trends in intermittency across the two regions.
Recommendation. My recommendation is to accept with minor revisions. I am reviewing this paper for the first time, but it is in its second round of revision. This paper provides a unique solution to estimating large scale spatial and temporal patterns in intermittency. Further, authors should be commended for quantifying the fit [or in some cases lack thereof] between the simulated and observed streamflow regime. My one major concern is the method used to define the threshold for intermittence. I hope the authors will consider providing more information to their readers. Below, I provide several comments in an attempt to help authors improve an already good manuscript.
-At L205, authors outline their process of defining ‘thresholds’ for intermittence in the simulated flow data. I encourage the authors to provide information about the distributions of the flow quantiles of these thresholds (either intext, or better yet, a figure!).
-Related to the last comment, authors need to provide more information about the linear models. What were the model fits like, were a split sample design employed, and did the authors consider spatial autocorrelation? The reader currently cannot adequately evaluate this portion of the analysis, which is a critical step when defining intermittency.
-It’s unclear why authors use parametric comparative statistics (i.e., t-test). I would suggest switching to a nonparametric analogs (i.e., Wilcoxon test) or showing normality to prove t-test is appropriate.
- After my first read-through, I was unclear why the WaterDyn model was incorporated into the manuscript. While this is reported in L220, I would strongly recommend authors make this more apparent [i.e., not hidden at the bottom of a paragraph.]
-Related to the comment above, there is a lot of important information packed in the paragraph starting at 370. [Infact, I would go so far as to say this is one of the main findings of the paper!] My suggestion is to rewrite this paragraph; lead with the argument that “our results suggest that the temporal resolution of analysis should be dictated by the resolution of input streamflow data;” and potentially further develop this discussion point. This is a useful result – one that will help others build on this studies results! Authors might even consider adding this result to the abstract.
-Finally, I was disappointed the authors did not do more to explore the spatial variation of intermittence in their data. I encourage them to add an additional figure to highlight spatial variability. Note, I’m not suggesting they add an additional map. More appropriately, I would encourage authors to add a figure that quantifies spatial variability in intermittence.