Response of catchment water storage capacity to the 1 prolonged meteorological drought and asymptotic 2 climate variation 3

Abstract. Studies on the hydrological response to continuous extreme and asymptotic climate change can improve our ability to cope the intensified water-related problems. Most of the existing literature focused on the runoff response to different climate change patterns, while neglected the impacts by the potential variation in the catchment water storage capacity (CWSC) that plays an important role in the transfer of climate input to the catchment runoff. This study aims to identify the response of the CWSC to the long-term meteorological drought and asymptotic climate change systematically. Firstly, the time-varying parameter is derived to reflect the CWSC periodic/abrupt variations under both drought and non-drought periods. Secondly, the change points and varying patterns of the CWSC are analysed based on the Bayesian change point analysis with multiple evaluation criteria. Finally, multiple catchment properties and climate characteristics are used to explore the possible relationship between these variables and the temporal variation characteristic of the CWSC. The catchments suffered from prolonged meteorological drought in southeast Australia are selected as the case study. Results indicate that: (1) the increase of CWSC amplitude change has been observed in 83/92 catchments during the prolonged drought period and the significant shifts in the mean value of the CWSC are detected in 77/92 catchments; (2) the median response time of CWSC for all 92 catchments with significant changes is 641.3 days; (3) the values of CWSC are changed significantly in the catchments with small area\\low elevation\\small slope range\\large forest coverage and high soil water holding capacity. This study might enhance our understanding to the variations in catchment property under different climate-changing patterns.


four parameters, and its structure is shown in Fig.3. The meanings of the four model 187 parameters are as follows: θ1 is the capacity of runoff producing reservoir in the 188 catchment (mm); θ2 is the groundwater exchange coefficient (mm); θ3 is the capacity 189 of catchment reservoir (mm); θ4 is the unit line confluence time (day).
After the change-point: 220 where, where p means the probability of likelihood.  Table 2. 242

250
To evaluate whether the CWSC has changed significantly under climate change, the 251 following three criteria are adopted. 252 (1) The requirement of NSE 253 In order to guarantee the reasonable simulation results of the GR4J model, the 254 Nash-Sutcliffe efficiency (NSE) coefficient values before and after the change point 255 should be greater than 0.6. Furthermore, the difference of NSE values between the two 256 periods should be less than 20%  . 257 (2) The minimum requirements for significant changes in storage capacity 258 The change rate of the estimated parameter θ1 (θ'1) before and after the change 259 point should exceed 20%  . i.e., 1 1

260
(3) Robustness requirements of the results 261 The initial values of the model parameters will be changed three times. Only when 262 the results of the three calculations all show that the CWSC has changed significantly, 263 the catchments would be selected.

280
The process that leads to the change of the CWSC cannot be measured directly, so some 281 measurable factors are used to probe their lurking correlation between the change of 282 CWSC and the catchment response time. We select 33 potential factors of catchment 283 and list in Table 3, which including 9 catchment features and 24 local climate variables. 284 It is noted that because of the limitation of available data for catchment characteristics,  interval of forest coverage is 15%-92%. These catchment features are selected as 308 potential impact factors and analysed further in Section 4.3. 309

310
The most likely change point is confirmed when those three criteria had been satisfied. 311 The change pattern of the CWSC is determined by Equation (1) and (2). In other words, 312 Equation (1)/Equation (2) reflects the potential periodic/asymptotic feature during the 313 period before/after the change point. It is obvious that 1 ( 2 ) and 1 ( 2 ) are the most 314 important parameters in the regression function, which refer to the amplitude and 315 intercept of the time-varying parameter 1 , respectively. Furthermore, the variation 316 between 1 and 2 denoted the average difference between 1 and 1 ′ , reflecting the 317 potential change between the CWSC of periods before and after the change point. 318 Table 5  The regression parameter , refers to the intercept/mean value of the CWSC 338 during the specific period, is used to evaluate the average difference of the CWSC 339 during two periods. Significantly upward change in the mean value has been 340 identified in 84% of catchments (77 in 145 catchments) during the drought period, no 341 catchment has been found with the significant downward change of in the drought 342 period. In addition, the number of catchments with non-significant change in is 15; 343 6.9% of catchments (10 in 145 catchments) and 3.5% of catchments (5 in 145 344 catchments) have been identified with non-significant upward and downward trend 345 during the drought period, respectively. These results illustrated that most of catchments 346 (77 in 92 catchments) experienced a significantly upward trend in the average CWSC 347 during the transformation from non-drought period to the prolonged drought period, 348 indicating the increased CWSC during the latter period. 349 The spatial distribution of the set of 92 catchments that satisfied the criteria of 350 https://doi.org/10.5194/hess-2021- 646 Preprint. Discussion started: 26 January 2022 c Author(s) 2022. CC BY 4.0 License. NSE performance and results robustness is presented in Fig. 5. As shown in Fig. 5(a), 351 94.5% catchments (87 in 92 catchments) were found with the significantly upward 352 change in the amplitude during the drought period. Similarly, As presented in  negatively correlated with all other catchment features (see Fig.10(a)). Furthermore, no 449 significant correlation has been found between the absolute change of amplitude ( ) 450 and all catchment features. As presented in Fig.10(b), positive association has been 451 found between the relative change of amplitude ( ) and the AWHC of the topsoil, between the relative change of amplitude ( ) and all climate variables (Fig.10(d)). 460

Factors for shifts in the CWSC
Since no strong correlation between the amplitude ( ) and a single factor is found, 461 therefore we speculate that the potential change of the variation range of the CWSC is 462 the result of the combination of various catchment features and climate factors. with both catchment features (see Fig. 12(a)). For instance, the most related estimate 481 of CC was acquired by the absolute variation in and the Ks of topsoil with CC 482 estimate of -0.362, followed by the AWHC of the subsoil (CC=-0.341), Ks of subsoil 483 (CC=-0.267), Forest percentage (CC=-0.242), subsequently. As illustrated in Fig. 12(b), 484 the relative change of mean value (  ) of 1  is negatively correlated with all catchment 485 features (except for A3 (Slope range), and A6 (AWHC of topsoil)), but both correlations 486 are weak. In general, soil and forest percentage are the most related variables to the 487 mean value (  ). The water holding capacity of various soil types is different as the 488 dissimilarity of void and adhesion in different soil types, which directly affects the 489 ability of the catchment to absorb and store water, and then affects the CWSC of the 490 catchment. Furthermore, the coverage of multiple forest percentage would affect the 491 water holding capacity and water assumption ability, resulting the potential changes in 492 the CWSC. Fig.12(c) and (d)  precipitation, CC=-0.245) (Fig.12 (c)). The correlation between the relative change 499 values of these two is shown in Fig.12 (d). Only the correlation between relative change 500 of mean value (  ) and B20 (Mean annual runoff index, CC=-0.215), B24 (Annual base 501 flow ratio, CC=-0.279) are significantly negative, respectively. 502

503
The Pearson correlation coefficient between the response time with catchment features 504 and climate variables are presented in Fig.13, which indicate that: a strong positive 505 correlation was identified between the response time with A2 (mean elevation, 506 CC=0.239), and A6 (AWHC of the topsoil, CC=0.249). While a strong negative 507 correlation was found between response time with A5 (forest coverage, CC=-0.225). 508 The potential reasons for this finding lie that the increased forest coverage of the 509 catchment resulted in the larger water demand of the ecosystem, and thus a shorter 510 response time of the CWSC to the meteorological drought. In other words, when a 511 catchment has experienced a prolonged meteorological drought, it would respond fast 512 due to its large water demand. As for the climate variables, the absolute variations of 513 most climate variables had negative correlations with response time in Fig.13(b).  Fig.13(c). 518