|The Authors have addressed all comments from the previous review round and adopted some of the suggestions. I kindly ask them to address the comments below, most of which are based on the replies to the previous review round:|
1) Thank you for addressing this comment.
2) The authors have included in their reply of this comment the difference between forecast and projection. However, both weather-forecast and climate-projections should be accompanied by a probability of occurrence, since they both include expectations (or in other words, a statistical estimation of the expected/mean value of an envelope of events). In high-complex systems (such as climate dynamics), this can be achieved by a sensitivity analysis of the input parameters in order to create the envelope of different possible projections to see “what might happen if the world warms by +4-5 oC”, as for example, for the RCP8.5 scenario that the authors use (e.g., see, for example, recent discussions in Schwalm et al., 2020; Hausfather and Peters, 2020).
Therefore, a sensitivity analysis of the input parameters is required to create such an envelope of events. For example, the comparison between the observed and simulated river stages, groundwater levels and snow cover at the 3 stations indicated in Figures R1a-d, illustrate that (with the exception of snow cover) there is a large difference between them. However, if the Authors perform a sensitivity analysis on the several input parameters of their model, they would see that the cloud of simulations is expected to cover the observed records and to also assign a probability of occurrence to the end of century expectations.
Moreover, the high-dependence of the simulations on the initial conditions is a known issue in atmospheric dynamics and thermodynamics. One of the first studies (there are many since then) is by Lorenz (1963), who discovered the existence of the phenomenon of chaos in the atmospheric dynamics and the high-dependence on the initial conditions, which increases with time. Please note that only the statistical estimations, as for example the expectation/mean of an envelope of events, is expected to be independent to initial conditions, but this requires a sensitivity analysis and not the application of a deterministic model with fixed values of the input parameters and initial conditions.
I understand it is difficult for the authors to perform such a huge sensitivity analysis at this stage, and so, this is why I kindly ask to at least discuss the issue of uncertainty and variability on the input parameters and initial conditions, and thus, on the simulated output.
3) The so-called Long-Range Dependence (LRD; or else known as long-term persistence or the Hurst phenomenon; Hurst, 1951) was the first study that gave a justification of why the so-called clustering of events (i.e., the tendency of wet/dry years to occur together in a non-predictive manner forming clusters) may occur in natural processes. Earlier, and independently, Kolmogorov (1940) introduced the so-called fractional-Gaussian-noise (fGn) model that was able to simulate this clustering behaviour. Since then, there are many developments towards the simulation of the LRD in natural processes and atmospheric dynamics (the literature is huge; see, for example, studies by Mandelbrot Wallis, 1968; Klemes, 1974; Tsonis, 1999; Dimitriadis et al., 2021, etc., which all include many references on this issue and identify the LRD in key hydrological-cycle processes - like streamflow, precipitation, air temperature, specific humidity, atmospheric pressure, wind speed, solar radiation, evapotranspiration, etc.- from analysis of thousands of stations and billions of records). The impact of the LRD behaviour on the atmospheric dynamics extends to the climatic scale (over 30-years), and so, it is not enough to show a validation of a 5 years with the model but rather to validate the model in an over-30-years window. I understand that it may be difficult to perform such analyses at this stage, however, I kindly ask the authors to at least discuss the impact and need for preservation of the LRD behaviour in the atmospheric dynamics model. This is in line with the previous comment, since the LRD is a stochastic attribute that arises from the intrinsic uncertainty of the chaotic behaviour apparent in atmospheric dynamics and thermodynamics models.
4) Thank you for addressing this comment. From the provided Tables, it can be observed that there are in total 29 input parameters (8 for the hydrodynamic properties based on the geology, 12 for the surface roughness and crop properties based on land use, and 9 for the numerical model set-up) and 5 output variables.
5) Please consider further discussing the effect of groundwater and present both scientific opinions. It is mentioned by the Authors that the simulations without pumping do not significantly change the observed dynamics of the system; however, the overexploitation of groundwater is considered an important anthropogenic impact at the local and global climate and the hydrological cycle with a visible effect on sea level rise (see, for example, a recent analysis by Koutsoyiannis, 2020).
Dimitriadis, P., D. Koutsoyiannis, T. Iliopoulou, and P. Papanicolaou, A global-scale investigation of stochastic similarities in marginal distribution and dependence structure of key hydrological-cycle processes, Hydrology, 8 (2), 59, doi:10.3390/hydrology8020059, 2021.
Hausfather, Z., and G.P. Peters, RCP8.5 is a problematic scenario for near-term emissions, Proc. Natl. Acad. Sci. USA, 117, 27791–27792, 2020.
Hurst, Long-Term Storage Capacity of Reservoirs, Trans. Am. Soc. Civ. Eng., 116, 770–799, 1951.
Klemes, V., The Hurst phenomenon: A puzzle?, Water Resour. Res., 10 (4) 675-688, 1974.
Kolmogorov, A.N., Wiener spirals and some other interesting curves in a Hilbert space, Dokl. Akad. Nauk SSSR, 26, 115–118, 1940.
Koutsoyiannis, D., Revisiting the global hydrological cycle: is it intensifying?, Hydrology and Earth System Sciences, 24, 3899–3932, doi:10.5194/hess-24-3899-2020, 2020.
Lorenz, E.N., Deterministic nonperiodic flow. Journal of the Atmospheric Sciences, 20 (2), 130–141, 1963.
Mandelbrot, B.B. and J.R. Wallis, Noah, Joseph and operational hydrology, Water Resour. Res., 4, 909–918, 1968.
Schwalm, C.R., S. Glendon, and P.B. Duffy, RCP8.5 tracks cumulative CO 2 emissions, Proc. Natl. Acad. Sci. U.S.A., 117, 19656–19657, 2020.
Tsonis A.A., P.J. Roebber, and J.B. Elsner, Long-range correlations in the extratropical atmospheric circulation: origins and implications, J. Clim, 12, 1534–41, 1999.
The manuscript simulates End of Century (EOC) extremes and their effects on the water-energy balance in the Cosumnes river basin, using cutting-edge global climate and integrated hydrologic models (ParFlow-CLM). I really like the way the authors used to analyze the hydroclimatic changes by median WY, dry WY and wet WY (e.g., Figures 3-5). The manuscript is overall clearly written, and the results are well discussed.
My first concern is the insufficient validation of the models’ simulations in the historical period. Besides temperature and precipitation outputs, other watershed-integrated fluxes, and storages (e.g., ET, soil moisture, TWS and streamflow) should also be validated as much as possible using the observations, remote sensing data and reanalysis, to ensure the models’ simulations reasonable. Only then will we believe the further analysis between future and historical periods is valid. In my opinion, the historical simulation of VR-CESM is not so good because the simulated dry, median, and wet water years are distinct from the PRISM (Figure A2).
The authors may argue the historical simulations are acceptable, because a global climate and integrated hydrologic models are used (more complex and larger simulation domain). However, one can use a finer-resolution hydrological model (e.g., VIC, SWAT, and many others) driven by statistically or dynamically downscaled regional climate model outputs to obtain more reasonable (maybe more accurate from the perspective of validation) simulations in this river basin (7000 km2), and to do further analysis like the authors did in this study. Please explain why the global climate and integrated hydrologic models are more suitable for this case study?