Spatial and temporal variations in hydro-climatic variables and runoff in response to climate change in the Luanhe River basin, China

Abstract

Climate change in North China would result in significant changes in temperature, precipitation and their spatial/temporal distributions. Consequently, these induced changes will have profound effects on the hydrological cycle and water resources in both agricultural and natural ecosystems. Panjiakou reservoir in the middle Luanhe River basin—a tributary of the Haihe River basin—is one of the important sources of water for industrial and agricultural development in Beijing, Tianjin and Hebei province, China. Any significant change in the magnitude and/or timing of runoff from the reservoir induced by changes in climatic variables would have significant implication for the economic prosperity in North China. This paper investigates the impacts of climate change on hydrological processes in the Luanhe River basin as follows. Firstly, spatial and temporal patterns of precipitation, temperature and runoff at both annual and seasonal scales from 1957 to 2000 in the Luanhe River basin are analyzed using Mann–Kendall trend analysis, linear regression methods and inverse distance weighted interpolation. For the impact study, four Global Climate Models (GCMs) (named CSIRO, HadCM3, CNRM and GFDL) were used to produce precipitation and temperature data under A2 scenario by mean of a widely used quantile–quantile transformation. Projected meteorological variables were used to force a two-parameter hydrologic model to simulate the hydrological response to climate change in the future (2021–2050). Moreover, a sensitivity analysis is conducted to assess how precipitation and temperature affect the runoff. Results suggested that most part of the Luanhe River basin was dominated by significant increasing trends of temperature and no significant trends of precipitation in annual and seasonal scale during the past decades. Annual, spring and autumn runoffs present significant decreasing trends in the Panjiakou reservoir basin. Meanwhile, runoff is more strongly related to precipitation than to temperature. All GCMs projected precipitation and temperature series after bias correction indicated increasing temperature and increasing precipitation trends for the period 2021–2050 except that CNRM showed a slight decreasing trend in precipitation. Great enhancements can be found in projected runoff except CNRM by driving the two-parameter water balance model. The study provides valuable information on the assessment of the impact of the climate change on water resources in the Luanhe River basin as well for allocating and designing water resources projects.

Appendix: Criteria for model performance

Normally model parameters are estimated by calibration and model performance is tested through verification. Correspondingly, the whole data set is divided into two parts, the calibration period (N _c ) and the verification period (N _v ). Only when the performance of the model is satisfactory in both calibration and verification periods could the model be used with confidence in practice.

The percentage of volume difference D _v (in percentage) and the Nash and Sutcliffe coefficient of efficiency E _ns are used to measure the model performance. The percentage of volume difference is defined by

$$ D_{v} = \sum {{{\left( {Q_{i} -{\mathop Q\limits^{ \wedge }}_{i} } \right)} \mathord{\left/ {\vphantom {{\left( {Q_{i} - \mathop Q\limits^{ \wedge }_{i} } \right)} {\sum {Q_{i} } }}} \right. \kern-0pt} {\sum {Q_{i} } }}} \times 100\;\% $$

(6)

where Q _i is the observed monthly runoff, ${\mathop Q\limits^{ \wedge }}_{i}$ is the simulated monthly runoff. A simulation is considered to be satisfactory when D _v is below 10 % and excellent when D _v is less than 5 %. The coefficient of efficiency describes how well the volume and timing of the calibrated hydrograph compared to the observed hydrograph and is defined by

$$ E_{ns} = 1 - \frac{{\sum\limits_{i} {\left( {Q_{i} - {\mathop Q\limits^{ \wedge }}_{{i}} } \right)^{2} } }}{{\sum\limits_{i} {\left( {Q_{i} - {\mathop Q\limits^{{ - }}}_{C} } \right)^{2} } }} \times 100\;\% $$

(7)

with ${ \mathop Q\limits^{ - }}_{C}$ being the mean monthly runoff in the simulation term. The closer the value of E _ns is to 1, the more successful the model calibration/validation.