Interactive comment on “ Seasonality of hydrological model spin-up time : a case study using the Xinanjiang model ”

Authors estimated spin-up time of a conceptual hydrologic model by initializing the model from two extreme initial conditions (wet and dry states) at different times of a year, and defining the equilibrium state by computing the Mahalanobis distance between the soil moisture states of two model runs. By initializing the model at different times of a year, authors were able to present variability of spin-up time from models started from different seasons. While model initialization is an important issue in hydrologic modelling and erroneous initial conditions lead to bias on simulated model output(s), there are a few issues with the current study as highlighted below.

Author response: Considering the differences in model definition and structure, it is quite risky to consider the same control for all the models.Available spin-up literatures are mainly model specific.Lumped and conceptual models are less expensive to run for sure.However, spin-up time could be of a concern for data scarce situation or seasonal simulation (Rahman and Lu, 2015).Common practices consider first 2 or 3 years of simulation as the spin-up period and removes from the rest of the analysis including model calibration.The problem here is to define the spin-up period.This is usually done based on personal feeling, experience, available data records and purpose.In some cases exclusion of one year model outputs could be a very costly task in developing countries where hydro-climatic data is very scarce (say only 2-3 years of available data records).Over-estimating the spinup period would lead to a loss of important information.Likewise, an underestimation would affect the conclusion by incorporating erroneous initial model outputs.
On the other hand, researchers often consider various spin-up time even for the same model.Lin et al (2016) considered a spin-up period of 19 days for the Xianjiang (XAJ) model during a four-month streamflow simulation for the Shiguanhe River Basin, China.In another study, Lu et al. (2008) considered only 12 h of spin-up time while forecasting floods at the Huaihe River Basin's Wangjiaba sub-basin.Even if the spin-up times are dissimilar for conceptual and physical models, this study serves important information for the XAJ model as well as other modeling communities.Firstly, it provides a basis for estimating the spin-up time for the XAJ model using widely available data sets.Secondly, it establishes a conceptual basis and shows the variations of spin-up time based on the simulation start time that provides new insights even for the physical models (the controls might be different).Thirdly, it ascertains new approaches to explore and define model spin-up time based on broadly acceptable Mahalanobis Distance that over comes the limitations of available spin-up detection techniques.
3) Following from (1) and ( 2), there are systems that are known to require longer spinup times, such as deep groundwater aquifers, large surface storages, etc.Is the study model configuration capturing any of these slower processes, or purely focused on shallow soil water storage?How appropriate is this model configuration for the study basins' dominant hydrologic regime?How generally applicable are the findings if these processes are not represented?I recommend the authors provide a bit more detail on the model and its appropriateness for the study basins.

Author response:
The studied model mainly focuses on hydrologically active soil water storage zones.This model is extensively used in humid and semi-arid regions of China and other parts of the world.The runoff formation in the XAJ model is based on the repletion of storage concept, the runoff will start to generate once the soil moisture content of the unsaturated zone reaches its field capacity, and subsequently runoff equals the rainfall excess without further loss (Zhao, 1992).The model accepts areal mean precipitation and pan evapotranspiration as the inputs and produce streamflow from the whole basin.The applicability of the XAJ model of the study basins' have been tested by Kyi, 2014.At this point this study outcome might be true for humid and semi-humid basins of USA.Commenting on the appropriateness of this conclusion at outside USA requires further verification.
4) How does the model calibration affect the results?The calibration procedure is a bit difficult to understand based on the description provided, so I recommend going into a bit more detail on what was done and how it may/may not be impacting spinup times.
Author response: The calibration process does not affect the result.The model calibration and exploration of model spin-up time were performed separately (please see page-5, line-11-14).The XAJ model was firstly calibrated with saturated initial condition and thereafter the daily streamflow was validated against those of the observed by taking spin-up time long enough (10 year) to avoid the effects of the initial condition.Thereafter, these calibrated parameter values were exercised for the subsequent simulations that explores the impact of initial condition.
The model calibration section has been updated with more clarification.
5) The authors present some interesting patterns in seasonality of spinup time.However, there is little discussion of the potential physical reasons for the patterns.Are these patterns simply mirroring seasonal patterns of precipitation?Are there other physical or climatic controls that might explain some of the spread, or help us determine how we can apply these results to other models or domains?I recommend expanding the discussion section to address some of these questions, which should give the paper a much broader relevance.