Quantifying the contribution of climate- and human-induced runoff decrease in the Luanhe river basin, China

Quantifying climate variability and human activities on mean annual runoff (MAR) has received much attention, since climate change and human activities are known to affect hydrologic cycles and exert impacts on our environment on a global scale (Barnett et al. 2008; Zhang et al. 2008; Wang & Cai 2010). Human activities such as land use changes, water withdrawals, reservoir operations, and return flow can affect hydrological processes (Wang & Cai 2010; Schilling et al. 2010). Climate change and its impact on water resources is a problem that has to be coped with worldwide (Piao et al. 2010), because climate change can affect runoff by the redistribution of precipitation and temperature change (Karl et al. 1996). However, hydrological processes are subject to the combined effects of climate change and human activities (Berry et al. 2005; Rodriguez-lturbe & Porporato 2005; Donohue et al. 2007), resulting in basin-scale changes in runoff or the water balance (Sun 2007; Zhang et al. 2008). Therefore, quantification of the impacts of climate change and human activities on hydrology and water resources is of vital necessity (Lane et al. 2005; Siriwardena et al. 2006; Tuteja et al. 2007; Huo et al. 2008; Yang & Tian 2009; Wei & Zhang 2010; St. Jacques et al. 2010).

In order to quantify the climate- and human-induced impacts on MAR accurately, various methods have been proposed. One way was to adopt hydrological models such as SWAT (Liu et al. 2013a), HBV (Liu et al. 2013b) and SIMHYD (Bao et al. 2012), etc., by varying the meteorological inputs to obtain the quantitative assessment of human impacts. However, there were structural errors for hydrological models, suggesting that the estimated human impacts may be inaccurate or insensitive to climate change (Sankarasubramanian et al. 2001; Sun 2007; Zheng et al. 2009; Xu et al. 2014). Another way was to use empirical approaches or conceptual models. Empirical approaches such as statistical regression methods have always been short of physical meaning (Ma et al. 2008; Schilling et al. 2010; Xu et al. 2014). Conceptual models built on the principle of water-energy balance have been useful for investigating the hydrological response (Renner et al. 2012b). Some conceptual models were developed based on the Budyko hypothesis (Budyko 1974). Wang & Hejazi (2011) quantified the climate- and human-induced impacts for 413 gauge stations by the Budyko decomposition method across the USA. Wang et al. (2013) used hydrological sensitivity analysis, the Budyko decomposition method and Zhang's curve (Zhang et al. 2001) to evaluate the effects of climate variability and human activities on runoff in the Haihe River basin, suggesting that the Budyko decomposition method was valid in the Haihe River basin. Li (2014) developed a stochastic soil moisture model within the Budyko framework (Fu's equation), and distinguished the impact of interannual variability of precipitation and potential evaporation on evapotranspiration in the USA.

Moreover, in a remarkable paper, Tomer & Schilling (2009) introduced a method to distinguish climate effects from land use change effects on runoff. They observed different soil conservation treatments, and the watershed showed different evaporation ratios. They found that the shift within this hydro-climatic state space was perpendicular to the observed shift over time, due to the conservation treatments. Renner et al. (2014) used this separation method to quantify the contributions of environmental factors to evaporation in 68 small–medium river basins, which cover the greatest part of the German Federal State of Saxony, and they confirmed the validity of this method. However, the geometric approach is rarely used in China.

Many studies in recent years have focused on quantification of climate change and human activities on MAR in the Luanhe river basin, China. The Luanhe River is a very significant water resource for Tianjin City, an important municipality city in China. Previous studies chose different models/methods to quantify the influence of climate change and human activities on MAR, and they achieved opposite results about whether climate variability or human activity was the main driving factor for the reduction of the MAR in the Luanhe river basin. Bao et al. (2012) used the Variable Infiltration Capacity (VIC) model to prove that climate variability was the major driving factor in the Luanhe river basin. Zeng et al. (2014) combined a hydrological model (Distributed Time Variant Gain Model) and a global terrestrial biogeochemical model (CASACNP) to estimate the effects of climate change, land use/land cover (LUCC) and increase in CO₂ concentration on runoff in the Luanhe river basin, and they reported that the effects of climate change and LUCC on runoff are stronger. Xu et al. (2013) made a statistical analysis and found a 79.5% decrease of annual inflow in the Panjiakou Reservoir caused by climate change and human activities. However, they did not separate the contributions of each factor. Wang et al. (2013) discovered that the impact of human activities was the main driving factor for the decline of annual runoff in the Luanhe river basin. Therefore, there is a need for a consistent understanding of the dominant cause for the reduction of runoff in the Luanhe river basin.

The main objectives of this study were to: (1) illustrate runoff decrease in the study regions by observing the annual rainfall-runoff relationships; and (2) quantify the contributions of climate change and human activities to MAR in the Luanhe river basin by the Budyko decomposition method and the geometric approach.

The spatial and temporal distribution of the precipitation within the Luanhe river basin is uneven, and about 70–80% of the annual precipitation falls in the rainy months of June to September. The region receives an average runoff of 4.69 billion cubic meters per year. Floods in the basin are often caused by rainstorms and occur in July and August, since the rainstorms with characteristics of short duration and high intensity are likely to result in floods of high peaks and large amounts.

The Luanhe river basin is well known for its water supply function for the Tianjin city, an important metropolis of China. It was planned to introduce a billion cubic meters of water to Tianjin per year from the basin. However, the actual amount of water transferred to Tianjin has been less than this for several years, especially in the last decade (Li et al. 2014). The deficient water supply to Tianjin is mainly due to the decrease of annual water storage in the Panjiakou reservoir in the 21st century. Due to rainfall reduction, land use change and construction of many small check dams for soil and water conservation, the average annual runoff decreased by about 30% after 1980 (Li & Feng 2007). The long-lasting water shortage aggravated the water crisis in Tianjin city (Yi et al. 2011; Liu & Wu 2012). In this study, six sub-watersheds were selected, Luanhe, Yixunhe, Wuliehe, Liuhe, Baohe and Laoniuhe, with areas of 17,100 km², 6,761 km², 2,460 km², 626 km², 1,661 km² and 1,615 km², respectively.

In this study, the Budyko decomposition method and the geometric approach were used to quantify the contributions of climate-caused and human-induced impact on MAR. The procedures of both methods are listed in the following sections.

Budyko (1948) and Budyko & Zubenok (1961) postulated that the mean annual evaporation from a watershed could be determined from precipitation and net radiation, known as the ‘Budyko curve’. Based on worldwide data on a large number of watersheds, Budyko (1974) developed a framework for estimating actual evapotranspiration based on a dryness index. Dooge (1992) studied how the decomposition method might approach the problem of analyzing the sensitivity of streamflow to climate change. In 1999, Dooge et al. (1999) considered that both precipitation and potential evapotranspiration can lead to changes in water balance. On this basis, Ma et al. (2008) postulated that the total change in MAR can be estimated as the change in MAR due to climate variability and the change in MAR caused by land use/cover change.

In order to calculate the contribution of climate- and human-induced impact on the runoff, the specific steps (referred to by Renner et al. (2014)) are as follows:

1. Calculate the annual average of water partitioning and energy partitioning in the pre-change and post-change periods, respectively, i.e., q₀=E₀/P₀ and g₀=E₀/E_P0, q₁=E₁/P₁ and g₁=E₁/E_P1, where P₀, P₁, E₀, E₁, E_P0 and E_P1 represent the average annual precipitation, actual evaporation and the potential evaporation in different periods, respectively.

2. Solve the angle of θ, the angle between both vectors, which can be described by the scalar product divided by the vector magnitudes in two different ways, given as:

   

   13
3. Derive the numerical of g_b and q_b alternating the use of Equations (12) and (13). After conversion, the mathematical expressions are as follows:

   

   14

   

   15
4. E_b at point (q_b, g_b) can be computed as E_b = q_bP₀. The absolute differences between evaporation rate (E₀, E₁) and E_b can be available for determining the climatic (ΔE_C) and human-induced (ΔE_H) parts of change:

   

   16
5. Calculate the climate- (ΔQ_C) and human-induced (ΔQ_H) contributions to the runoff change. They can be computed directly by assuming zero storage change in the water balance equation. The specific expressions are:

   

   17

In order to assess the influence of environmental change on runoff variation, the change point of the runoff series should be detected first in the six sub-watersheds. Wang et al. (2013) identified 1979 as the change point of streamflow time series in the Luanhe river basin by Pettitt's test; Li & Feng (2007) gave the same change point in the six sub-watersheds. Therefore, we divided the long-term period into two periods: the reference period, 1956–1979, was the pre-change period, and the impaired period, 1980–2011, was the post-change period.

To quantify the magnitude of the MAR decrease, we calculated the MAR in the pre-change and post-change periods of the six sub-basins. The results are shown in Table 1. As can be seen, MAR decreased 22–44% relative to that in the pre-change period.

However, the contributions of the driving factors were unclear, and should be quantified by the methods in the following sections.

The climate- and human-induced percentage changes of MAR using Equation (10) are shown in Table 2. Human activities have mainly driven the decrease of runoff, which was consistent with results obtained in previous studies (Liu et al. 2013a, 2013b). Human activities accounted for more than 70% in five sub-basins out of six. The contribution of human activity to the reduced runoff in the Sandaohezi sub-basin was estimated as 85%, which was the highest among all the six sub-basins. Seventy-nine and 76% were the human-induced percentage of annual runoff decrease in the Liuhe and Wuliehe sub-basins, respectively. The smallest contribution of human activities occurred in the Baohe sub-basin, which was 62%. By contrast, the climate contribution was less than 30% in five sub-basins except in the Baohe sub-basin which accounted for 38%.

As can be seen, the human-induced percentage change ranges from 51 to 93%. For the six sub-basins, the quantified results were almost the same as those estimated by Liu et al. (2013a) via the SWAT model. In the Baohe sub-basins, climate-caused impact accounted for a part of the change in MAR, though the human-induced impact was still an important factor, accounting for more than 50%. Therefore, the impact of human activities is the main driving factor for the decline of annual runoff.

The Budyko decomposition method and the geometrical approach were employed in this study. Both methods required the same basic meteorological data such as precipitation, wind speed, humidity, and temperature, and the same hydrological data such as the observed runoff series. The absolute differences in the human-induced contribution, i.e., Δ/H/ = /H_g-H_b/, derived using the two methods are shown in Table 2.

From Table 2, the differences in human-induced percentage change were from 1 to 17%. The largest difference in human contribution was 17% in the Sandaohezi sub-basin, and it was 11% less by the geometric approach for the Baohe sub-basin.

The differences between the Budyko decomposition method and the geometric approach were mainly: (1) the Budyko decomposition method provided more explicit physical meaning (Renner et al. 2012b; Wang et al. 2013), but the geometric approach was a conceptual model (Renner et al. 2012b); (2) a single-parameter Budyko-type curve was used as the Budyko decomposition method, and the parameter was related to the complex interaction between vegetation types, soil properties, and topography (Ma et al. 2008), although the geometric approach required two non-dimensional hydrologic state variables to describe the hydro-climatic state of a basin; and (3) the largest difference between both approaches occurred under limiting conditions. The geometric approach did not adhere to the water and energy limits, while the Budyko decomposition method accounted for limiting conditions, and was the reason why the difference in the human contribution from those two approaches in the Sandaohezi sub-basin was the highest of all. The value of ω was 3.65 in the Sandaohezi sub-basin, almost close to the water limit line (see Figure 6).

Runoff decrease in the Luanhe river basin was assessed by the rainfall–runoff relationship in our study. Runoff in most sub-basins was affected by climate change and human activities, such as agricultural irrigation, industry development and dam construction. According to Wang et al. (2013), the change points of runoff in the Luanhe river basin occurred in the year 1979 using Pettitt's test. The late 1970s saw the start of the construction of hydraulic structures and land reform in China, and the increasing agricultural land and the amount of irrigation water led to a large increase in water use, which was the main driving factor of runoff decline (Yang & Tian 2009; Wang et al. 2013).

In this study, the Budyko decomposition method and the geometric approach were adopted to explicitly quantify the relative contributions of climate and human activities to MAR in the Luanhe river basin during the period 1956–2011. Using the change point of 1979 (Li & Feng 2007; Wang et al. 2013), we split the time period into two sub-periods: ‘pre-change’ and ‘post-change’. In addition, the Budyko decomposition method and the geometric approach rely on the assumption of zero storage change. The period (1956–2011) is long enough to make sure that the assumption holds, since the significant interannual storage change can be attributed to the human contribution.

Both the Budyko decomposition method and the geometric approach obtained similar results for the impact of human activities on runoff, which inspired great confidence in the impact assessment of this study. The quantitative evaluation of climate impacts in these six sub-basins was 15–40%, approximately, and the contribution of human activities was 60–85%, based on the Budyko decomposition method. According to the geometric approach, the effects of climate change in these six sub-basins was 7–49%, and the contribution of human activities was approximately 51–93%.

Other researchers have carried out this work in the Luanhe river basin using different methods. Wang et al. (2013) quantified the impact of climate variability and human activities for the reduction of runoff during 1957–2000 in the Luanxian hydrological station. They used the hydrological model method, the hydrological sensitivity method and the climate elasticity method, and found that the relationship between precipitation and runoff had changed abruptly. Sixty-one, 67 and 57% of runoff reduction were attributed to human activities by these three methods, respectively. Xu et al. (2013) adopted the geomorphology-based hydrological model (GBHM) and a climate elasticity model to distinguish effects of climate change and human activities during 1956–2005 at Sandaohezi station. The contributions of human activities and climate variability were 61% and 39%, respectively. The major causes of runoff decrease in the Luanhe river basin were consistent with our findings. However, there was a degree of difference in the climate-caused percentage change. These differences might come from the different model structures, the drainage area, the study period and the uncertainty of the model parameters. Additionally, the geometric approach, rarely used in China, can be demonstrated to be an alternative way to separate the effects of streamflow by comparing its consistent results with other methods in the Luanhe river basin.

The uncertainties should be noted and are related to separating the effects of climate variability and human activities on streamflow. Firstly, the one-parameter model of the Budyko composition method easily satisfies the data requirement, but it can provide less information about a detailed description of the hydrological process. Secondly, both of the methods would be uncertain when the distribution of precipitation changes. At the same time, the streamflow can be influenced by changes in other precipitation characteristics, such as seasonality, intensity and concentration. For example, we generally used P-Q to estimate E. However, in a wet year (high P, low aridity index E_P/P), if we still assumed ΔS_W=0 (Equation (1)), basin E would be overestimated. Whereas, in a dry year (low P, higher aridity index E_P/P), basin E would be underestimated. Furthermore, in order to simplify the issue, we assume the soil moisture change to be zero. Actually, the assumption could lead to an error in the final results. The change of soil water storage (ΔS_W) is assumed to be imbalanced between P-E and Q. For example, in a wet year, high infiltration and high precipitation would lead to high soil moisture (ΔS_W>0), when we ignore the soil moisture change, P-Q would result in the overestimation of E. Therefore, the uncertainty may be caused by assuming the soil moisture change to be zero. Moreover, these uncertainties can inevitably affect the results. Therefore, future studies should take them into consideration to improve the accuracy of the results.

Our research could be useful for water resources planning and management to choose the methodology and to separate the effects of climate change and human impacts on MAR. The deficiency is that we assumed the soil moisture change to be zero, which would result in a small error in the water budget. In the future, the extension of this work will take the soil moisture change into consideration, and further, the impact of different human activities on streamflow will also be considered.

This work was supported by the National Natural Science Foundation of China (No. 51479130). We are also grateful to the Hydrology and Water Resource Survey Bureau of Hebei Province for providing the hydrometeorological data.