the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Oct 2025
Characterizing runoff response to rainfall in permafrost catchments and its implications for hydrological and biogeochemical fluxes in a warming climate
Abstract. Understanding how Arctic catchments respond to rainfall is critical for anticipating hydrological and biogeochemical effects of a warming climate. We use ensemble rainfall-runoff analysis (ERRA) to identify how runoff response to rainfall varies with meteorological, subsurface, and geomorphic conditions across three permafrost catchments: Upper Kuparuk (Alaska) and the Goose and Ptarmigan catchments (Cape Bounty, Canadian High Arctic). ERRA enables us to quantify event-scale runoff responses to rainfall using high-resolution, multi-year hydrometeorological datasets, and test how variations in rainfall intensity, thaw depth, antecedent wetness, and active layer detachments (ALDs) affect runoff behavior. Our results show that peak runoff response increases by more than five-fold in response to increases in antecedent streamflow (a proxy for antecedent moisture), and is also higher in summers with higher average precipitation. By contrast, warmer winters and springs, likely linked to deeper thaw and increased subsurface storage capacity, are associated with reduced runoff sensitivity to rainfall. Furthermore, a paired watershed comparison shows that streamflow and riverine fluxes of dissolved solids, suspended sediment, and particulate organic carbon are more readily mobilized by rainfall inputs when ALDs are present. Considered together, these findings highlight the difficulty in generalizing climate-driven runoff trends in permafrost regions subject to competing and interacting controls, such as precipitation intensity, storage capacity and permafrost stability. Our findings offer a more nuanced alternative to broad classifications of Arctic landscapes as “drying” or “wetting” under climate change.
- Preprint
(3869 KB) - Metadata XML
-
Supplement
(1748 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-4275', Anonymous Referee #1, 19 Nov 2025
-
AC1: 'Reply on RC1', Cansu Culha, 28 Feb 2026
We thank Reviewer 1 for their detailed, thoughtful, and constructive review of the manuscript. Their comments prompted substantial revisions that significantly strengthened the paper. In response, we clarified the scope of the study as explicitly rainfall-driven runoff response, expanded and updated the literature review to better reflect recent work across Alaska, Canada, Scandinavia, China, and Russia, refined the interpretation of antecedent conditions in permafrost catchments, and more clearly articulated the novelty and limitations of the ERRA framework in this context. We believe these changes substantially improve both the clarity and positioning of our contribution within the existing literature.
A central takeaway from Reviewer 1’s comments is that, although the underlying hydrologic processes examined are well established in the permafrost literature, the manuscript did not initially articulate clearly how ERRA provides new insight beyond existing conceptual and modeling frameworks. The revisions therefore focus on clarifying that the primary contribution of the study lies in using ERRA as a diagnostic tool to quantify event-scale runoff and material-flux responses under variable atmospheric and physical conditions, including nonstationary thaw and disturbance conditions, rather than in proposing new runoff-generation mechanisms.
The paper entitled "Characterizing runoff response to rainfall in permafrost catchments and its implications for hydrological and biogeochemical fluxes in a warming climate" utilizes the ensemble rainfall-runoff analysis (ERRA) framework across three permafrost-underlain catchments to understand how rainfall driven runoff varies with biophysical and precipitation. The authors utilize data from the large and well-studied Upper Kuparuk River in Alaska and two very small watersheds in the Canadian high arctic; one of which is impacted by thermokarst.
ERRA is a new method and I will take the approach and framework 'as is' and will not present any criticisms to its methodology. Like a unit hydrograph, it provides a response function of catchments runoff (or other factors) to precipitation inputs, yet can handle nonlinear, non-stationary, autocorrelated and heterogenous systems as outlined an earlier paper that has been published in HESS. As noted, this manuscript seeks to use ERRA for three permafrost underlain watersheds, which are inherently non-linear and heterogeneous, to improve our process understanding and interrogate the influence of phenomenon like active layer detachments.
I read this manuscript several times with considerable interest. There are few people working in cold environments, and applications of new and novel methods are welcome.
Thank you for appreciating the methodology.
That said, there is considerable process understanding developed from decades of hard-won observation and careful analysis. As it stands, I do not believe that his paper is suitable for publication in its present form. I have a number of comments that I hope the authors will consider to improve this manuscript to make it more impactful to the permafrost hydrology and catchment science community.
We want to acknowledge the current state of the art and make sure that we clearly communicate the novelty of our research. We look forward to improving the manuscript with your feedback.
First, and I hesitate to say this, the writing is colloquial and perhaps not up to publication standard. Suggesting improved writing is an easy throw-away comment, but I found it very causal with many interruptions. More detailed comments are below, and I strongly suggest that the authors separate the Results and Discussion to remove speculative comments and to identify where this work sits within the broader and more recent published literature. At times, I was jumping back and forth between what were new results, results that supported the literature (if referenced) and speculation.
Comments by line:
L51. Why ‘seemingly’. Is permafrost not impermeable? Or relatively so? If not, please provide references.
We thank the reviewer for this feedback. We now write “continuous permafrost… are often largely impermeable, and thus can rapidly transport…”. We now cite Woo, M.-K., Marsh, P., & Pomeroy, J. W. (2000). Snow, frozen soils and permafrost hydrology in Canada, 1995-1998. Hydrological Processes, 14(9), 1591–1611.
L55. Wang et al. 2021 is referenced twice.
Thank you for catching this mistake.
L57. Wolvord & Kurylk is 2016.
Thank you for catching this mistake.
L57. Is there a referencing order/format? It seems random here and in other places.
We have standardized the referencing order/format.
L60. It is likely Koch not Coch et al. 2018
We appreciate the reviewer’s suggestion. We reference Caroline Coch in Coch et al., 2018 this is a different person from Joshua Koch. The reference should be correct.
L61. It is likely DeBeer et al. 2016 not De Boer et al. 2016. Neither are in the reference list.
Thank you for recognizing this mistake.
L66. The authors in the previous line discuss rainfall, and here discuss how peak flow and flood response relate to permafrost thaw. Clearly, the expansive permafrost literature indicates that peak flows are most often (although obviously not always) associated with spring snowmelt freshet. Higher peak flows are most likely associated with increased snowfall and/or intense melt along with rain on snow, not summer rainfall. There is literature that discusses this.
The goal of this paragraph is to discuss peak flow following a rainfall event. We acknowledge that peak flow usually happens during spring snowmelt, but our work does not investigate the spring freshet. To make the point clearer, we write, “peak flows following rainfall events”.
L68. I am unsure as to why the ‘peak streamflow’ is discussed here as noted above. Peak streamflow in almost all permafrost rivers is associated with melt, not rainfall, except at potentially very large scales and in certain exceptional years.
We apologize for the confusion, we are characterizing peak flow following rainfall events, so we added “peak streamflow following a rainfall event”.
L70. This will arise later, but the authors focus on thermokarst as a key process affecting flows. I would like to push back against this. While a startling and important process at local (and even scales up the ~100 km2), thermokarst is not a pan-arctic driver of changes in flows. I’m happy to be proven wrong. There is an observation bias simply because of its dramatic effect, yet it is confined to ice-rich regions, and there are many areas underlain with permafrost that are not influenced by thermokarst. Probably the vast majority of the arctic.
Thank you for this comment. Our goal is not to suggest that the entire Arctic landscape will experience thermokarst, but that in regions with thermokarst, even more specifically Active Layer Detachment slides, we may see higher peak streamflow following a rainfall event. To make this more clear, we wrote “potentially accelerating landscape erosion”. The sentence now reads, “On the other hand, understanding how climatic shifts may change the timing and magnitude of peak streamflow following a rainfall event is challenging due to confounding effects of changing precipitation patterns, increasing storage, and potentially accelerating landscape erosion as permafrost thaws”.
L72. I’m unsure as to the terminology here. ‘granular pore-scale storage capacity’. Is storage capacity not a clear enough term?
Here we use the phrase ‘granular pore-scale storage capacity’ simply to distinguish soil-matrix storage (i.e., water stored within pore spaces of mineral and organic grains) from other forms of storage such as macropores, subsurface cavities, and surface water bodies. For simplicity, we now use “soil-matrix storage capacity”.
L74. Again, ‘pore-scale’ storage capacity? I’m curious as to what other storage capacity would be in play here?
Thank you, see above comment.
L76. I agree that thermokarst affects runoff at local scales, but is there any supporting literature as to its ubiquity and influence at larger scales? I’m unsure if there is.
We agree that we do not know of any literature that has examined the impacts across scales beyond the local scale, nor has there been any comparison of the effects of different types of thermokarst on runoff (gullying, retrogressive thaw slumps, active layer detachment slides, disintegration of polygonal terrain, etc). We have clarified the purpose of the manuscript to make this the focus of the study.
L80. There has been considerable ground-based observations in permafrost underlain regions. The Upper Kuparuk River and its sub-basins are some of the most well studied in the world, but the literature citations here are almost 30 years old and neglect much of the newer Alaska North Slope work that looks at runoff, chemical fluxes and seasonal responses. Simple searches, articles by Shorgren, etc., will reveal a wealth of process information not cited but pertinent.
We now include much more recent research in our manuscript. We include a wide range of authors like Shorgren, Tank, Grewal, Kokelj, Quinton, and Youcha. The paragraph on L80 is replaced with: “Ground-based observations and modeling are commonly used to understand controls on streamflow dynamics in permafrost landscapes, each with distinct strengths and limitations. Long-term ground-based studies have provided some of the most detailed hydrologic, chemical, and biogeochemical observations available for permafrost catchments. These studies have documented strong seasonal controls on runoff generation, solute transport, and flowpath activation associated with active layer development and changing hydrologic connectivity (e.g., McNamara et al., 1998; Kane et al., 2000; Stieglitz et al., 2003; Shogren et al., 2019, 2021, 2024; Tank et al., 2020; Beel et al., 2021; Youcha et al., 2023; Grewal et al., 2025). Runoff and material flux responses in Arctic catchments reflect complex interactions among precipitation timing, antecedent wetness, thaw depth, and spatially variable subsurface permeability, often resulting in nonlinear or near-chemostatic concentration–discharge behavior (e.g., Shogren et al., 2019, 2021; Tank et al., 2020; Grewal et al., 2025). Conceptual and process-based frameworks developed in tundra and permafrost environments likewise emphasize the importance of vertical and lateral contrasts in hydraulic conductivity and threshold-driven hydrologic connectivity in shaping runoff response (e.g., Quinton & Marsh, 1999; Quinton et al., 2000; Kokelj et al., 2013; Kokelj et al., 2017).
"Despite these advances, collecting long-term, high-frequency observational data in Arctic environments remains challenging and costly, and many existing analyses focus on seasonal or annual responses rather than event-scale rainfall–runoff dynamics. Simulation models have advanced substantially in their ability to represent coupled cold-region processes and to project hydrologic change in permafrost landscapes, including recent pan-Arctic runoff simulations and integrated surface–subsurface thermal hydrology models (e.g., Stieglitz et al., 2003; Painter et al., 2013, 2016; Evans et al., 2020; Jan et al., 2020; Rawlins et al., 2021, 2024). However, quantitative constraints on how permeability, hydraulic gradients, and effective subsurface storage evolve at event timescales in response to rainfall and snow/ice melt remain limited (Kurylyk et al., 2013; Favaro et al., 2014; Walvoord & Kurylyk, 2016; Chen et al., 2022). These gaps motivate the use of data-driven approaches that can directly quantify nonlinear and nonstationary runoff responses to rainfall at event scales.”
L87. I’m unsure as to what the linkages between remote sensing and process-scale subsurface characteristics are. Is this easy anywhere? Is remote sensing useful in understand rainfall-runoff response in other environments and not permafrost ones? I’m not following this thread.
We removed how remote sensing is being used in characterizing streamflow dynamics.
L91. There are many models that have worked to predict runoff from permafrost environments, many in the past five years that are not referenced. Papers by Rawlins and Painter are just two of many authors to consider.
We agree and have revised the introduction to better reflect recent progress in permafrost hydrology modeling. We now cite recent pan-Arctic runoff modeling (e.g., Rawlins et al., 2021, 2024) and integrated surface–subsurface thermal hydrology models (e.g., Painter et al., 2016; Jan et al., 2020), while clarifying that remaining uncertainties are concentrated in how effective permeability, hydraulic gradients, and storage evolve under heterogeneous thaw at event time scales relevant to rainfall forcing:
“Simulation models have advanced substantially in their ability to represent coupled cold-region processes and to project hydrologic change in permafrost landscapes, including recent pan-Arctic runoff simulations and integrated surface–subsurface thermal hydrology models (e.g., Stieglitz et al., 2003; Painter et al., 2013; Painter et al., 2016; Jan et al., 2020; Evans et al., 2020; Rawlins et al., 2021, 2024). However, quantitative constraints on how permeability, hydraulic gradients, and effective subsurface storage evolve at event timescales in response to rainfall remain limited (Kurylyk et al., 2013; Walvoord & Kurylyk, 2016; Chen et al., 2022). ”L92. I agree that we do not fully understand runoff processes in permafrost watersheds, but the statement: “we still do not adequately understand how permeability, hydraulic potentials, and subsurface storage capacity will change with rainfall intensity and contributions from snow/ice melt” is somewhat peculiar. There is a wealth of literature showing how subsurface permeability, transmissivity and flow changes with seasonality and active layer development. As thaw progresses, the classical transmissivity feedback processes occur in permafrost catchments as porous near-surface layers rapidly convey water, yet as thaw progresses and water tables decline, runoff also does. There is almost 30 years of literate on this that is not referenced. While I understand that the Cape Bounty watersheds do not follow this conceptual model well, the Upper Kuparuk certainly does as it was in part developed there. The vertical aspect of water transmission and differences in depth-dependent permeability is important in interpreting the results yet is absent from the manuscript.
We thank the reviewer for this careful and well-taken comment. We agree that there is a substantial body of literature demonstrating how subsurface permeability, transmissivity, and flow pathways evolve seasonally with active layer development in permafrost catchments, including the classical transmissivity feedback framework developed and applied at Upper Kuparuk and elsewhere. Our original wording was imprecise and may have inadvertently suggested that these processes are not understood in a general sense, which was not our intent.
Our intended point was narrower: while the seasonal evolution of transmissivity and near-surface flow pathways is relatively well characterized, there remains limited quantitative understanding of event-scale rainfall intensity and the resulting runoff response, particularly under nonstationary thaw conditions and across contrasting Arctic catchments. In particular, vertical variations in permeability and the partitioning of flow between shallow, transmissive layers and deeper storage remain difficult to observe directly and are not well constrained at the temporal resolution of individual storm events.
We agree that the vertical structure of subsurface flow and depth-dependent permeability is important for interpreting our results, and we have revised the manuscript to better acknowledge the existing transmissivity feedback literature and have added additional citations to reflect the breadth of prior work in this area.
We change the text to read: “The mechanistic basis for such models remains incomplete, however, because there remains limited quantitative understanding of event-scale rainfall intensity and the resulting runoff response, particularly under nonstationary thaw conditions and across contrasting Arctic catchments”
L97. I believe that this manuscript is confined to rainfall and should be clearly stated.
We agree and have revised the manuscript to explicitly state that the analysis is confined to rainfall-driven runoff during the summer season. The revised text now clarifies this scope and avoids ambiguity regarding other precipitation or melt processes. “We characterize and quantify how summer runoff in response to rainfall in landscapes underlain by permafrost is affected by precipitation intensity and antecedent effective wetness on timescales of hours to days, as well as precipitation and average temperatures during preceding seasons, the seasonal development of the active layer, and the erosion of the land surface”.
L97. This will arise later, but flows in permafrost catchments do not correspond to antecedent wetness – or at least antecedent storage. I believe the authors understand this and should reframe things. For example, in early June, when flows are high, antecedent water storage is very low, although streamflow (defined as wetness?) is very high. At this time, thaw depths are shallow, there is very little available storage for water. Flows are high and rainfall is rapidly transmitted to the stream (along with water released form the thawing active layer). As the seasons progresses, potential storage increases as the active layer expands. Flows can be equivalent in August and June, yet the ‘antecedent storage’ in August would be many times greater than June - with respect to liquid water anyway, but flows may be the same. Antecedent wetness can vary considerably from antecedent discharge, which is likely true in many places but amplified in permafrost environments. This is a clearly established conceptual model from permafrost catchments that is not well addressed in this manuscript.
Thank you for this comment, we will address this further later in the comments to the reviewer.
L106. I’m wondering when the last time unit hydrographs appeared in HESS? A simple search shows only ‘reflective’ articles. While useful for teaching, the unit hydrograph (to my knowledge) is not widely applied in catchment science.
Part of the purpose of this paragraph was to highlight that “unit hydrographs” are no longer used because they are not readily applied to non-linear, heterogeneous, and non-stationary conditions in the catchment. That is why we add the word “historically”. We believe that still including unit hydrographs in the introduction is helpful to help readers orient themselves in understanding our methods.
L111. Is “real-world” needed?
We replace “real-world” with “observed” runoff.
L123. “Relatively free” of glacial melt inputs? To my knowledge, there are no glaciers in either of these catchments. In addition, the vast majority of arctic watersheds at the headwater scale are free from glacial influence (and aufeis). While glaciers and incredibly important at larger continental and regional scales, and aufeis is a dynamic source of streamflow in catchments with extensive surface-groundwater interactions, the literature referencing here is somewhat confusing. They are not ubiquitous nor globally important across pan-arctic regions.
To address your concern, we changed the text to “Unlike larger Arctic river basins, which may receive glacial melt or aufeis contributions, small headwater catchments such as the Upper Kuparuk generally lack these inputs.”
L130. Are thermo-erosional features common? They are dramatic and not rare, but ‘common’ is likely untrue. There is literature to cite as to their occurrence and terranes vulnerable to thermokarst that can be cited.
We now write “other thermoerosional features occur widely in ice-rich and thermokarst-susceptible terrain” in order to be more specific and we cite more manuscripts including those by Kokelj, Balser, Biskaborn, Bowden, and Krieger. We also add “They can exert disproportionate geomorphic and hydrologic impact and their areal extent and frequency have increased in many regions in recent decades, consistent with more extreme weather events and permafrost degradation (Bowden et al., 2008; Balser et al., 2009; Krieger, 2012; Biskaborn et al., 2019). “
L131. Yokely et al. in review is not in the citations.
We have removed this reference.
L139. The Kuparuk is 139 km2. The Cape Bounty Catchments are 0.18 and 0.21 km2. Nowhere is there a discussion of scale and its influence on the integration of results.
We thank the reviewer for highlighting the importance of scale, which we agree was insufficiently discussed in the original manuscript. The Upper Kuparuk and Cape Bounty catchments differ by nearly three orders of magnitude in area, and this difference has important implications for hydrologic responses across space.
ERRA characterizes runoff response as an emergent, catchment-integrated signal resulting from the aggregation of input (e.g., hourly, daily) responses. As catchment size increases, ERRA increasingly reflects the spatial integration of multiple flow pathways, soil depths, and contributing areas, each with potentially different rainfall thresholds, storage capacities, and timing of response. Consequently, ERRA curves for larger catchments are expected to be smoother and delayed in response, while ERRA curves for small catchments are more sensitive.
We have revised the manuscript to explicitly discuss how catchment scale influences ERRA interpretation, including the role of spatial heterogeneity, travel time dispersion, and the averaging of vertical and lateral flow processes. We also clarify that we do not conduct a direct comparison between Upper Kuparuk and Cape Bounty. The use of these two regions is intended to illustrate how similar analytical frameworks manifest differently across scales, rather than to imply scale-invariant behavior.
In the beginning of the results section, we write “We note that as catchment size increases, ERRA increasingly reflects the spatial integration of multiple flow pathways, soil depths, and contributing areas, each with potentially different rainfall thresholds, storage capacities, and timing of response. Consequently, RRD curves for larger catchments are expected to be smoother and delayed in response, whereas RRD curves for smaller catchments are more sensitive to event-scale variability.”
L144. How would a single tipping bucket rain gauge in a very large catchment influence the interpretation of the results? In summer, storms on the North Slope are convective and likely there are events that are missed/mismatched. Does this have any influence on the ERRA or its interpretation?
We agree that the use of a single rain gauge in a large, convectively influenced catchment introduces uncertainty due to spatial variability in precipitation. In the Upper Kuparuk, localized convective storms may not be fully captured by a point measurement, resulting in mismatches between recorded precipitation and the effective catchment-average rainfall.
For ERRA, this uncertainty primarily introduces additional scatter and timing uncertainty in event-scale responses rather than systematic bias. Because ERRA integrates runoff response over many inputs (e.g., hourly, daily), it is relatively robust to occasional precipitation misrepresentation. We therefore interpret ERRA results at Kuparuk as reflecting integrated catchment behavior rather than event-by-event runoff response to rainfall.
We added, “As ERRA integrates runoff responses across many input events (e.g., hourly), these high-frequency measurements enable characterization of integrated catchment behavior with reduced sensitivity to event-scale uncertainty.”
L145-149. I’m unsure as to the need for this sentence. Likely high frequency measurements are useful everywhere and the historical referencing is somewhat peculiar.
We replaced this with “As ERRA integrates runoff responses across many input events (e.g., hourly), these high-frequency measurements enable characterization of integrated catchment behavior with reduced sensitivity to event-scale uncertainty.”
L150. Limited potential water sources? Most catchments receive snow and rain. Active layer thaw provides water certainly, but please consider comments above.
We updated the sentence to “Unlike larger Arctic river basins that may receive glacial melt or aufeis inputs, the Upper Kuparuk headwater catchment has limited active water sources following spring snowmelt, primarily precipitation and permafrost thaw”
L156. Is it simple?
We updated the sentence to “Following spring snowmelt, the limited number of active water sources allows for a focused analysis of the impacts of rainfall and thaw on permafrost hydrology.”
L158. This is somewhat repetitive from above and could be streamlined with a cleaner site description.
Thank you, we condensed this paragraph and it now reads:
“Unlike larger Arctic river basins that may receive glacial melt or aufeis inputs, the Upper Kuparuk headwater catchment has limited active water sources following spring snowmelt, primarily precipitation and permafrost thaw (McNamara et al., 1998). The catchment typically starts its flow season in May and discharges almost all of its snowmelt by June (Kane et al., 2000), after which roughly two-thirds of summer precipitation leaves the catchment as streamflow (McNamara et al., 1998), implying that evapotranspiration is a relatively small component of the water balance. During peak thaw, continuous permafrost extends to 250–600 m depth with an active layer of approximately 50 cm (Kane et al., 1985, 2009; Nelson et al., 1997; Arp et al., 2015), precluding deep groundwater contributions to summer streamflow, which is instead derived from precipitation, melting permafrost and active-layer ice, and shallow subsurface flow. Following spring snowmelt, the limited number of active water sources allows for a focused analysis of the impacts of rainfall and thaw on permafrost hydrology”.
L161. Permafrost thaws, it does not melt.
Thank you for this comment, while permafrost thaws, the ice within permafrost and the active layer melts. We are consistent with our terminology. We include “ice melting” to indicate a water source; whereas “permafrost thawing” could simply be a change in temperature.
L172. There is an indication of lower vegetation, although vegetation characteristics of the Upper Kuparuk are not given.
Vegetation characteristics are not explicitly quantified here and are noted only as a qualitative distinction between sites.
L179. Is there a need to speculate as to why one site had active layer detachments. Such a small slope difference may not be the reason as other simple topographic features like aspect and convergence can be invoked.
We agree with the reviewer. There is no need to speculate as to why one site had active layer detachments, so we have removed the explanation.
L237. The focus is on rainfall. Somewhere the reader needs to understand that this is a small fraction of the total water export from these watersheds. I doubt these are peak flows.
We thank the reviewer for this clarification request. Our analysis does not imply that rainfall dominates the total annual water export from these watersheds, nor that rainfall-generated flows exceed the spring snowmelt peak. Rather, our focus is on event-scale runoff response during the summer season, when streamflow variability is primarily driven by rainfall inputs following snowmelt. Previous studies at Upper Kuparuk have shown that while snowmelt produces the dominant annual peak flow, a substantial fraction of summer runoff and many of the largest post-snowmelt discharge events are rainfall-driven. We have revised the manuscript to clarify this distinction between annual water export and event-scale peak flows, “While snowmelt can dominate the annual peak discharge in these catchments, rainfall is the primary driver of streamflow variability and event-scale peak flows during the summer season following snowmelt.”
L246. There are 12 predictor variables to related RRD to environmental factors. I am uncertain as to how they were chosen, and there are some concerns I have. Number 1 and 3 appear to be the same. Number 9, 10 and 12 are highly correlated. Is number 11 well captured? I’m curious as to how previous summer rainfall intensity affects current year RRD. In the end, I believe that there should be some rationalizing of these 12 predictor variables. I can think of others from the literature that should be considered such previous season or fall total precipitation.
We appreciate the reviewer’s careful reading and agree that several of the candidate predictor variables are correlated. This was intentional rather than inadvertent. Our goal was not to construct a parsimonious predictive model of RRD, but rather to explore how RRD co-varies with a physically motivated set of climatic and hydrologic descriptors that are known to be coupled in permafrost catchments.
Many of the listed variables (e.g., summer temperature, degree days, thaw rate; or total precipitation and rainfall intensity) represent different but related expressions of the same underlying processes. We include them separately because they capture distinct physical interpretations (e.g., energy availability versus thaw dynamics; cumulative input versus intensity), and because different variables may be more directly observable or transferable across sites. As such, collinearity among predictors is expected and does not undermine their use for interpretive comparison.
Regarding the reviewer’s specific points:
- Variables (1) and (3) are not redundant, as total summer precipitation includes snowfall and rainfall, whereas total summer rainfall isolates liquid inputs during the active layer period. We now make this distinction clear in the manuscript.
- Variables (9), (10), and (12) are indeed correlated, reflecting the coupled nature of summer energy input and thaw progression; we explicitly note that thaw rate and snow depth are used as supplementary variables only due to limited temporal overlap.
- Maximum spring streamflow (11) is included as an integrated proxy for pre-summer water availability and catchment wetness, rather than as a precisely resolved process variable.
We agree that additional predictors (e.g., previous fall precipitation) could plausibly influence RRD; however, our intent was to limit the analysis to variables with direct physical relevance to summer runoff generation and with sufficient data availability for robust comparison. We have clarified in the manuscript that these predictors are treated as candidate, physically motivated descriptors rather than independent explanatory variables, and that the analysis is interpretive rather than predictive, “These variables are treated as candidate, physically motivated descriptors rather than independent predictors; collinearity among them is expected given the coupled nature of precipitation, temperature, and thaw processes in permafrost catchments, and they are used to aid interpretation rather than to construct a parsimonious predictive model.”
L252. Calculating total summer precipitation and rainfall as the average of the available data multiplied by the combined three summer months seems problematic, and is certainly an ‘extrapolation’ of reality. I’d like to see some analysis of how this affects analysis with complete data being downgraded and this method used to assess its influence on RRD. I understand that data provision is a challenge, but the influence of this on the interpretation seems to perhaps be large.
We thank the reviewer for raising this point and agree that missing precipitation data require careful treatment. We would like to clarify that this approach is not intended as an extrapolation of storm structure or intensity, but rather as a normalization step to allow interannual comparison of cumulative seasonal metrics in the presence of uneven data availability.
Importantly, total summer precipitation and rainfall are used here only as contextual, candidate predictors for interpreting variations in RRD, not as primary drivers of the RRD analysis itself. The RRD and ERRA calculations are based on observed rainfall–runoff pairs at hourly scales and therefore do not depend on seasonal precipitation totals. As a result, any uncertainty introduced by estimating seasonal totals does not propagate directly into the RRD computation.
We agree that using only years with complete precipitation records would avoid this normalization, but doing so would substantially reduce the number of analyzable years and weaken the ability to examine interannual variability. Given that these seasonal metrics are used qualitatively rather than for predictive modeling, we consider this a reasonable compromise.
To avoid overinterpretation, we have clarified in the manuscript that seasonal precipitation and rainfall totals are approximate indicators used for contextual comparison, and that conclusions regarding RRD are not sensitive to their precise values.
“These seasonal precipitation and rainfall totals are used as approximate, contextual descriptors to facilitate interannual comparison and do not enter directly into the RRD calculations, which are based on observed event-scale rainfall–runoff pairs.”
L256. Was freshet caught each year? I know this is a challenge.
We agree that capturing the full spring freshet each year in Arctic catchments is challenging. At Upper Kuparuk, discharge measurements begin early in the flow season and typically capture the dominant snowmelt pulse; however, in some years the very onset of freshet may not be fully resolved due to logistical and instrumental constraints common in Arctic monitoring.
L274. I do not believe that this analysis evaluates peak runoff. Rainfall driven runoff perhaps. I also do not know why this paragraph appears here.
Our intent is not to evaluate seasonal or annual peak discharge, but rather to examine event or input-scale increases in runoff and associated material fluxes in response to rainfall inputs.
The motivation for introducing this paragraph is to explain how we can use ERRA to understand carbon and material fluxes: specifically, to quantify how rainfall events mobilize dissolved and particulate constituents, the magnitude of these event-scale responses, and their persistence in time. This complements the runoff response to rainfall analysis by linking hydrologic responses directly to biogeochemical export, rather than focusing on long-term or annual flux totals.
We have revised the manuscript to clarify that this analysis targets rainfall-driven event responses rather than peak runoff in the seasonal sense, and to more clearly motivate the application of ERRA to carbon and material fluxes. We now write, “One motivation for examining rainfall-driven runoff responses is to better quantify event-scale mobilization of carbon and other materials from permafrost landscapes”.
In the introduction we have added a paragraph to motivate the carbon flux analysis:
“Also, while many studies have focused on how permafrost thaw alters annual or seasonal material export from Arctic catchments (e.g., Beel et al., 2020), far less is known about how individual rainfall events mobilize carbon and sediment on timescales of hours to days. Yet such event-scale responses are critical, as short-lived increases in runoff can disproportionately control the timing, magnitude, and form (dissolved versus particulate) of material export. Quantifying how rainfall can shape these event-driven fluxes is therefore essential for linking hydrologic change to biogeochemical export in permafrost landscapes. “L283. Separating the results and discussion would help identify the contributions of this manuscript with respect to the published literature.
We appreciate the reviewer’s suggestion. We have chosen to present results and interpretation together because the ERRA framework yields diagnostics that are most meaningfully interpreted in direct connection with their physical context. Separating results and discussion would require repeating figures and explanations across sections and would risk obscuring the linkage between observed patterns and their hydrologic and biogeochemical interpretation.
This integrated structure follows precedent in prior studies applying diagnostic, data-driven approaches to hydrologic time series, where interpretation is closely tied to the presentation of results. To improve clarity and better situate our findings within the existing literature, we have strengthened the contextual discussion within the relevant sections and added cross-references where appropriate, while retaining the combined Results–Discussion format. We also add more clarity to show the clear contributions of this manuscript.
Figure 2C. This information is a reproduction of 2B and likely not needed. Labelling the ‘potential point of saturation’ is also misleading as there is no empirical evidence for this.
Following the reviewers comments, we are removing this labeling. However, we think 2C is helpful in seeing the linearity of the response. We make this more clear in the manuscript.
L301. How does 57% compare with the published literature, even for this catchment or others form the North Slope.
We add: “This rainfall-conditioned runoff coefficient is consistent with previous water-balance studies at Upper Kuparuk, which report that approximately 60–70% of annual precipitation is exported as runoff due to limited subsurface storage in continuous permafrost terrain (McNamara et al., 1998; Déry et al., 2004; Youcha and Stuefer, 2025).”
Line 302-304. This is hard to contextualize and understand without literature to reference. These are very small numbers. I’m uncertain as to the meaning.
We appreciate the reviewer’s comment and agree that the magnitude of the peak runoff response may initially appear small. The peak amplitude represents the instantaneous runoff response to an individual hour at a specific lag time, not the cumulative runoff generated by the whole rainfall event. Because the runoff response at Upper Kuparuk is distributed over multiple days (~300 hours), the response at any single lag is necessarily small, even though the integrated response is large. The physical interpretation of the peak value is fully defined by the ERRA framework and by its relationship to the integrated runoff coefficient reported in the same paragraph. We revise the manuscript text to help with the interpretation and include the quote from our response, “Because the runoff response at Upper Kuparuk is distributed over multiple days (~300 hours), the response at any single lag is necessarily small, even though the integrated response is large.”
L305. Again, discharge in permafrost catchments is a weak reflection of catchment wetness. I’m not disputing the data, just the process interpretation. Maximum flows (even post freshet) are associated with low antecedent ‘wetness’ in terms of total mm of water available in active storage.
We agree with the reviewer that discharge is not a direct measure of total water stored in the active layer or soil profile, particularly in permafrost catchments where storage is constrained and hydrologic connectivity can change rapidly. In this study, however, antecedent streamflow is not intended to represent absolute catchment water storage. Rather, it is used as an operational proxy for the hydrologic state of the drainage network, including the degree of near-surface connectivity and effective drainage efficiency at the time rainfall occurs.
Our interpretation does not imply that higher antecedent streamflow corresponds to greater volumetric water storage in the active layer. Instead, higher antecedent streamflow reflects conditions under which shallow flow pathways are already active and hydraulically connected, such that rainfall inputs are transmitted more efficiently to the channel. The observed sharpening of the runoff response with increasing antecedent streamflow therefore reflects changes in hydrologic responsiveness, not total available storage. We therefore revise the manuscript to replace the term “antecedent wetness” with “antecedent effective wetness,” clarifying that lagged streamflow is used as an operational proxy for the hydrologic state of the drainage network, reflecting near-surface wetness, connectivity, and drainage efficiency rather than absolute storage. This change aligns the terminology with the process interpretation without altering the analysis. We now write, “Thus, we suggest that antecedent streamflow may be a proxy for antecedent effective wetness, where lagged streamflow is an operational proxy for the hydrologic state of the drainage network, reflecting near-surface wetness, connectivity, and drainage efficiency rather than the absolute storage”.
L310. There is considerable literature from the US, Canada, China, Russia and Scandinavia to reference here that is missing, most of which is more recent than a seminal 54 year old paper. The temperate catchment references in my opinion are not appropriate analogues to these catchments.
We reference a few more manuscripts that we found on hydraulic efficiency with antecedent conditions in permafrost landscapes. We write, “The five-fold increase in peak RRD with antecedent runoff is consistent with other permafrost catchments and cold-region catchments, where antecedent discharge reflects the degree of hydrologic connectivity and effective storage within the active layer rather than soil moisture alone. Numerous studies across permafrost regions have shown that higher antecedent flows are associated with enhanced rainfall responsiveness (e.g., Dingman, 1971; Quinton & Marsh, 1999; Xiao et al., 2022; Massari et al., 2023; Knapp et al., 2024)”.
L315-323. The discussion of water tracks is important, but at the scale of the Upper Kuparuk, there are other issues such as beaded streams and channel routing that need to be considered. There are papers and literature on this. While water tracks are an important feature in this environment, there are other processes that become important as scale increases. In addition, there is no discussion here of depth-dependent hydraulic differences that are well studied and referenced in this environment.
Thank you for highlighting this point and for noting where the presentation caused confusion. We agree that additional clarity is helpful.
Our objective in Fig. 3 is not to analyze event-scale runoff volumes, which are intrinsically controlled by total precipitation amount, but rather to characterize the catchment’s hydraulic response to rainfall intensity. Specifically, we use the nonlinear response function (NRF) to quantify how the magnitude of streamflow response to a short rainfall pulse varies with precipitation intensity and antecedent wetness. This approach is conceptually similar to examining the impulse response of the catchment, allowing us to isolate how efficiently rainfall is converted to runoff under different states of the system.
While each NRF bin reflects the runoff generated per time step at a given rainfall rate (and therefore implicitly includes the rainfall volume within that time step), our analysis does not integrate over entire storm events. Instead, we focus on the peak response within each NRF curve because this metric directly tracks changes in the strength and nonlinearity of the peak runoff response, largely independent of event duration or cumulative rainfall totals. Peak response is particularly informative for diagnosing shifts in catchment behavior in this case to rainfall intensity and gives us a sense of how runoff response may be changing as a result of rainfall intensity as opposed to other processes.
Figure 3 demonstrates that increasing rainfall intensity leads to systematically larger peak NRF values, and that this sensitivity increases under wetter antecedent conditions (Fig. 3c). This indicates that higher-intensity rainfall pulses are converted into runoff more efficiently as the basin becomes wetter, consistent with increasing hydraulic connectivity and reduced transient storage.
We agree that total precipitation volume is an important control on event runoff magnitude, but evaluating event-integrated volumes was outside the scope of the present analysis because it adds a level of complexity we are trying to simplify in this work. Our goal is to focus on diagnosing changes in the intensity of the catchment response rather than predicting flood volumes. We have clarified the text to emphasize this distinction and to better explain that Fig. 3 examines intensity-controlled pulse responses rather than storm-integrated runoff totals.
L345. I’m unsure as to the value of this speculative comment.
We removed the sentence: It is possible that runoff response becomes more strongly nonlinear at higher precipitation intensities than those shown here, but there are not enough of these events to robustly quantify the non-linearity at higher rainfall rates.
Figure 4 is interesting as there is considerable variability, but I am not sure that this is particularly clear to the reader or interpreted correctly. The 12 potential explanatory variables are used to provide a reason for this variability (with some factors having less data than others). Line 369 suggests that total summer precipitation (rainfall) is a fist order control on peak runoff response to rainfall. The variance simply could be because in one year all the rain was in the early season and in another it was all in the late season. The emphasis of interannual differences of event scale metrics can be highly influenced simply by timing. Was this considered?
We agree that seasonal rainfall timing can influence event-scale runoff responses and contribute to interannual variability. This effect is explicitly examined later in the manuscript, where we analyze how runoff response varies across summer months. While interannual differences in rainfall timing are present (Fig. 6), their magnitude is small relative to the variability in peak runoff response shown in Fig. 4. Specifically, Fig. 4 exhibits approximately an order-of-magnitude variation in peak runoff response, whereas Fig. 6 indicates that seasonal timing accounts for on the order of ~25% variability. This comparison suggests that rainfall timing alone is insufficient to explain the observed spread in event-scale runoff response.
Also, the discussion relating to snow versus rain is confusing things. Is there much rain below 2 degrees? If so, these values should be clearly presented and this seems more like a repetition of methods.
Thank you, we removed that sentence.
L373. The average temperature during the preceding winter and spring as a predictor as a key factor controlling RRA seems somewhat peculiar and looking at Figure 5c suggests a very weak relationship. An average winter temperature, in such cold environments, has very little impact on summer thaw and active layer development. There are many decades of work on heat transfer in permafrost soils that would suggest that summer temperature and thew rates far exceed any antecedent winter thermal conditions. The thought that a cold or a warm winter in such extremely cold environments overrides the ‘in season’ climate needs some considerations. There are other factors like previous season or late season precipitation that process studies have used to explain current season responses that make more process sense. This goes to the selection of the 12 factors - which could lean a bit more on the process literature.
We agree with the reviewer that, in permafrost environments, summer energy input and thaw rates exert the primary control on active layer development. We now write:
“Winter–spring temperatures may act as an integrated antecedent thermal condition that influences snowpack evolution, the timing of early-season thaw, and the initial hydrologic state of near-surface flow pathways entering the summer season. In this sense, average winter–spring temperature may modulate the efficiency with which rainfall is converted to runoff during summer events, rather than directly controlling active layer development.”We also acknowledge that the relationship between peak runoff response and winter–spring temperature is weaker than that with summer precipitation, as reflected in Fig. 5c. Nevertheless, when considered jointly with summer precipitation, average winter–spring temperature emerges as a statistically significant covariate, suggesting that antecedent thermal conditions may modulate the peak runoff response. We have revised the manuscript to clarify this distinction and to avoid implying that winter temperature supersedes in-season energy inputs.
Finally, we agree that other antecedent factors, such as late-season precipitation, may also influence current-season responses. Our selection of predictors reflects a balance between process motivation and data availability, and the regression analysis is intended to be exploratory rather than a definitive process attribution.
L404. Yes, this is well established. Work from across Alaska in different permafrost environments from Caribou-Poker Creek to Toolik Lake based catchments show this. More appropriate referencing to this response that is related to active layer thaw, storage and vertical profiles of conductivity would be welcome.
We thank the reviewer for this suggestion and agree that the role of active layer thaw, subsurface storage, and depth-dependent hydraulic conductivity in regulating runoff response is well established across Alaskan permafrost environments, including Caribou–Poker Creek and Toolik Lake–based catchments. In response, we have revised this section to more explicitly situate our results within this body of work.
Specifically, we now reference both classic and more recent studies that document how progressive active layer deepening alters runoff generation by increasing subsurface storage and modifying vertical and lateral profiles of hydraulic conductivity (e.g., Neilson et al., 2018; O’Connor et al., 2020; Sjöberg et al., 2021; Hamm et al., 2021; Jin et al., 2022; Jorgenson et al., 2025). We also clarify that our contribution is not to introduce a new conceptual mechanism, but to quantify the magnitude of this well-established seasonal effect at event scales using ERRA.
These additions place our findings in direct continuity with prior process studies and highlight that, while thaw-related increases in storage reduce rainfall responsiveness through the summer, their influence on peak event-scale runoff response in the Upper Kuparuk is modest relative to controls such as antecedent effective wetness.
L412. This is not peak runoff which would likely be during freshet.
To clarify, we write “runoff response to rainfall “.
L420. The decline has been noted in the literature and is related to active layer development and flow pathways.
We agree that the seasonal decline in runoff response has been noted previously and is commonly attributed to active layer development and evolving flow pathways. Our intent here is not to challenge that interpretation, but to evaluate whether concurrent changes in atmospheric forcing or evapotranspiration could also explain the observed decline. By explicitly examining these factors and finding no consistent seasonal trends, we aim to rule out alternative explanations before attributing the decline to thaw-related subsurface processes, which we discuss subsequently. We add “This diagnostic step is intended to rule out alternative atmospheric explanations before attributing the seasonal decline in runoff response to thaw-related changes in subsurface storage and flow pathways, as discussed below.”
L423. There is more recent literature on evapotranspiration in permafrost watersheds and its coupling to runoff response and seasonality. Certainly, they are decoupled, but the process of runoff generation and ET are well documented.
We now also include Eugster et al., 2000; Liljedahl et al., 2011; Helbig et al., 2020.
L430. The observation that runoff decreases with season in permafrost catchments is one of the most common observations and should be cited with references more recent than 1998 for Upper Kuparuk.
We now include this update:
“These findings support the hypothesis that seasonal expansion of subsurface storage capacity through thawing can decrease runoff response (e.g., McNamara et al., 1997, 1998; Stieglitz et al., 2003; Liljedahl et al., 2011; Jin et al., 2022). This agrees with previous studies in the Upper Kuparuk where sequential storms through a summer had increasing old water contributions (McNamara et al., 1997), and lower runoff ratios (Q/P)(McNamara et al., 1998), as well as more recent work documenting enhanced subsurface storage, delayed flow pathways, and reduced rainfall responsiveness later in the summer in permafrost catchments (Stieglitz et al., 2003; Liljedahl et al., 2011; Jin et al., 2022). Our results are consistent with these findings, but indicate that, even after accounting for event-to-event variability in rainfall magnitude, intensity, and structure, the seasonal influence of thaw depth on peak runoff response is smaller than the variability driven by antecedent effective wetness”.
L434. I disagree with this statement. Antecedent wetness is not a process and its interpretation does not square with process understanding from permafrost watershed research.
We thank the reviewer for introducing this clarity in our text. We changed the text to “Our results are consistent with these findings, but indicate that, even after accounting for event-to-event variability in rainfall magnitude, intensity, and structure, the seasonal influence of thaw depth on peak runoff response is smaller than the variability driven by other states like the antecedent effective wetness, as proxied by antecedent streamflow”.
L437. I am unsure how vegetation enters the discussion of hydraulic conductivity, and the following discussion of infiltration with respect to soil texture is simply speculative in the face of again, considerable literature. The hypothetical paragraph staring line 444 does not advance our underrating of process.
We thank the reviewer for this comment and agree that the original paragraph was overly hypothetical and did not sufficiently anchor the discussion of hydraulic conductivity in the existing literature. In response, we have revised this section to remove speculative examples and to clarify how soil texture and vegetation influence runoff generation through well-documented effects on vertical and lateral hydraulic conductivity, active layer structure, and flowpath connectivity.
We write, “In other permafrost catchments, thawed-layer storage capacity may play a more prominent role in regulating runoff response, depending on soil properties and flowpath structure. Numerous studies have documented that hydraulic conductivity in permafrost terrains varies by orders of magnitude both vertically within the active layer and laterally across hillslopes, reflecting differences in soil texture, ice content, organic horizons, and soil structure (Hinzman et al., 1991; Kane et al., 2001; Stieglitz et al., 2003; Neilson et al., 2018; O’Connor et al., 2020; Sjöberg et al. 2021; Hamm et al. 2021). These contrasts exert strong controls on infiltration capacity, subsurface storage, and the partitioning between rapid surface runoff and delayed subsurface flow. Vegetation is relevant in this context insofar as it co-varies with soil development and organic layer thickness, which influence near-surface hydraulic conductivity and water retention, rather than acting as an independent control on runoff response. As a result, the importance of thawed-layer storage in regulating runoff varies across permafrost landscapes and depends on the combined effects of soil texture, active layer structure, and anisotropy in hydraulic conductivity, rather than on thaw depth alone.”
Another process absent from discussion is one of lateral connectivity, which can be facilitated in permafrost underlain catchments as wetness increases. It is an environment where thresholds dominate. Consider the work of Quinton from the late 90s.
We have revised the manuscript to explicitly incorporate this process, drawing on the conceptual framework developed by Quinton and colleagues (Quinton and Marsh, 1999; Quinton et al., 2000) and linking lateral connectivity to our interpretation of antecedent effective wetness. This addition clarifies that the strong sensitivity of peak runoff response to antecedent conditions reflects not only storage effects, but also the activation of lateral flow pathways under threshold-controlled connectivity.
“In addition to vertical storage effects, runoff generation in permafrost catchments is strongly influenced by threshold-controlled lateral hydrologic connectivity. As soil moisture increases and the active layer deepens, previously disconnected portions of the near-surface soil profile can become hydraulically connected to downslope flow pathways and the channel network, producing nonlinear increases in runoff response (Quinton and Marsh, 1999; Quinton et al., 2000). This behavior is characteristic of permafrost environments, where low permeability at depth constrains flow to shallow horizons and promotes rapid lateral transmission once saturation thresholds are exceeded. In this context, antecedent effective wetness reflects not only the availability of subsurface storage, but also the degree to which laterally connected flow paths are activated at the time rainfall occurs. Such threshold behavior provides a physical basis for the strong sensitivity of peak runoff response to antecedent hydrologic state observed in Figure 2b–c, and may explain why antecedent streamflow exerts a larger control than gradual seasonal increases in thaw depth.”
Section 4.5 seems out of place and an add-on to the manuscript. The inclusion of material transport here at one site is a distraction this late in the manuscript. I think a more focussed analysis on rainfall responses across permafrost watersheds would have provided the potential for process insights than data from a very small single site.
We thank the reviewer for this comment and agree that Section 4.6 required clearer integration with the central focus of the manuscript. We now motivate this section earlier on and have rewritten the transition.
Literature on active layer detachments and their influence on runoff and water quality exist for Alaska and the Northwest Territories that should be cited. Undoubtedly ALD will influence runoff response, although I’m struggling to understand how ERRA sheds new light here on what is going on.
We agree that ALDs are well recognized as hydrologically important landscape disturbances, and our intent is not to suggest otherwise. Rather, our goal is to clarify how ERRA provides a complementary and quantitative perspective on how ALDs modify rainfall–runoff response at the event scale.
Previous studies have clearly documented hydrologic and biogeochemical impacts associated with ALDs. For example, Lamoureux and Lafrenière (2009) showed increased discharge, enhanced hydrologic connectivity among detachment features, and flow impoundment at Cape Bounty, while Lamoureux et al. (2014) demonstrated persistent heterogeneity in runoff following disturbance. More recently, Beel et al. (2018) examined catchment-scale fluvial responses to hydrometeorological variability and landscape disturbance, and Beel et al. (2020) provided a detailed assessment of how thermal and physical permafrost disturbances, including ALDs, alter dissolved and particulate fluxes in High Arctic streams. Together, these studies establish that ALDs can substantially influence hydrologic connectivity, material transport, and seasonal discharge patterns.
Our contribution builds directly on this body of work by addressing a related but distinct question: how ALDs modify the magnitude, timing, and persistence of runoff response to individual rainfall events, and whether this response recovers following stabilization. Whereas prior studies have largely emphasized seasonal discharge behavior, concentration changes, or cumulative fluxes, ERRA enables an event-scale diagnostic that isolates runoff response to standardized 2-hour rainfall inputs. By leveraging high-frequency (hourly) observations and pre–post comparisons, ERRA allows us to separate disturbance-driven hydrologic change from meteorological variability and to quantify changes in runoff “flashiness” during and after ALD activity.
We also place ALDs within the broader context of abrupt permafrost disturbances by acknowledging that other mass movement processes, such as retrogressive thaw slumps, are known to affect catchment hydrology (e.g., Connon et al., 2014; Lafrenière and Lamoureux, 2019; Kokelj et al., 2021). However, these studies typically infer hydrologic change indirectly, through sediment or solute delivery, channel burial or impoundment, or downstream turbidity, rather than explicitly quantifying runoff ratios, peak timing, or rainfall–runoff responsiveness at the event scale.
We have revised the manuscript accordingly to better reflect this literature and to clarify the specific contribution of ERRA. The revised text now reads:
“Lamoureux and Lafrenière (2009) document increased discharge, enhanced hydrologic connectivity among ALDs, and flow impoundment associated with extensive ALD activity at Cape Bounty, while Lamoureux et al. (2014) note persistent heterogeneity in runoff following disturbance. More recent work has shown that permafrost disturbances, including ALDs, strongly influence dissolved and particulate fluvial fluxes and catchment-scale hydrologic behavior (e.g., Beel et al., 2018, 2020). Similarly, other mass movement events such as retrogressive thaw slumps are known to impact catchment hydrology (e.g., Connon et al., 2014; Lafrenière and Lamoureux, 2019; Kokelj et al., 2021), but it has been difficult to quantify these effects at the event scale and to assess whether hydrologic responses recover following stabilization. ERRA enables a complementary event-scale characterization by quantifying how hourly rainfall inputs modify runoff response during periods of ALD activity versus quiescent conditions, and relative to an undisturbed reference catchment.”
L491. There is considerable literature on the geomechanics of active layer detachments. ‘lubricating subsurface layers’ is not particularly technical.
In the new version of the manuscript, we do not go into the geomechanics of active layer detachments.
L502. The statement discussing ALD on flash floods and damage critical infrastructure does not make much sense considering the scale of these ALD. More likely, an ALD will simply block a road or destabilize a foundation as mentioned in the cited literature. I’m not sure the alarmism is well founded and the final comment on L506-508 is beyond what can be stated by the analysis here.
We appreciate the reviewer’s concern regarding scale and potential overinterpretation. Our intent is not to equate ALD activity with basin-scale flood hazards, but rather to highlight that ERRA explicitly quantifies the flashiness of runoff response to rainfall, defined here as the magnitude and rapidity of event-scale runoff response. The paragraph now reads: “Although the spatial extent of ALDs is limited to merely 10% of the catchment, our results demonstrate that ALD activity substantially increases the flashiness of runoff response to rainfall, as quantified by ERRA through amplified peak response. This enhanced flashiness does not imply basin-scale flooding, but it does indicate a greater sensitivity of affected headwater catchments to rainfall inputs, with potential consequences for localized drainage exceedance, erosion, and infrastructure performance at small spatial scales. (e.g., Van der Sluijs et al., 2018; Ackerson et al., 2021; Luo et al., 2021)”.
L520. Yes, rainfall mobilizes solutes and sediments, but it can also dilute them. The picking and choosing of literature here is peculiar considering an extremely large body of work on water quality and flows at catchments (including the Upper Kuparuk) and the development of conceptual models that are not referenced. Again, see more recent work by Shogren.
In response to this comment, we have revised the manuscript to explicitly acknowledge dilution effects, to reference established conceptual models of concentration–discharge behavior in permafrost catchments (including recent work by Shogren and colleagues), and to clarify that rainfall-driven material responses may involve enrichment, dilution, or hysteresis depending on flow-path activation and antecedent conditions. We have rewritten this section to better motivate the novelty of our research.
L546. While it is acknowledged that material flux is positively associated with discharge (because discharge is in the flow calculation) how can you suggest that this is true as there is no association between discharge and the concentrations or values. I’m uncertain as to the reasoning here.
We thank the reviewer for this point. We have rewritten this section to better put the emphasis on material flux response to rainfall instead of discharge.
L560. If you state that the data suggests that rain mobilizes DOC and TDS, how does this square with the line 456 that states that there is no association between rain and the material flux?
We appreciate the reviewer pointing out a confusion in the way the text reads and how this section is motivated. We reorganized the section and include a paragraph that writes:
“Although there does not appear to be a correlation between material fluxes and precipitation, there can still be a relationship that is not well captured in a simple correlation analysis. Rainfall influences material export in Arctic catchments through both mobilization and dilution processes, and the resulting concentration–discharge relationships can exhibit strong hysteresis or near-chemostatic behavior depending on solute availability and flow-path activation (e.g., McNamara et al., 1998; Frey et al., 2009; Shogren et al., 2019; 2021). Event-scale studies have shown that rainfall can generate short-lived but disproportionate contributions to suspended sediment, particulate organic carbon, and solute fluxes (Lamoureux and Lafrenière, 2014; Beel et al., 2018; Xiao et al., 2022), while dissolved constituents such as DOC may remain elevated for days following rainfall (Koch et al., 2021). Collectively, this work demonstrates that rainfall exerts strong control on material export, but the timing and persistence of event-scale flux responses remain incompletely constrained. Here, we extend ERRA beyond runoff to quantify the event-scale material flux response to rainfall, allowing direct comparison of the timing, magnitude, and persistence of DOC, TDS, SS, and POC export following rainfall, independent of background seasonal trends or cumulative fluxes at Ptarmigan and Goose.”In the end, I’m unsure as to what this analysis adds to the already good work of Beel documenting this process at this site. The authors need to emphasize this and also cite work of Bowden, Tank and their team (just two scientists that come to mind).
We thank the reviewer that we should bring in more analysis that has been done in this space. Our goal here is not to do an extensive literature review, but point to the new insights and opportunity. ERRA is able to quantify the event-scale material flux response to rainfall, allowing direct comparison of the timing, magnitude, and persistence of DOC, TDS, SS, and POC export following rainfall, independent of background seasonal trends or cumulative fluxes at Ptarmigan and Goose. We have re-written the section to highlight the specific contributions.
Figure 9 is similar to other conceptual models in the literature over the past 30 years defining active layer expansion and water/material fluxes. Recent ones by Grewal come to mind and older ones by Quinton from an environment similar to the Upper Kuparuk in Canada. For 9c, d and e, the stream seems far above the slope/riparian zone and is a bit confusing. I’d suggest permafrost not be blue but perhaps grey or a color not immediately associated with water.
We agree that Figure 9 builds on a well-established body of conceptual work describing runoff generation, active-layer development, and hydrologic connectivity in permafrost landscapes. We have revised the manuscript to explicitly reference this literature, including foundational conceptual frameworks developed by Quinton and colleagues (Quinton & Marsh, 1999; Quinton et al., 2000; Quinton & Carey, 2008; Quinton et al., 2009) and recent synthesis work linking active-layer thaw, discharge, and stream chemistry across permafrost catchments (Grewal et al., 2025).
We clarify that Figure 9 is intended as a synthesis that connects these established process frameworks to quantified, event-scale rainfall–runoff and material flux responses derived using ERRA, rather than as a new conceptual model of permafrost hydrology. We have also revised the figure and caption to emphasize its conceptual nature and improve visual clarity.
L628. The year to year variation in peak runoff being negatively correlated with average temperatures during pervious winter is a weak (panel 5c) and in-season conditions are a much greater proxy on how rapid thaw would be. The degree day models while empirical are pretty good at predicting thaw so I’m a bit surprised. I suggest that the timing of precipitation within the year has much to do with this variability (early v. late). Winter temperatures have warmed much more dramatically across circumpolar regions over the last several decades, yet no appreciable influence on rainfall-runoff processes have been observed form this warming so I struggle to understand causal mechanisms that warm winters facilitate more rapid thaw.
We agree that in-season conditions, particularly summer temperature, rainfall timing, and event sequencing, exert a more direct control on thaw dynamics and runoff generation than mean winter and spring temperature. We agree that in-season conditions, including the timing of precipitation relative to active layer development (early versus late summer), are expected to influence runoff generation and thaw dynamics. We have explicitly examined this effect by comparing runoff response across summer months (Fig. 6). While some seasonal modulation is evident, we find that differences associated with rainfall timing within the summer are modest relative to the interannual variability in peak runoff response and substantially smaller than variability driven by antecedent hydrologic state.
We have revised the manuscript to clarify that, although precipitation timing may contribute to year-to-year differences, our analysis indicates that early- versus late-summer rainfall timing alone cannot explain the observed variability in peak runoff response. This strengthens our interpretation that antecedent effective wetness and catchment connectivity exert a more dominant control than seasonal timing or background winter temperature metrics.
SM Figure 6. It is extremely hard to interpret either the precipitation or temperature.
Thank you for identifying that the figure is hard to interpret. We have updated it to improve its interpretation.
SM Figure 7. As noted, there is mixed responses between discharge and material/chemical fluxes that confound some of the explanations in the manuscript. DOC is largely chemostatic, perhaps. TDS as well with a bit of dilution, SS flux and POC mobilize a bit at Ptarmigan, but relations are relatively weak.
Thank you, the newer version of the manuscript better captures the novelty of applying ERRA to datasets like these. We agree that concentration–discharge relationships for material fluxes in permafrost catchments are often mixed, weak, or near-chemostatic, particularly for DOC and TDS, as has been widely documented in the literature. This variability is evident in Supp. Fig. 7 and this is why ERRA is a powerful tool to unveil certain catchment characteristics at the eventscale.
We have revised the manuscript to clarify that ERRA is not intended to reinterpret or replace traditional concentration–discharge analyses. Instead, ERRA provides a complementary event-scale perspective by quantifying the timing, magnitude, and persistence of rainfall-driven flux responses, even in cases where concentration responses are muted or chemostatic. This allows us to isolate rainfall sensitivity in material fluxes from background variability and long-term differences in baseflow chemistry, which is difficult to achieve using C–Q relationships alone.
We have adjusted the text accordingly to better reflect this distinction..
Citation: https://doi.org/10.5194/egusphere-2025-4275-AC1
-
AC1: 'Reply on RC1', Cansu Culha, 28 Feb 2026
-
RC2: 'Comment on egusphere-2025-4275', Anonymous Referee #2, 20 Nov 2025
This manuscript describes an application of the Ensemble Rainfall Runoff Analysis framework developed by Kirchner (2022), who is a co-author on this paper. The original theoretical framework was applied in a proof of concept by Kirchner (2024). This is the first application of the ERRA framework to an Arctic environment. The authors utilize long-term rainfall and discharge data from the Kuparuk River (Alaska, USA) and similar data augmented with material concentrations (e.g., DOC, SS) from two smaller catchments in Canada, to explore how well-known seasonal changes in active layer depth and surficial disturbance (active layer detachments) affect runoff and material transport in response to rain events. The authors then speculate on how expected changes in rainfall and surface air temperature may change runoff and material transport in the future. I think the greater value of this analysis is the information it provides about the credibility of the ERRA approach in a new environment. The analysis does less to increase our current understanding of the seasonal hydrology of Arctic headwater landscapes as they function now or as they may function in the future. This paper provides another useful proof of concept of the ERRA framework, which is likely to be of increasing utility as more and longer high-frequency solute datasets are developed to complement high-frequency rainfall and runoff data.
I have numerous suggestions for the authors to consider, that I think could improve this manuscript. In the following comments “L##s” refer to line numbers in the manuscript and are followed by the relevant text. The text following “Reviewer:” is my comment or suggestion.
L51: seemingly impermeable
Reviewer: Vague. Replace with "continuous"? Delete?L 74: Rainfall
Reviewer: Perhaps link the paragraph to the previous one with "On the other hand..."?L85-86: For example, McNamara et al. (1998) investigated rainfall–runoff response over three years at four distinct catchment scales but faced challenges in identifying the causes of annual changes.
Reviewer: Other more recent citations could be added. For example:- Shogren, A. J., et al. (2022), Multi-year, spatially extensive, watershed-scale synoptic stream chemistry and water quality conditions for six permafrost-underlain Arctic watersheds, Earth Syst Sci Data, 14(1), 95-116, doi:10.5194/essd-14-95-2022.
- Shogren, A. J., J. P. Zarnetske, B. W. Abbott, F. Iannucci, R. J. Frei, N. A. Griffin, and W. B. Bowden (2019), Revealing biogeochemical signatures of Arctic landscapes with river chemistry, Sci Rep-Uk, 9, doi:ARTN 12894. 10.1038/s41598-019-49296-6.
L152: or aufeis
Reviewer: To be accurate, perhaps say "or extensive aufeis" as there is some limited, seasonal aufeis in the mid-catchment.L 154-156: Roughly two-thirds of summer precipitation leaves the catchment as streamflow (McNamara et al., 1998), implying that evapotranspiration is a relatively small component of the water balance.
Reviewer: I don’t think you can infer that ET is small on this basis. Given this data only, ET is somewhere between “small” and 1/3 of summer precipitation. Either delete this final statement or clarify that unknown portions of the 1/3 balance went ET, temporary soil storage, and/or deeper percolation (which may occur even in “continuous” permafrost).L21-218: typical environmental time series are characterized by autocorrelated noise
Reviewer: It would be helpful to identify typical sources (types) of this autocorrelated noise. Some are described in Kirchner (2024).L222: where now beta-k
Reviewer: beta-k is not a part of equation 2. What is intended here?L222: Equation 15 of Kirchner, 2022
Reviewer: Equation 2 in this manuscript does not conform to Equation 15 in Kirchner 2022, which is a matrix. Do you mean Equation 11 in the latter?L226: b sub k
Reviewer: This parameter does not appear to be defined, nor is h as an index value.L269: Second instance of "predictor"
Reviewer: Should be "predictorS"? PluralL279-281: Given that the material flux data are only available for 92 individual days, we set the maximum response time to be 5 days (m = 5 days in Eq. (1) ) to avoid overfitting.
Reviewer: Are these single observations on 92 different days? If so, how do you account for the likelihood that identical points in different hydrographs may be associated with different levels of material concentrations? I understand how this works when the time series are continuous/regular and the time intervals are small, but not when the measurements are infrequent/random and the time intervals are large.L295-296: up to an apparent saturation point
Reviewer: It is interesting to note that this asymptote is at a discharge of about 0.11 mm/h which translates to 4.4 m3/s for the Kuparuk. This is well above the median flow and closer to bankfull.L300: 300 hours
Reviewer: This is nearly two weeks (12.5 days). How to subsequent storms and thaw runoff that is independent of rainfall influence this RRD curve?L320-323: This saturation threshold may represent the point at which water tracks become hydrologically connected to stream channels, enabling efficient runoff without requiring full saturation of the entire landscape, consistent with modeling results within the basin (Stieglitz et al., 2003).
Reviewer: What is meant by "efficient runoff" and "full saturation" of the watershed? It is not clear to me that "full saturation of the entire landscape" is a requirement to display the runoff behavior depicted in Figure 2c.L327: Figure 3.
Reviewer: It’s a minor point, but why connect the dots in panel 3c but not 3b?L351: Figure 4.
Reviewer: Why is the lag range in Fig 4a so much smaller than the lag range in either Fig. 3a or (especially) Fig. 2a? Shouldn't these be similar?L 354-355: aggregating hourly Upper Kuparuk time series to 4-hour intervals
Reviewer: Why change the intervals from 10 h bins to 4 h bins?L365-366 (and Fig. S3) In Supp. Fig. 3, we compare each explanatory variable individually with the peak runoff response to rainfall
Reviewer: In Fig. S3 in the Supplement it would be helpful to include the p values for each regression in addition to r values and to exclude a (red) line for those correlations that are not significant.L372-373: The pair of predictor variables with the highest R² value in our multiple linear regression analyses is total summer precipitation and average temperatures during the preceding winter and spring.
Reviewer: More specifically do you mean the best pair of predicator variables ignoring the first-order control variables mentioned in the previous paragraph?L438-441: On the timescale of peak runoff response in Upper Kuparuk (~24 hours), coarse, silty soils with permeabilities around 10 −5 m/s can allow rainfall to infiltrate to depths on the order of tens of centimeters, comparable to the thickness of the active layer, enabling much of the precipitation to be stored in the subsurface
Reviewer: A citation should be provided for this permeability value. While the overlying organic soils in the Kuparuk basis likely have high permeability, the mineral soils are rich in loess and likely behave more like clay-rich soils. Thaw depths as great as cited in line 427 (up to 50 cm) would extend into this loess rich mineral soil. Differences in the hydraulic conductivities of these two materials are substantial (e.g. Hinzman, L. D., D. L. Kane, R. E. Gieck, and K. R. Everett (1991), Hydrologic and thermal properties of the active layer in the Alaskan Arctic, Cold Regions Science and Technology, 19(2), doi:10.1016/0165-232X(91)90001-W.L448-449: These localized pools, due to the high heat capacity of water, can enhance thaw and generate spatial feedback in both storage capacity and runoff routing.
Reviewer: Here and elsewhere, there are statements of fact that are likely to be true but are unsupported by relevant citations.L473: significant
Reviewer: If you use the word "significant" please provide a p value based on a test. Otherwise say "important", "substantive" or similar.L486-488: We hypothesize that it is not the act of stabilization per se that reduces runoff, but rather the gradual thaw and drainage of the exposed ice-rich layer. As the active layer deepens and drains out pore water in the scar zone over time, increasing its capacity to store water.
Reviewer: Is there evidence that the active layer has deepened and dried? Might this also be due to the reestablishment of vegetation and organic layers in the stabilized ALD that begin to insulate the soil in a manner like the undisturbed Goose site?L512: Figure 8.
Reviewer: Perhaps stack panels e and f under c and d in a 3x2 configuration rather than 2x3.L558-559: This implies that rainfall mobilizes DOC and TDS flux at these catchments.
Reviewer: Then why is there no substantial correlation with rainfall described in the previous paragraph (L545-546)? Is there a contradiction here? Or perhaps simply a need to more clearly state that rainfall mobilizes (dislodges) DOC and sediment in place, but runoff is the mechanism responsible for lateral transport of these materials?L583-584: Unfortunately, the sparsity of the available data precludes a clear assessment of differences in runoff response between the periods when the ALDs were active versus stable at Ptarmigan
Reviewer: Given this statement I think any further discussion of what might have happened is entirely speculative. This speculation is not necessarily wrong, but it can't be substantiated and so I think it is best to delete this paragraph.L610-619 in Figure 9: Thawing of the subsurface can increase storage capacity and thus decrease runoff sensitivity to rainfall….etc.
Reviewer: Most of the rest of this lengthy legend repeats interpretation in the narrative of the manuscript and should be deleted here.648-649: they are comparably influential Reviewer:
Replace with "their influence is similar"?L650-651: Interannual variability in runoff response is also significant ( Fig. 4 ), underscoring the need for year-specific predictive modeling.
Reviewer: The intent here is not clear. Interannual variability in runoff response is likely. But if you know the response has changed, why do you need predictive modeling? Perhaps delete this reference to annual predictive modeling here as you take this up again, more cogently, in the final paragraph.Citation: https://doi.org/10.5194/egusphere-2025-4275-RC2 -
AC2: 'Reply on RC2', Cansu Culha, 28 Feb 2026
This manuscript describes an application of the Ensemble Rainfall Runoff Analysis framework developed by Kirchner (2022), who is a co-author on this paper. The original theoretical framework was applied in a proof of concept by Kirchner (2024). This is the first application of the ERRA framework to an Arctic environment. The authors utilize long-term rainfall and discharge data from the Kuparuk River (Alaska, USA) and similar data augmented with material concentrations (e.g., DOC, SS) from two smaller catchments in Canada, to explore how well-known seasonal changes in active layer depth and surficial disturbance (active layer detachments) affect runoff and material transport in response to rain events. The authors then speculate on how expected changes in rainfall and surface air temperature may change runoff and material transport in the future. I think the greater value of this analysis is the information it provides about the credibility of the ERRA approach in a new environment. The analysis does less to increase our current understanding of the seasonal hydrology of Arctic headwater landscapes as they function now or as they may function in the future. This paper provides another useful proof of concept of the ERRA framework, which is likely to be of increasing utility as more and longer high-frequency solute datasets are developed to complement high-frequency rainfall and runoff data.
We thank the reviewer for their thoughtful, detailed, and constructive assessment of the manuscript. We agree with their characterization that this work represents the first application of the Ensemble Runoff Response Analysis (ERRA) framework in an Arctic permafrost environment, and we recognize that this study provides an important test of the credibility and utility of ERRA beyond its original proof-of-concept applications.
We acknowledge the reviewer’s point that many of the broad hydrologic processes examined here, including seasonal active layer development and antecedent conditions, are well established in the Arctic hydrology literature. Our intent is not to redefine these processes, but rather to introduce and evaluate a new diagnostic toolset that allows them to be quantified at event scales that are difficult to resolve using conventional approaches. Furthermore, this manuscript demonstrates two novel applications: (1) the use of ERRA to quantify how active layer detachments (ALDs) modify the magnitude, timing, and persistence of runoff response to rainfall, revealing a five-fold increase in rainfall-driven runoff response during periods of ALD activity; and (2) the first application of ERRA to material fluxes, enabling event-scale characterization of DOC, TDS, SS, and POC responses to rainfall, rather than relying solely on seasonal or annual summaries.
We fully agree with the reviewer that the broader value of this work lies in demonstrating the applicability of ERRA in complex, data-limited Arctic systems, and in illustrating how this framework can be extended beyond discharge to investigate coupled hydrologic–biogeochemical responses. In response to the reviewer’s comments, we have substantially revised the manuscript to more clearly articulate this methodological contribution, better situate our findings within the extensive Arctic hydrology literature, and avoid overstating process novelty where it is not warranted. We believe these revisions have significantly strengthened the manuscript and clarified its primary contribution.
We thank the reviewer again for their careful reading and constructive suggestions, which have materially improved the clarity, rigor, and positioning of this work.
I have numerous suggestions for the authors to consider, that I think could improve this manuscript. In the following comments “L##s” refer to line numbers in the manuscript and are followed by the relevant text. The text following “Reviewer:” is my comment or suggestion.
L51: seemingly impermeable
Reviewer: Vague. Replace with "continuous"? Delete?We thank the reviewer for this feedback. We now write “continuous permafrost… are often largely impermeable, and thus can rapidly transport…”
L 74: Rainfall
Reviewer: Perhaps link the paragraph to the previous one with "On the other hand..."?We thank the reviewer for this comment and have made this change.
L85-86: For example, McNamara et al. (1998) investigated rainfall–runoff response over three years at four distinct catchment scales but faced challenges in identifying the causes of annual changes.
Reviewer: Other more recent citations could be added. For example:- Shogren, A. J., et al. (2022), Multi-year, spatially extensive, watershed-scale synoptic stream chemistry and water quality conditions for six permafrost-underlain Arctic watersheds, Earth Syst Sci Data, 14(1), 95-116, doi:10.5194/essd-14-95-2022.
- Shogren, A. J., J. P. Zarnetske, B. W. Abbott, F. Iannucci, R. J. Frei, N. A. Griffin, and W. B. Bowden (2019), Revealing biogeochemical signatures of Arctic landscapes with river chemistry, Sci Rep-UK, 9, doi:ARTN 12894. 10.1038/s41598-019-49296-6.
We thank the reviewer and include these citations throughout the manuscript.
L152: or aufeis
Reviewer: To be accurate, perhaps say "or extensive aufeis" as there is some limited, seasonal aufeis in the mid-catchment.We thank the reviewer and made these changes.
L 154-156: Roughly two-thirds of summer precipitation leaves the catchment as streamflow (McNamara et al., 1998), implying that evapotranspiration is a relatively small component of the water balance.
Reviewer: I don’t think you can infer that ET is small on this basis. Given this data only, ET is somewhere between “small” and 1/3 of summer precipitation. Either delete this final statement or clarify that unknown portions of the 1/3 balance went ET, temporary soil storage, and/or deeper percolation (which may occur even in “continuous” permafrost).We thank the reviewer and made this change. The sentence now reads “The catchment typically starts its flow season in May and discharges almost all of its snowmelt by June (Kane et al., 2000), after which roughly two-thirds of summer precipitation leaves the catchment as streamflow (McNamara et al., 1998), implying that evapotranspiration, temporary soil storage, and/or deeper soil storage are relatively small components”.
L21-218: typical environmental time series are characterized by autocorrelated noise
Reviewer: It would be helpful to identify typical sources (types) of this autocorrelated noise. Some are described in Kirchner (2024).We thank the reviewer and now write: “Time series runoff data can exhibit autocorrelated residuals because catchments integrate inputs over characteristic response times that can be longer than the sampling interval. As a result, successive observations are not statistically independent; streamflow and other fluxes at a given time partly reflect earlier precipitation and storage conditions. This temporal persistence leads to autocorrelated noise, which can bias uncertainty estimates and overstate statistical significance without necessarily affecting the estimated mean response itself (Kirchner, 2024)”.
L222: where now beta-k
Reviewer: beta-k is not a part of equation 2. What is intended here?The reviewer is correct to point out that the explanation of the series of equations was confusing. There were also notation errors in equation 2. Since the technical details of the autocorrelation correction are already documented in Kirchner (2022) and do not need to be repeated here, we will replace lines 219-226 with “Using the procedure explained in Section 2 of Kirchner (2022), ERRA estimates the residual autocorrelation and corrects for its effects on the coefficients beta_k and their estimated standard errors.”
L222: Equation 15 of Kirchner, 2022
Reviewer: Equation 2 in this manuscript does not conform to Equation 15 in Kirchner 2022, which is a matrix. Do you mean Equation 11 in the latter?As noted above, we will remove this equation and the associated text.
L226: b sub k
Reviewer: This parameter does not appear to be defined, nor is h as an index value.As noted above, we have removed equations 2 and 3 and the associated text.
L269: Second instance of "predictor"
Reviewer: Should be "predictorS"? PluralYes, thank you.
L279-281: Given that the material flux data are only available for 92 individual days, we set the maximum response time to be 5 days (m = 5 days in Eq. (1) ) to avoid overfitting.
Reviewer: Are these single observations on 92 different days? If so, how do you account for the likelihood that identical points in different hydrographs may be associated with different levels of material concentrations? I understand how this works when the time series are continuous/regular and the time intervals are small, but not when the measurements are infrequent/random and the time intervals are large.We thank the reviewer for raising this important point. ERRA can work with data gaps, but it is true that it ignores data blocks smaller than the number of hours or days we interpret the data over (e.g., m+h+1 in Kirchner, 2022). In our analysis, precipitation and discharge are available as continuous daily time series, and material fluxes are indexed to this same daily timeline. Thus, the 92 individual days are not randomly scattered through the data set, but had continuous blocks. We set the maximum response window to be 5 days, which means there are multiple continuous blocks that are 5 days long within the 92 individual days we have data. Extending the response window beyond 5 days would substantially reduce the effective sample size. So, the response curves will be representative of the times we have data.
ERRA estimates the ensemble-average response of material flux to rainfall across many events, allowing for event-to-event variability in concentration. This ensemble approach is precisely what makes ERRA suitable for sparse but temporally indexed biogeochemical datasets. We now write, “Given that the material flux data are only available for 92 individual days, we set the maximum response time to be 5 continuous days (m = 5 days in Eq. (1)) to maximize the sample size for the ensemble. The response curves will only apply to the behavior during those times that we have data.”
L295-296: up to an apparent saturation point
Reviewer: It is interesting to note that this asymptote is at a discharge of about 0.11 mm/h which translates to 4.4 m3/s for the Kuparuk. This is well above the median flow and closer to bankfull.Thank you for this observation. The asymptote indeed occurs at Q≈0.11 mm/h, corresponding to ∼4.4 m^3/ s for the Upper Kuparuk. We agree this is substantially above median discharge (which is 3.13m^3/s) and could plausibly be near-bankfull. We have revised the text to describe this feature as an asymptote at high discharge and to clarify that we interpret it as a transition toward high hydrologic connectivity rather than as bankfull condition. We now write “It corresponds to discharge values of 4.4m3 /s for Upper Kuparuk, which is above the median discharge values (3.13m3 /s ).”
L300: 300 hours
Reviewer: This is nearly two weeks (12.5 days). How do subsequent storms and thaw runoff that is independent of rainfall influence this RRD curve?The reviewer brings up a point that shows the novelty of using ERRA. ERRA explicitly accounts for subsequent rainfall by including precipitation at all lags as separate predictors in the regression prior to the hour at which runoff is accounted for, so subsequent storms are accounted for. As a result, subsequent storms do not systematically bias the estimated runoff response. However, at longer lags (on the order of days to weeks), multiple processes with similar timescales, such as thaw-driven drainage and storm clustering, can contribute to runoff variability. While the ensemble approach reduces random overlap among events, interpretation of the far tail of the RRD is therefore more cautious.
L320-323: This saturation threshold may represent the point at which water tracks become hydrologically connected to stream channels, enabling efficient runoff without requiring full saturation of the entire landscape, consistent with modeling results within the basin (Stieglitz et al., 2003).
Reviewer: What is meant by "efficient runoff" and "full saturation" of the watershed? It is not clear to me that "full saturation of the entire landscape" is a requirement to display the runoff behavior depicted in Figure 2c.We agree with the reviewer that “full saturation of the entire landscape” is not a requirement to display enhanced runoff. We also agree “efficient runoff” and “full saturation” of the watershed were not well phrased. We now write: “This asymptotic point may represent the antecedent discharge at which water tracks become hydrologically connected to stream channels, enabling rapid routing of rainfall to the stream with limited additional storage attenuation, consistent with modeling results within the basin (Stieglitz et al., 2003). Thus, we suggest that antecedent streamflow may be a proxy for antecedent effective wetness.”
L327: Figure 3.
Reviewer: It’s a minor point, but why connect the dots in panel 3c but not 3b?We now include lines to connect the dots in all of the figures.
L351: Figure 4.
Reviewer: Why is the lag range in Fig 4a so much smaller than the lag range in either Fig. 3a or (especially) Fig. 2a? Shouldn't these be similar?We interpret “lag range” to refer to the time lag over which the runoff response distribution is displayed. The lag window shown in each figure reflects the scale of the ensemble being analyzed, not a different assumed response time of the catchment.
In Fig. 4a, the RRDs are computed for subsets of the data conditioned on individual years and predictor-variable bins, which substantially reduces the number of events contributing to each ensemble. To avoid overfitting and unstable estimates at long lags, we therefore restrict the displayed lag window to the portion of the response that is robustly resolved by the data.
The location of the peak runoff response in Figs. 2a, 3a, and 4a is approximately 10–30 hours, less than two days. Explaining the variability within this range is beyond the scope of this manuscript and is worth investigating in future research. However, the smaller lag window in Fig. 4a likely reflects reduced statistical support at longer lags for these more finely stratified ensembles, rather than a physical difference in catchment response time.
L 354-355: aggregating hourly Upper Kuparuk time series to 4-hour intervals
Reviewer: Why change the intervals from 10 h bins to 4 h bins?In Fig. 4, we aggregate the hourly Upper Kuparuk time series to 4-hour intervals to better resolve differences in peak runoff response across years and predictor-variable bins. Using a shorter aggregation interval improves temporal resolution near the response peak and increases the number of independent observations contributing to each ensemble, which reduces uncertainty and narrows error bars without altering the inferred peak lag or total runoff response.
Importantly, sensitivity tests show that the qualitative behavior of the runoff response and the inferred peak timing are consistent across aggregation intervals; the change in bin width affects only the smoothness and uncertainty of the curves, not the underlying response structure.
L365-366 (and Fig. S3) In Supp. Fig. 3, we compare each explanatory variable individually with the peak runoff response to rainfall
Reviewer: In Fig. S3 in the Supplement it would be helpful to include the p values for each regression in addition to r values and to exclude a (red) line for those correlations that are not significant.We agree this will be helpful for the manuscript. We update the Supplementary Figure in the new version.
L372-373: The pair of predictor variables with the highest R² value in our multiple linear regression analyses is total summer precipitation and average temperatures during the preceding winter and spring.
Reviewer: More specifically do you mean the best pair of predicator variables ignoring the first-order control variables mentioned in the previous paragraph?Yes, we now write “Ignoring this first-order control, the pair of predictor variables with the highest R² value in our multiple linear regression analyses is total summer precipitation and average temperatures during the preceding winter and spring.”
L438-441: On the timescale of peak runoff response in Upper Kuparuk (~24 hours), coarse, silty soils with permeabilities around 10 −5 m/s can allow rainfall to infiltrate to depths on the order of tens of centimeters, comparable to the thickness of the active layer, enabling much of the precipitation to be stored in the subsurface
Reviewer: A citation should be provided for this permeability value. While the overlying organic soils in the Kuparuk basis likely have high permeability, the mineral soils are rich in loess and likely behave more like clay-rich soils. Thaw depths as great as cited in line 427 (up to 50 cm) would extend into this loess rich mineral soil. Differences in the hydraulic conductivities of these two materials are substantial (e.g. Hinzman, L. D., D. L. Kane, R. E. Gieck, and K. R. Everett (1991), Hydrologic and thermal properties of the active layer in the Alaskan Arctic, Cold Regions Science and Technology, 19(2), doi:10.1016/0165-232X(91)90001-W.We thank the reviewer for this clarification and for pointing out the important distinction between organic surface soils and underlying loess-rich mineral soils in the Upper Kuparuk basin, as well as for providing an appropriate reference (Hinzman et al., 1991). In response to this comment, and in light of related reviewer concerns regarding the speculative nature of this discussion, we have removed this paragraph from the revised manuscript. This revision avoids overinterpretation of site-specific permeability values and ensures that the discussion remains closely grounded in the observational constraints of the data.
We now write:
“In other permafrost catchments, thawed-layer storage capacity may play a more prominent role in regulating runoff response, depending on soil properties and flowpath structure. Numerous studies have documented that hydraulic conductivity in permafrost terrains varies by orders of magnitude both vertically within the active layer and laterally across hillslopes, reflecting differences in soil texture, ice content, organic horizons, and soil structure (Hinzman et al., 1991; Kane et al., 2001; Stieglitz et al., 2003; Neilson et al., 2018; O’Connor et al., 2020; Sjöberg et al. 2021; Hamm et al. 2021). These contrasts exert strong controls on infiltration capacity, subsurface storage, and the partitioning between rapid surface runoff and delayed subsurface flow. Vegetation is relevant in this context insofar as it co-varies with soil development and organic layer thickness, which influence near-surface hydraulic conductivity and water retention, rather than acting as an independent control on runoff response. As a result, the importance of thawed-layer storage in regulating runoff varies across permafrost landscapes and depends on the combined effects of soil texture, active layer structure, and anisotropy in hydraulic conductivity, rather than on thaw depth alone.In addition to vertical storage effects, runoff generation in permafrost catchments is strongly influenced by threshold-controlled lateral hydrologic connectivity. As soil moisture increases and the active layer deepens, previously disconnected portions of the near-surface soil profile can become hydraulically connected to downslope flow pathways and the channel network, producing nonlinear increases in runoff response (Quinton and Marsh, 1999; Quinton et al., 2000). This behavior is characteristic of permafrost environments, where low permeability at depth constrains flow to shallow horizons and promotes rapid lateral transmission once saturation thresholds are exceeded. In this context, antecedent effective wetness reflects not only the availability of subsurface storage, but also the degree to which laterally connected flow paths are activated at the time rainfall occurs. Such threshold behavior provides a physical basis for the strong sensitivity of peak runoff response to antecedent hydrologic state observed in Figure 2b–c, and may explain why antecedent streamflow exerts a larger control than gradual seasonal increases in thaw depth.”
L448-449: These localized pools, due to the high heat capacity of water, can enhance thaw and generate spatial feedback in both storage capacity and runoff routing.
Reviewer: Here and elsewhere, there are statements of fact that are likely to be true but are unsupported by relevant citations.We agree with the reviewer and also removed this section.
L473: significant
Reviewer: If you use the word "significant" please provide a p value based on a test. Otherwise say "important", "substantive" or similar.We thank the reviewer for this point. We now use the term “substantive”.
L486-488: We hypothesize that it is not the act of stabilization per se that reduces runoff, but rather the gradual thaw and drainage of the exposed ice-rich layer. As the active layer deepens and drains out pore water in the scar zone over time, increasing its capacity to store water.
Reviewer: Is there evidence that the active layer has deepened and dried? Might this also be due to the reestablishment of vegetation and organic layers in the stabilized ALD that begin to insulate the soil in a manner like the undisturbed Goose site?In the revised manuscript, we now explicitly acknowledge the lack of direct measurements and broaden the discussion to include alternative, non-exclusive mechanisms, including gradual surface recovery and organic layer reestablishment:
“One plausible explanation is that the hypothetical ice-rich layer at the sliding interface progressively thaws and drains, increasing subsurface storage capacity over time, reducing the efficiency with which rainfall is converted to runoff. Although vegetation cover at Ptarmigan remains minimal during the study period (Fig. 1c), even limited organic layer development may influence near-surface hydrologic behavior. Disentangling the relative roles of thaw depth evolution, subsurface drainage, and surface recovery would require targeted measurements of active layer thickness, soil moisture, and surface properties within scar zones, which are beyond the scope of the present dataset”.
L512: Figure 8.
Reviewer: Perhaps stack panels e and f under c and d in a 3x2 configuration rather than 2x3.We thank the reviewer for this suggestion and make this change in the newer version.
L558-559: This implies that rainfall mobilizes DOC and TDS flux at these catchments.
Reviewer: Then why is there no substantial correlation with rainfall described in the previous paragraph (L545-546)? Is there a contradiction here? Or perhaps simply a need to more clearly state that rainfall mobilizes (dislodges) DOC and sediment in place, but runoff is the mechanism responsible for lateral transport of these materials?We thank the reviewer for raising this point, which was also highlighted by another reviewer and prompted clarification in the revised manuscript. The apparent lack of correlation between precipitation and material fluxes does not contradict our conclusion that rainfall mobilizes DOC and TDS. Rather, it reflects the fact that simple correlation analyses are poorly suited to capture event-scale mobilization processes in systems characterized by dilution, hysteresis, and near-chemostatic behavior. We reorganize this section to highlight the novelty of using ERRA and how it enables a lag response even when there seems to be a lack of correlation between precipitation and material fluxes. We write:
“Although there does not appear to be a correlation between material fluxes and precipitation, there can still be a relationship that is not well captured in a simple correlation analysis. Rainfall influences material export in Arctic catchments through both mobilization and dilution processes, and the resulting concentration–discharge relationships can exhibit strong hysteresis or near-chemostatic behavior depending on solute availability and flow-path activation (e.g., McNamara et al., 1998; Frey et al., 2009; Shogren et al., 2019; 2021). Event-scale studies have shown that rainfall can generate short-lived but disproportionate contributions to suspended sediment, particulate organic carbon, and solute fluxes (Lamoureux and Lafrenière, 2014; Beel et al., 2018; Xiao et al., 2022) of shallow and near-stream flow paths (e.g., Bowden et al., 2008; Tank et al., 2016), while dissolved constituents such as DOC may remain elevated for days following rainfall (Koch et al., 2021). Collectively, this work demonstrates that rainfall exerts strong control on material export, but the timing and persistence of event-scale flux responses remain incompletely constrained. Here, we extend ERRA beyond runoff to quantify the event-scale material flux response to rainfall, allowing direct comparison of the timing, magnitude, and persistence of DOC, TDS, SS, and POC export following rainfall, independent of background seasonal trends or cumulative fluxes at Ptarmigan and Goose.
In this analysis, we substitute runoff in Eqs. (1)–(2) with material fluxes (concentration multiplied by runoff), while retaining precipitation (P) as the driver. Figure 8 illustrates how runoff fluxes of DOC, TDS, SS, and POC respond to rainfall at Ptarmigan and Goose. Figure 8 shows that both catchments exhibit notable DOC and TDS flux runoff responses that peak within a day of rainfall and decline over the subsequent two days. In Figs. 8e and f, we show the daily water flux runoff response distribution to rainfall, which produces a similar rapid response that declines over the course of two days. This implies that rainfall mobilizes DOC and TDS flux at these catchments. Together with the previous findings of Beel et al. (2020), our results support the conclusion that rainfall mobilizes DOC and TDS.”
L583-584: Unfortunately, the sparsity of the available data precludes a clear assessment of differences in runoff response between the periods when the ALDs were active versus stable at Ptarmigan
Reviewer: Given this statement I think any further discussion of what might have happened is entirely speculative. This speculation is not necessarily wrong, but it can't be substantiated and so I think it is best to delete this paragraph.We agree with the reviewer and further discussion on the differences between when ALDs were active versus stable at Ptarmigan is speculative. As suggested, we remove this paragraph.
L610-619 in Figure 9: Thawing of the subsurface can increase storage capacity and thus decrease runoff sensitivity to rainfall….etc.
Reviewer: Most of the rest of this lengthy legend repeats interpretation in the narrative of the manuscript and should be deleted here.We agree with the reviewer that this is redundant and so we have removed these sentences from the figure caption.
648-649: they are comparably influential Reviewer:
Replace with "their influence is similar"?We thank the reviewer and have made this change.
L650-651: Interannual variability in runoff response is also significant ( Fig. 4 ), underscoring the need for year-specific predictive modeling.
Reviewer: The intent here is not clear. Interannual variability in runoff response is likely. But if you know the response has changed, why do you need predictive modeling? Perhaps delete this reference to annual predictive modeling here as you take this up again, more cogently, in the final paragraph.We thank the reviewer and have made this change.
Citation: https://doi.org/10.5194/egusphere-2025-4275-AC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 333 | 119 | 26 | 478 | 54 | 23 | 28 |
- HTML: 333
- PDF: 119
- XML: 26
- Total: 478
- Supplement: 54
- BibTeX: 23
- EndNote: 28
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The paper entitled "Characterizing runoff response to rainfall in permafrost catchments and its implications for hydrological and biogeochemical fluxes in a warming climate" utilizes the ensemble rainfall-runoff analysis (ERRA) framework across three permafrost-underlain catchments to understand how rainfall driven runoff varies with biophysical and precipitation. The authors utilize data from the large and well-studied Upper Kuparuk River in Alaska and two very small watersheds in the Canadian high arctic; one of which is impacted by thermokarst.
ERRA is a new method and I will take the approach and framework 'as is' and will not present any criticisms to its methodology. Like a unit hydrograph, it provides a response function of catchments runoff (or other factors) to precipitation inputs, yet can handle nonlinear, non-stationary, autocorrelated and heterogenous systems as outlined an earlier paper that has been published in HESS. As noted, this manuscript seeks to use ERRA for three permafrost underlain watersheds, which are inherently non-linear and heterogeneous, to improve our process understanding and interrogate the influence of phenomenon like active layer detachments.
I read this manuscript several times with considerable interest. There are few people working in cold environments, and applications of new and novel methods are welcome. That said, there is considerable process understanding developed from decades of hard-won observation and careful analysis. As it stands, I do not believe that his paper is suitable for publication in its present form. I have a number of comments that I hope the authors will consider to improve this manuscript to make it more impactful to the permafrost hydrology and catchment science community.
First, and I hesitate to say this, the writing is colloquial and perhaps not up to publication standard. Suggesting improved writing is an easy throw-away comment, but I found it very causal with many interruptions. More detailed comments are below, and I strongly suggest that the authors separate the Results and Discussion to remove speculative comments and to identify where this work sits within the broader and more recent published literature. At times, I was jumping back and forth between what were new results, results that supported the literature (if referenced) and speculation.
Comments by line:
L51. Why ‘seemingly’. Is permafrost not impermeable? Or relatively so? If not, please provide references.
L55. Wang et al. 2021 is referenced twice.
L57. Wolvord & Kurylk is 2016.
L57. Is there a referencing order/format? It seems random here and in other places.
L60. It is likely Koch not Coch et al. 2018
L61. It is likely DeBeer et al. 2016 not De Boer et al. 2016. Neither are in the reference list.
L66. The authors in the previous line discuss rainfall, and here discuss how peak flow and flood response relate to permafrost thaw. Clearly, the expansive permafrost literature indicates that peak flows are most often (although obviously not always) associated with spring snowmelt freshet. Higher peak flows are most likely associated with increased snowfall and/or intense melt along with rain on snow, not summer rainfall. There is literature that discusses this.
L68. I am unsure as to why the ‘peak streamflow’ is discussed here as noted above. Peak streamflow in almost all permafrost rivers is associated with melt, not rainfall, except at potentially very large scales and in certain exceptional years.
L70. This will arise later, but the authors focus on thermokarst as a key process affecting flows. I would like to push back against this. While a startling and important process at local (and even scales up the ~100 km2), thermokarst is not a pan-arctic driver of changes in flows. I’m happy to be proven wrong. There is an observation bias simply because of its dramatic effect, yet it is confined to ice-rich regions, and there are many areas underlain with permafrost that are not influenced by thermokarst. Probably the vast majority of the arctic.
L72. I’m unsure as to the terminology here. ‘granular pore-scale storage capacity’. Is storage capacity not a clear enough term?
L74. Again, ‘pore-scale’ storage capacity? I’m curious as to what other storage capacity would be in play here?
L76. I agree that thermokarst affects runoff at local scales, but is there any supporting literature as it its ubiquity and influence at larger scales? I’m unsure if there is.
L80. There has been considerable ground-based observations in permafrost underlain regions. The Upper Kuparuk River and its sub-basins are some of the most well studied in the world, but the literature citations here are almost 30 years old and neglect much of the newer Alaska North Slope work that looks at runoff, chemical fluxes and seasonal responses. Simple searches, articles by Shorgren, etc., will reveal a wealth of process information not cited but pertinent.
L87. I’m unsure as to what the linkages between remote sensing and process-scale subsurface characteristics are. Is this easy anywhere? Is remote sensing useful in understand rainfall-runoff response in other environments and not permafrost ones? I’m not following this thread.
L91. There are many models that have worked to predict runoff form permafrost environments, many in the past five years that are not referenced. Papers by Rawlins and Painter are just two of many authors to consider.
L92. I agree that we do not fully understand runoff processes in permafrost watersheds, but the statement: “we still do not adequately understand how permeability, hydraulic potentials, and subsurface storage capacity will change with rainfall intensity and contributions from snow/ice melt” is somewhat peculiar. There is a wealth of literature showing how subsurface permeability, transmissivity and flow changes with seasonality and active layer development. As thaw progresses, the classical transmissivity feedback processes occur in permafrost catchments as porous near-surface layers rapidly convey water, yet as thaw progresses and water tables decline, runoff also does. There is almost 30 years of literate on this that is not referenced. While I understand that the Cape Bounty watersheds do not follow this conceptual model well, the Upper Kuparuk certainly does as it was in part developed there. The vertical aspect of water transmission and differences in depth-dependent permeability is important in interpreting the results yet is absent from the manuscript.
L97. I believe that this manuscript is confined to rainfall and should be clearly stated.
L97. This will arise later, but flows in permafrost catchments do not correspond to antecedent wetness – or at least antecedent storage. I believe the authors understand this and should reframe things. For example, in early June, when flows are high, antecedent water storage is very low, although streamflow (defined as wetness?) is very high. At this time, thaw depths are shallow, there is very little available storage for water. Flows are high and rainfall is rapidly transmitted to the stream (along with water released form the thawing active layer). As the seasons progresses, potential storage increases as the active layer expands. Flows can be equivalent in August and June, yet the ‘antecedent storage’ in August would be many times greater than June - with respect to liquid water anyway, but flows may be the same. Antecedent wetness can vary considerably from antecedent discharge, which is likely true in many places but amplified in permafrost environments. This is a clearly established conceptual model from permafrost catchments that is not well addressed in this manuscript.
L106. I’m wondering when the last time unit hydrographs appeared in HESS? A simple search shows only ‘reflective’ articles. While useful for teaching, the unit hydrograph (to my knowledge) is not widely applied in catchment science.
L111. Is “real-world” needed?
L123. “Relatively free” of glacial melt inputs? To my knowledge, there are no glaciers in either of these catchments. In addition, the vast majority of arctic watersheds at the headwater scale are free from glacial influence (and aufeis). While glaciers and incredibly important at larger continental and regional scales, and aufeis is a dynamic source of streamflow in catchments with extensive surface-groundwater interactions, the literature referencing here is somewhat confusing. They are not ubiquitous nor globally important across pan-arctic regions.
L130. Are thermo-erosional features common? They are dramatic and not rare, but ‘common’ is likely untrue. There is literature to cite as to their occurrence and terranes vulnerable to thermokarst that can be cited.
L131. Yokely et al. in review is not in the citations.
L139. The Kuparuk is 139 km2. The Cape Bounty Catchments are 0.18 and 0.21 km2. Nowhere is there a discussion of scale and its influence on the integration of results.
L144. How would a single tipping bucket rain gauge in a very large catchment influence the interpretation of the results? In summer, storms on the North Slope are convective and likely there are events that are missed/mismatched. Does this have any influence on the ERRA or its interpretation?
L145-149. I’m unsure as to the need for this sentence. Likely high frequency measurements are useful everywhere and the historical referencing is somewhat peculiar.
L150. Limited potential water sources? Most catchments receive snow and rain. Active layer thaw provides water certainly, but please consider comments above.
L156. Is it simple?
L158. This is somewhat repetitive from above and could be streamlined with a cleaner site description.
L161. Permafrost thaws, it does not melt.
L172. There is an indication of lower vegetation, although vegetation characteristics of the Upper Kuparuk are not given.
L179. Is there a need to speculate as to why one site had active layer detachments. Such a small slope difference may not be the reason as other simple topographic features like aspect and convergence can be invoked.
L237. The focus is on rainfall. Somewhere the reader needs to understand that this is a small fraction of the total water export from these watersheds. I doubt these are peak flows.
L246. There are 12 predictor variables to related RRD to environmental factors. I am uncertain as to how they were chosen, and there are some concerns I have. Number 1 and 3 appear to be the same. Number 9, 10 and 12 are highly correlated. Is number 11 well captured? I’m curious as to how previous summer rainfall intensity affects current year RRD. In the end, I believe that there should be some rationalizing of these 12 predictor variables. I can think of others from the literature that should be considered such previous season or fall total precipitation.
L252. Calculating total summer precipitation and rainfall as the average of the available data multiplied by the combined three summer months seems problematic, and is certainly an ‘extrapolation’ of reality. I’d like to see some analysis of how this affects analysis with complete data being downgraded and this method used to assess its influence on RRD. I understand that data provision is a challenge, but the influence of this on the interpretation seems to perhaps be large.
L256. Was freshet caught each year? I know this is a challenge.
L274. I do not believe that this analysis evaluates peak runoff. Rainfall driven runoff perhaps. I also do not know why this paragraph appears here.
L283. Separating the results and discussion would help identify the contributions of this manuscript with respect to the published literature.
Figure 2C. This information is a reproduction of 2B and likely not needed. Labelling the ‘potential point of saturation’ is also misleading as there is no empirical evidence for this.
L301. How does 57% compare with the published literature, even for this catchment or others form the North Slope.
Line 302-304. This is hard to contextualize and understand without literature to reference. These are very small numbers. I’m uncertain as to the meaning.
L305. Again, discharge in permafrost catchments is a weak reflection of catchment wetness. I’m not disputing the data, just the process interpretation. Maximum flows (even post freshet) are associated with low antecedent ‘wetness’ in terms of total mm of water available in active storage.
L310. There is considerable literature from the US, Canada, China, Russia and Scandinavia to reference here that is missing, most of which is more recent than a seminal 54 year old paper. The temperate catchment references in my opinion are not appropriate analogues to these catchments.
L315-323. The discussion of water tracks is important, but at the scale of the Upper Kuparuk, there are other issues such as beaded streams and channel routing that need to be considered. There are papers and literature on this. While water tracks are an important feature in this environment, there are other processes that become important as scale increases. In addition, there is no discussion here of depth-dependent hydraulic differences that are well studied and referenced in this environment.
Figure 3. The rainfall intensities are very low, but perhaps this is just how things are presented vis-à-vis a unit hydrograph type approach. It would be nice to understand how these intensities match up with the intensities of arctic storms, which are certainly increasing. Figure 3c suggests that as intensity increases, peak runoff increases along different ‘wetness’ (flow) classes. I’m wondering why the focus on peaks and intensity as opposed to total volumes? Perhaps I am caught a bit up in the weeds here but I’m struggling to understand the implications of this figure. Greater intensity (not total?) precipitation events generate greater NRF across lag times? Does total precipitation matter? Was it analyzed? Probably just a bit of clarity needed.
L345. I’m unsure as to the value of this speculative comment.
Figure 4 is interesting as there is considerable variability, but I am not sure that this is particularly clear to the reader or interpreted correctly. The 12 potential explanatory variables are used to provide a reason for this variability (with some factors having less data than others). Line 369 suggests that total summer precipitation (rainfall) is a fist order control on peak runoff response to rainfall. The variance simply could be because in one year all the rain was in the early season and in another it was all in the late season. The emphasis of interannual differences of event scale metrics can be highly influenced simply by timing. Was this considered?
Also, the discussion relating to snow versus rain is confusing things. Is there much rain below 2 degrees? If so, these values should be clearly presented and this seems more like a repetition of methods.
L373. The average temperature during the preceding winter and spring as a predictor as a key factor controlling RRA seems somewhat peculiar and looking at Figure 5c suggests a very weak relationship. An average winter temperature, in such cold environments, has very little impact on summer thaw and active layer development. There are many decades of work on heat transfer in permafrost soils that would suggest that summer temperature and thew rates far exceed any antecedent winter thermal conditions. The thought that a cold or a warm winter in such extremely cold environments overrides the ‘in season’ climate needs some considerations. There are other factors like previous season or late season precipitation that process studies have used to explain current season responses that make more process sense. This goes to the selection of the 12 factors - which could lean a bit more on the process literature.
L404. Yes, this is well established. Work from across Alaska in different permafrost environments from Caribou-Poker Creek to Toolik Lake based catchments show this. More appropriate referencing to this response that is related to active layer thaw, storage and vertical profiles of conductivity would be welcome.
L412. This is not peak runoff which would likely be during freshet.
L420. The decline has been noted in the literature and is related to active layer development and flow pathways.
L423. There is more recent literature on evapotranspiration in permafrost watersheds and its coupling to runoff response and seasonality. Certainly, they are decoupled, but the process of runoff generation and ET are well documented.
L430. The observation that runoff decreases with season in permafrost catchments is one of the most common observations and should be cited with references more recent than 1998 for Upper Kuparuk.
L434. I disagree with this statement. Antecedent wetness is not a process and its interpretation does not square with process understanding from permafrost watershed research.
L437. I am unsure how vegetation enters the discussion of hydraulic conductivity, and the following discussion of infiltration with respect to soil texture is simply speculative in the face of again, considerable literature. The hypothetical paragraph staring line 444 does not advance our underrating of process.
Another process absent from discussion is one of lateral connectivity, which can be facilitated in permafrost underlain catchments as wetness increases. It is an environment where thresholds dominate. Consider the work of Quinton from the late 90s.
Section 4.5 seems out of place and an add-on to the manuscript. The inclusion of material transport here at one site is a distraction this late in the manuscript. I think a more focussed analysis on rainfall responses across permafrost watersheds would have provided the potential for process insights than data from a very small single site.
Literature on active layer detachments and their influence on runoff and water quality exist for Alaska and the Northwest Territories that should be cited. Undoubtedly ALD will influence runoff response, although I’m struggling to understand how ERRA sheds new light here on what is going on.
L491. There is considerable literature on the geomechanics of active layer detachments. ‘lubricating subsurface layers’ is not particularly technical.
L502. The statement discussing ALD on flash floods and damage critical infrastructure does not make much sense considering the scale of these ALD. More likely, an ALD will simply block a road or destabilize a foundation as mentioned in the cited literature. I’m not sure the alarmism is well founded and the final comment on L506-508 is beyond what can be stated by the analysis here.
L520. Yes, rainfall mobilizes solutes and sediments, but it can also dilute them. The picking and choosing of literature here is peculiar considering an extremely large body of work on water quality and flows at catchments (including the Upper Kuparuk) and the development of conceptual models that are not referenced. Again, see more recent work by Shogren.
L546. While it is acknowledged that material flux is positively associated with discharge (because discharge is in the flow calculation) how can you suggest that this is true as there is no association between discharge and the concentrations or values. I’m uncertain as to the reasoning here.
L560. If you state that the data suggests that rain mobilizes DOC and TDS, how does this square with the line 456 that states that there is no association between rain and the material flux?
In the end, I’m unsure as to what this analysis adds to the already good work of Beel documenting this process at this site. The authors need to emphasize this and also cite work of Bowden, Tank and their team (just two scientists that come to mind).
Figure 9 is similar to other conceptual models in the literature over the past 30 years defining active layer expansion and water/material fluxes. Recent ones by Grewal come to mind and older ones by Quinton from an environment similar to the Upper Kuparuk in Canada. For 9c, d and e, the stream seems far above the slope/riparian zone and is a bit confusing. I’d suggest permafrost not be blue but perhaps grey or a color not immediately associated with water.
L628. The year to year variation in peak runoff being negatively correlated with average temperatures during pervious winter is a weak (panel 5c) and in-season conditions are a much greater proxy on how rapid thaw would be. The degree day models while empirical are pretty good at predicting thaw so I’m a bit surprised. I suggest that the timing of precipitation within the year has much to do with this variability (early v. late). Winter temperatures have warmed much more dramatically across circumpolar regions over the last several decades, yet no appreciable influence on rainfall-runoff processes have been observed form this warming so I struggle to understand causal mechanisms that warm winters facilitate more rapid thaw.
SM Figure 6. It is extremely hard to interpret either the precipitation or temperature.
SM Figure 7. As noted, there is mixed responses between discharge and material/chemical fluxes that confound some of the explanations in the manuscript. DOC is largely chemostatic, perhaps. TDS as well with a bit of dilution, SS flux and POC mobilize a bit at Ptarmigan, but relations are relatively weak.