Responses of runoff to historical and future climate variability over China

China has suffered some of the effects of global warming, and one of the potential implications of climate warming is the alteration of the temporal–spatial patterns of water resources. Based on the long-term (1960–2008) water budget data and climate projections from 28 global climate models (GCMs) of the Coupled Model Intercomparison Project Phase 5 (CMIP5), this study investigated the responses of runoff (R) to historical and future climate variability in China at both grid and catchment scales using the Budyko-based elasticity method. Results show that there is a large spatial variation in precipitation (P ) elasticity (from 1.1 to 3.2) and potential evaporation (PET) elasticity (from−2.2 to −0.1) across China. The P elasticity is larger in northeastern and western China than in southern China, while the opposite occurs for PET elasticity. The catchment properties’ elasticity of R appears to have a strong non-linear relationship with the mean annual aridity index and tends to be more significant in more arid regions. For the period 1960– 2008, the climate contribution to R ranges from −2.4 to 3.6 % yr−1 across China, with the negative contribution in north-eastern China and the positive contribution in western China and some parts of the south-west. The results of climate projections indicate that although there is large uncertainty involved in the 28 GCMs, most project a consistent change in P (or PET) in China at the annual scale. For the period 2071–2100, the mean annual P is projected to increase in most parts of China, especially the western regions, while the mean annual PET is projected to increase in all of China, particularly the southern regions. Furthermore, greater increases are projected for higher emission scenarios. Overall, due to climate change, the arid regions and humid regions of China are projected to become wetter and drier in the period 2071–2100, respectively (relative to the baseline 1971–2000).

based on the Budyko-based elasticity approach. The results indicated that the P elasticity exhibits a large regional variation, 65 with a small range in southern China, the Songhua River basin and the northwest and a large range in the Hai River basin, 66 the Yellow River basin, and the Liao River basin. Although the aforementioned studies have certainly made advances in 67 understanding the climate elasticity of R in China, our knowledge about the responses of R to climate change over various 68 temporal and spatial scales remains rather limited due to the large regional variation in climate types and catchment 69 characteristics. The question of how climate change will affect R over China in the future is also an important problem to be 70 addressed. Developing a more accurate and quantitative understanding of the changing water resources over various 71 temporal and spatial scales under a changing environment is therefore a high priority for China.

182
The P elasticity and PET elasticity of R based on Equations (2) and (3) are estimated at each of the 0.5 o grids in China. As 183 shown in Figure 5, the spatial patterns of P elasticity and PET elasticity from Equations (2) and (3) are almost the same in all 184 regions of China. There is a large spatial variation in P elasticity and PET elasticity, i.e. ranging from 1.1 to 3.2 and from 185 -2.2 to -0.1 across China, respectively. In particular, P elasticity is more significant in the northeast and western areas than in 186 southern China, which is in contrast to PET elasticity. Figure 6 shows the relationship between  and climate (P and PET) 187 elasticity. As shown, the P (PET) elasticity first increases (decreases) and then decreases (increases) with the increase of  188 under not very dry conditions (i.e.  <10). However, when  becomes large enough (e.g.  >10), both P and PET 189 elasticity becomes insensitive to  .

191
The climate elasticity estimated for each of the 14 large basins is shown in

205
The spatial distributions of catchment properties elasticity from Equations (2) and (3) are displayed in Figure 5(e) and (f). As

206
shown, the catchment properties elasticities for these two equations are rather similar across China, and the values of elasticity is very weak (approximately equal to 0) in southern China and some regions of northeast China, but it tends to be 209 more significant in some water-limited regions of northwest China. Figure 6( Table 3. A positive contribution of P is detected in Southeast Drainage,

240
3.5 Future climate change 241 Figure 9 shows the uncertainty range of the relative change in mean annual P and PET in the basins for the period 2071-242 2100 under the RCP2.6, RCP4.5, and RCP8.5 scenarios as predicted by 28 CMIP5 models (relative to the baseline 1971-  close to our results for P elasticity ranging from 1.1 to 3.2, and for PET elasticity ranging from -2.2 to -0.1 in China. It is 303 worth noting that the values of P elasticity tend to be larger in the northeast and some parts of western China that are located 304 in arid climates. This is in good agreement with the findings by Sankarasubramanian et al. (2001), which indicated that a 305 larger P elasticity occurs in more arid regions. However, some parts of Xinjiang, which is more arid than southern China,

306
have smaller P elasticity. Meanwhile, some parts of southern China, which is more humid than other regions in China, have 307 larger P elasticity ( Figure 5). In addition, the Haihe River basin, located in less arid climates than that of the northwest,

308
shows the largest P elasticity in China (Table 2)

338
In general, the Thornthwaite method corrected by Equation (1)

393
In addition to uncertainty in PET calculation (as discussed in section 4.2), there are also uncertainties associated with the 395 estimates of other water budget components, such as R. As shown in Figure 14, the sensitivity of climate (i.e., P and PET) 396 elasticity to R varies considerably between basins and tends to be larger in more humid basins. Moreover, PET elasticity is 397 more sensitive to changes in R compared with P elasticity for all 14 basins.