Recent decrease in summer precipitation over the Iberian Peninsula 1 closely links to reduction of local moisture recycling 2

17 The inherently dry summer climate of the Iberian Peninsula (IP) is undergoing drought 18 exacerbated by more intense warming and reduced precipitation. Although many 19 studies have studied changes in summer climate factors, it is still unclear how the 20 changes in moisture contribution from the sources lead to the decrease in summer 21 precipitation. This study investigates the differences in the IP precipitationshed between 22 1980-1997 and 1998-2019 using the Water Accounting Model-2layers with ERA5 data, 23 and assesses the role of local moisture recycling and external moisture in reducing 24 summer precipitation. Our findings indicate that the moisture contributions from the 25 local IP, and from the west and the east of the precipitationshed contributed 1.7, 3.6 and 26 1.1 mm mon -1 less precipitation after 1997 than before 1997, accounting for 26 %, 57 % 27 and 17 % of the main source supply reduction, respectively. The significant downward 28 trend of the IP local moisture recycling closely links to the disappearance of the wet 29 years after 1997 as well as the decrease of local contribution in the dry years. Moreover, 30 the feedback between the weakened local moisture recycling and the drier land surface 31 can exacerbate the local moisture scarcity and summer drought.

1980-1997 and 1998-2019 using the Water Accounting Model-2layers with ERA5 data, 23 and assesses the role of local moisture recycling and external moisture in reducing 24 summer precipitation. Our findings indicate that the moisture contributions from the 25 local IP, and from the west and the east of the precipitationshed contributed 1.7, 3.6 and 26 1.1 mm mon -1 less precipitation after 1997 than before 1997, accounting for 26 %, 57 % 27 and 17 % of the main source supply reduction, respectively. The significant downward 28 trend of the IP local moisture recycling closely links to the disappearance of the wet 29 years after 1997 as well as the decrease of local contribution in the dry years. Moreover, 30 the feedback between the weakened local moisture recycling and the drier land surface 31 can exacerbate the local moisture scarcity and summer drought. 32 33 1. Introduction 34 The Iberian Peninsula (IP) is located in the Mediterranean basin, which is among 35 Based on the assumption of a well-mixed atmosphere (Burde, 2010; Goessling and 149 Reick, 2013), the moisture contribution, that is, the tagged evaporation , can be 150 calculated considering that the ratio of tagged to total atmospheric water storage is equal 151 to the ratio of tagged to total evaporation, as shown in Eq. (2). Considering the proposed 152 retention time of atmospheric moisture is about 1 week to 10 days (Numaguti, 1999), 153 we set the backtracking time as 1 month longer for summer precipitation to make sure 154 that more than 90 % of the precipitation can be redistributed to the surface (Zhang et 155 al., 2017). (2) 157 The main moisture source suppling IP summer precipitation is defined as the 90 th 158 percentile precipitationshed in this study. It is further divided into subregions to 159 evaluate the role of the contribution from each area, such as the local moisture recycling, 160 which demonstrates the contribution of local evaporation to the IP precipitation, and 161 the external advection moisture, which describes the non-local evaporation contribution 162 to the IP precipitation. The contribution ratio ( ) of a subregion ( ) is defined as the 163 proportion of the moisture contribution from it to the total contribution from the main 164 source region ( ) , which is calculated as the following Eq. (3). The precipitation 165 recycling ratio of the IP can be substituted with the IP local contribution ratio . 166

Significance test 168
The slope significance of trend fitting and the significance of the difference in the 169 means are tested using Student t-test in this study. Additionally, the sliding t-test, as a 170 method of mutation analysis, is used to detect whether and when the sample mean in 171 the IP precipitation series changed significantly, 172 ( 1 −1) 1 2 +( 1 −1) 2 2 where is the precipitation series to be tested, 1 and 2 are step lengths set for two 174 sequences before and after the moving point, and 1 2 and 2 2 are the variances of the two 175 sequences which can be calculated as following. 176 with the observed data on both annual and seasonal scales, together with all correlation 183 coefficients higher than 0.95. The average annual precipitation over the IP is about 184 55.66 mm mon -1 from ERA5 and 58.07 mm mon -1 from Iberia01, respectively. 185 Compared with the observed data, the reanalysis data slightly underestimates the IP 186 precipitation with the root mean square error (RMSE) of 3.46 mm mon -1 on the annual 187 scale. The comparison of seasonal precipitation shows that the ERA5 is lower than the 188 observed Iberia01 value in the rainy seasons (both winter and autumn), but higher in 189 the dry summer. The RMSE between the two datasets of seasonal precipitation is in the 190 range of 4.30-12.65 mm mon -1 . Since the Iberia01 data is the grid data interpolated from 191 observation site data (Herrera et al., 2019), some of the deviations between the ERA5 192 and the Iberia01 precipitation can be partially affected by the interpolation process 193 rather than solely the result of the error generated by the reanalysis process. In general, 194 the ERA5 precipitation data shows the characteristics of IP precipitation reasonably 195 well and thus is suitable for studying the changes.

219
For summer precipitation, the dry years (1991,1994,2005,2012 and 2016) and 220 the wet years (1983 1987 1988 1992 and 1997) are selected, which are circled in Fig.  221 2(c). A wet year is defined as the year in which the precipitation is more than one 222 standard deviation above the average precipitation, and similarly, the precipitation in a 223 dry year is lower than a standard deviation range. Accordingly, the division of time 224 period also applies to the precipitation series of the dry and wet years. It is specifically 225 observed that the dry years are distributed in both two periods, with the average 226 precipitation of 17.15 and 18.34 mm mon -1 before and after 1997, whereas the wet years 227 occur before 1997 with an average of 51.03 mm mon -1 but disappear after 1997. percentile precipitationsheds shown in orange and green color in Fig. 5(a, b). This 287 westward extension implies that the significant and substantial reduction in the 288 contribution of the local grids and its surrounding grids results in a decrease in the 289 proportion of these areas. Therefore, for the same percentile of the precipitationshed, 290 only a smaller area concentrated by high-contribution grids is sufficient before 1997. 291 However, a larger area is required for the same proportion after 1997. 292

357
The dry years in the two periods have been divided and compared with each other, 358 and the differences between the two periods are shown in Fig. 8. From the distribution 359 of differences, the grids with reduced moisture contribution are mainly located in the 360 IP and the east region, and the southern part of the IP has the largest decrease ( Fig. 8(a)). 361 Mainly dominated by these negatively changing grids, both the absolute contribution 362 and the contribution ratio of the local IP and the east have dropped significantly, with 363 0.53 and 0.42 mm mon -1 decrease in contributed precipitation and 3.58 % and 2.81 % 364 contribution ratio reduction, respectively ( Fig. 8(b, c)). For the west region, however, 365 it raises the moisture contribution to the summer precipitation by 1.22 mm mon -1 in the 366 dry years after 1997, causing a 6.39 % increase in its contribution ratio. Despite the dry 367 years with no decrease precipitation between the two periods, the decrease in the local 368 moisture recycling is still noticeable.

397
The remarkable decrement of summer precipitation can be attributed to the 398 simultaneous and large reduction of contributions from all three source regions. 399 Moisture transport from the west region contributing to the IP precipitation is mainly 400 through the tropical-subtropical North Atlantic corridor. In summer, air masses from 401 the west in this transportation process, as it gets closer to the destination, gradually shift 402 from a net supply to a net uptake of the IP precipitation (Gimeno et al., 2010). In terms of the total contribution from the three subregions, the west region 416 dominates more of the reduction in the IP precipitation due to its wide coverage with a 417 large number of grids. Nevertheless, in the local IP, which is much smaller than the 418 west, the high contribution per grid, the difference between the two periods, and the 419 consistent decline of the precipitation recycling ratio make the role of the local 420 contribution variation worth emphasizing. As an important indicator to describe the 421 interaction between the surface and atmospheric processes, the change in the 422 precipitation recycling ratio takes into account changes in both precipitation and the 423 contribution of local evaporation (Goessling and Reick, 2011;Zhang, 2020). For the IP, 424 its significant reduction in local moisture contribution is most likely due to the 425 weakening of local evaporation (Fig. 9) 1) The reduction of contribution to IP summer precipitation is mainly concentrated in 452 the IP and its neighboring grids. The local IP grids show the greatest reduction, and 453 the surrounding grids show a slight but significant decrease. 454 2) Compared with the summer of 1980-1997, the IP and the east contributed 1.7 and 455 1.1 mm mon -1 less IP precipitation during 1998-2019, accounting for 26% and 17% 456 of the main source supply reduction, respectively. Meanwhile, the importance of 457 the vast west region was clearly shown by reducing the IP precipitation by 3.6 mm 458 mon -1 , representing 57% of the decrease in precipitation originating from main 459 sources. 460 3) The contributions from the local IP and the east keep declining during the 40 years.