HCPD-CA High-resolution climate projection dataset in Central 1 Asia for ecological and hydrological applications

7 Central Asia (referred to as CA) is one of the climate change Hot-Spots due to the fragile 8 ecosystems, frequent natural hazards, strained water resources, and accelerated glacier 9 melting, which underscores the need of high-resolution climate projection datasets for 10 application to vulnerability, impacts, and adaption assessments in ecological and hydrological 11 systems in this region. In this study, a high-resolution (9km) climate projection dataset over 12 CA (the HCPD-CA dataset) is derived from dynamically downscaled results based on multiple 13 bias-corrected global climate models, and contains four geostatic variables and ten 14 meteorological elements that are widely used to drive ecological and hydrological models. 15 The reference and future periods are 1986-2005 and 2031-2050, respectively. The carbon 16 emission scenario is Representative Concentration Pathway (RCP) 4.5. The results evaluation 17 shows that the data product has good quality in describing the climatology of all the elements 18 in CA despite some systematic biases, which ensures the suitability of the dataset for future 19 research. The mMain features of projected climate changes in over CA in the near-term future 20 is are strong warming (annual mean temperature increasing by 1.62-2.02℃) and significant 21 increase in downward shortwave and longwave flux at surface, with minor changes in other 22 elements (e. g., precipitation, relative humidity at 2m, and wind speed at 10m). The HCPD- 23 CA dataset presented here serves as a scientific basis for assessing the potential impacts of 24 projected climate changes over CA on many sectors, especially on ecological and 25 hydrological systems. It has the DOI https://doi.org/10.11888/Meteoro.tpdc.271759 (Qiu, 26 2021)is publicly


29
Central Asia (referred to as CA, Fig. 1a) has complex terrain and diverse climates and is 30 among the most vulnerable regions to climate change due to fragile ecosystems (Zhang et al.,31 2016; Seddon et al., 2016;Gessner et al., 2013), frequent natural hazards (Thurman,32 2011;Burunciuc, 2020), strained water resources (Frenken, 2013), and accelerated glacier 33 melting (Narama et al., 2010;Sorg et al., 2012), which underscores the need to achieve high-34 resolution climate projection datasets for application to vulnerability, impacts, and adaption  Torma et al., 2015). Regional To obtain the accurate information on region-scale 41 climate change, dynamical downscaling has been developed and widely applied in regional 42 climate projections over many areas, like East Asia (Zou and Zhou, 2016;Tang et al., 43 2016; Jung et al., 2015;Jiang et al., 2021;Ji and Kang, 2013;Hong et al., 2017;Guo et al., 44 2021;Bao et al., 2015;Zou and Zhou, 2017), North America (Wang and Kotamarthi,45 2015; Racherla et al., 2012;Pierce et al., 2013;Giorgi et al., 1994;Di Luca et al., 2013, 46 2012; , and Europe (Vautard et al., 2013;Jacob et al., 2014;Kotlarski et al., 47 2014;Fischer et al., 2015;Kotlarski et al., 2015;Torma et al., 2015;Giorgi et al., 2016;Zittis et 48 al., 2019;Jacob et al., 2020;Déqué et al., 2007;Gao et al., 2006;Im et al., 2010). climate models 49 (RCMs) have been applied to downscale the GCM outputs to finer scalesSome efforts have 50 also been devoted on regional climate projection in CA with the dynamical downscaling 51 method (Zhu et al., 2020;Ozturk et al., 2017;Mannig et al., 2013). However, their resolutions 52 are still low (≥30km), especially for the mountainous areas in the southeast. Moreover, most 53 of the previous RCM simulations in CA used a single GCM as the lateral boundary conditions, 54 which harbor high uncertainties in the projected climate changes. 55 The present authors carried out a study that involves the dynamical downscaling of temperature and precipitation have been evaluated in a recent study (Qiu et al., 2021) and 60 meanwhile basic features of the projected climate changes have been demonstrated. The 61 results show that the high-resolution RCMs driven by bias-corrected GCMs are excellent in 62 simulating the local temperature and precipitation in CA and detect significant warming, 63 severer heatwaves, and drier conditions in this region in the near-term future.

64
To satisfy the urgent need of high-resolution climate data for assessing the potential 65 impacts of the projected climate changes on many sectors ecological and hydrological 66 applications in CA, especially on ecological and hydrological systems, the HCPD-CA (High-67 resolution Climate Projection Dataset in CA) dataset is derived from the 9-km resolution 68 downscaled results, which includes four geostatic (time-invariant) variables and ten 69 meteorological elements (  2.1 Regional model 85 The Weather Research and Forecasting (WRF) model with version 3.8.1 (Skamarock et 86 al., 2008) is used to downscale the GCMs. It has two domains (Fig. 1b). The outer one covers 87 a large region, with a 27-km resolution and 290×205 grids. The inner one covers the CA 88 region, with a 9-km resolution and 409×295 grids. The model has 33 levels in the vertical 89 4 direction with its top fixed at 50 hPa. Its physical schemes are set based on our previous work 90 about the sensitivity study of different physical parameterizations of the WRF model in 91 simulating the local climate in CA (Wang et al., 2020). Details about them are in Qiu et al. 92 (2021).. Spectral nudging with a weak coefficient of 3×10 -5 is applied in the outer domain 93 (not in the inner one), which prevents possible model drift during the long-term integration 94 by relaxing the model simulations of wind, temperature, and moisture toward the driving 95 conditions. In addition to greenhouse gases and solar constant, the WRF model also considers 96 other external forcing, such as aerosols, volcanoes, and ozone, to make its inner external 97 forcing consistent with the driving GCMs.

216
To sum up, the model evaluation shows that the HCPD-CA dataset has good quality in 217 describing the climatology of all the ten meteorological elements in CA despite some 218 systematic biases (e.g., stronger SWD), which ensures the suitability of the dataset for 219 ecological and hydrological applicationsassessment of future risk from climate change in CA. Climatic Data Center and have been quality controlled (Qiu et al., 2021). Note that a station 259 is compared with the model grid on which it is located. Fig. 8S shows the SCCs and RMSEs 260 of the simulated annual and seasonal T2MEAN over CA before and after adjusting. It is seen 261 that the simulated T2MEAN is more consistent with the observations after vertically 262 interpolating the model data to the elevation of the stations by the standard moist lapse rate 263 of 6.5 ℃/km (Qiu et al., 2017). For instance, after adjusting the SCC of the annual T2MEAN 264 increases from 0.93 to 0.96 and its RMSE decreases from 2.52 to 2.25℃. This proves that the 265 regional model's skills may be underestimated if the elevation differences between the 266 observations and the model grids is not considered.  Fig. 1a). which have been quality 276 controlled (Qiu et al., 2021). Note that a station is compared with the model grid on which it 277 is located.

278
Compared with the 27-km resolution data, the 9-km resolution data largely increases 279 SCCs and reduces RMSEs, especially over the mountainous areas (see the scope of subregion 280 "MT" in Fig. 1c). For instance, over the mountainous areas, the ensemble-mean SCC of 281 annual precipitation increases from 0.38 to 0.58 (Fig. 8c) and the ensemble-mean RMSE of 282 annual precipitation decreases from 1.30 to 1.14 mm/day (Fig. 8d). This highlights the 283 necessity of improving the model resolution from ≥30km to 9km and the advantages of using 284 the HCPD-CA dataset for researches in CA. projection becomes more critical to human development than that for the end of this century.

293
Therefore, this study focuses on projected climate changes over CA in the near-term future 294 (2031-2050). Long-term continuous (e.g., 1986-2100) regional climate projections in CA are 295 more useful for studies in this region and will be conducted in the next stage. Land-use and likely to be ongoing in the future (Micklin, 2007;Ma et al., 2021;Chen et al., 2013;Li et al., 300 2019), such as water extentthe shrinking of the Aral Sea (Micklin, 2007)