The Thirst for Power: The Impacts of Water Availability on Electricity Generation in China.

• Author: An, Yao

1. INTRODUCTION

   Economic development under restricted resource availability has become a complex challenge for both developing and well-established economies. The availability of resources, like water, has been drastically affected by global warming with increased frequency in droughts and heat waves (Milly et al. 2005; Olmstead 2010), which lowers agriculture production (Mendelsohn et al. 1994) and disrupts international trade (Debaere 2014). More damagingly, electricity supply becomes unreliable due to increased water temperature and reduced water availability, as water is both directly used for power generation in hydropower plants, and indirectly for cooling in thermal and nuclear generation. Firm productivity has been significantly limited by the increased electricity scarcity (Fisher-Vanden et al. 2015). To maintain a sustainable resource supply and mitigate the impact of water shortage on economic development, it is therefore important to understand how utility firms respond to the change in water availability and unpacks the underlying mechanisms of power outage.

   This paper addresses this by investigating the technologies that utility plants use to generate electricity. Although the correlation between frequent droughts and persistent electricity short-ages is well known, it remains unclear as to how and why electricity shortages are related to drought. Relevant studies on this relationship are scant, especially for developing countries, possibly because meteorological data on water shortage and temperature at the power plant level are not readily available. Furthermore, the literature is silent on whether the technology switch for cooling and power generation can be attributed to a decline in water availability. This lack of information reduces the efficiency of investment decisions in new technologies, as the technology substitution that is induced as a result of water shortages may be different from the choices suggested in the engineering models. Although generation capacity has increased significantly since the early 2000s, electricity shortages remained a major challenge for China in the early 2010s. It is reported that the prolonged drought in spring intensified the electricity supply in Hainan province in 2007. In addition, lower water level in Yangtze River triggered the decline in hydro power generation. Thus, the drought in central China leads to electricity brownouts and power rationing in 2011. In all, the electricity shortage lasted for almost whole year, with a cumulative power rationing of about 35.2 billion kWh in China (Yue, 2021). As an essential input for production, electricity shortages can lower firm revenue by 10% (Allcott et al. 2016) or increase production costs by 8% (Fisher-Vanden et al. 2015).

   In this study, we match a plant-level panel of electricity generation information with a fine-scale monthly meteorological dataset on electricity generation, installed capacity, and water-use characteristics, relevant social economic factors of individual plants, and climatic information for plant locations for the period from 2007 to 2014. The unique dataset makes it possible to identify the mechanism behind the connections between water availability and electricity shortage, which can provide important information for managerial practice for achieving sustainability with restricted resource constraints. This is because water is the necessary resource for most plants in China. Hydropower generation requires large quantities of water to be available to operate the turbines. In addition, water is the major source for cooling in thermal power plants and nuclear plants.

   To shed light on the channels through which an electricity shortage is related to drought, we apply a fixed-effect model to study the substitution effect of variations in water shortage on plant generation technology selection. Causal attribution in this paper relies on the extent to which water shortage variations affect electricity supply exogenously after controlling for cooling degree days, installed capacity, electricity price, and fixed assets investment. We follow Couttenier and Soubeyran (2014), Eyer and Wichman (2018) to use Palmer Drought Severity Index (PDSI) as a proxy for water scarcity. PDSI values are computed using observed air temperature and precipitation and self-calibrated dynamically with the climate and duration factors to capture spatially comparable long-term drought trends (Alley 1984; Dai 2011b; Dai 2017). In addition, we employ two alternative proxies for water scarcity: the Standardized Precipitation Evapotranspiration Index (SPEI) for the relative long-term and precipitation (P) for the short term. Precipitation in a given location measures the instantaneous water availability in the plant's surroundings, while the PDSI and SPEI are widely used indexes that incorporate past and current supply of precipitation and demand for potential evapotranspiration moisture, and capture the impact of global warming on drought severity (Dai 2011a, 2017; Vicente-Serrano, and National Center for Atmospheric Research Staff 2015). In this paper, water scarcity and draught are interchangeable terms used, and both terms indicate the decline in water availability. We show that based on fuel type, coal and nuclear are called upon to substitute for the forgone hydropower when water availability declines.

   We find that a one-standard-deviation decrease in water availability causes an approximate 205 GWh decline in hydro power generation, a 145 GWh increase in nuclear generation, and a 28 GWh increase in coal generation. To minimum the estimation biases stemming from omitted variables, we also control for time invariant firm fixed effects, year-of-sample fixed effects, power grid region by year, and fuel type by year fixed effects in all our estimations. The results are robust for alternative measures of the water scarcity index. We also rule out factors that may confound our results on the technology substitution, such as newly constructed generators and regular generator maintenance.

   To support our findings, following Chen and Yang (2019), we further construct quarterly average water scarcity as an alternative measure. The finding is consistent with that of using annual data. We also find that water scarcity may result in electricity shortages in the second and third quarters of a year in particular due to a technology shift. In addition, we introduce water availability bins to address the monthly nonlinear relationship between the severity of water scarcity (abundance) and power generation. A remarkable finding is that electricity generation responds positively to severe drought or extreme wet. We note that a change from moderate drought (moisture) to extreme drought (moisture) enhances electricity generation, especially coal-fired generation. Moreover, we analyse the effects of the characteristics of water withdrawal mechanisms for generators. Specifically, we focus on the water sources the plants rely on and the technology that the plants have installed to cool generators. We identify that technology selection in China's electricity sector is likely to move from relatively water-intensive generation technologies towards less water-intensive technologies. Our result is consistent with prior studies (DeNooyer et al. 2016; Eyer and Wichman 2018). By examining the spatial effects of water scarcity on technology substitution across grids, we find that water scarcity-induced electricity shortages cannot be alleviated by supply from other grids via inter-grid transmission, which may suggest that the inefficiency of the power grid dispatch and transmission system is another reason for the frequent power shortages that occurred in the early 2010s in China.

   In addition, we address the environmental consequences of the technology selected for electricity generation induced by water scarcity (Amor et al. 2014; Clancy et al. 2015; Jacobsen and Schroder 2012). As discussed by Moomaw et al. (2011) in their lifecycle fuel summary analyses, coal-fired plants emit 840 g/kWh carbon dioxide, whereas hydroelectric and nuclear power plants emit little carbon dioxide. If droughts lead to a shift from the relatively water-intensive fuel source of hydro power towards nuclear or coal, emissions will rise accordingly. Hence, it is necessary to clarify to what extent the increased coal-fired generation resulting from drought results in rising greenhouse gas emissions. Our results suggest that the rising use of coal for electricity is associated with an increase in C[O.sub.2] emissions. The results imply a hidden increase in carbon emission up to 32000 tons of each thermal power plant per year, resulting in an additional cost of 0.18 million USD. We also note that a change from moderate drought to extreme drought would sharply intensify C[O.sub.2] emissions. An additional month at (-[infinity], -5), [-5, -3), [-3, -1), [-1, 1) [1, 3), [3, 5), and [5, [infinity]) bins of PDSI can lead to a substantial increase in average yearly C[O.sub.2] emissions from each coal power plant by roughly 27064 tons, 12632 tons, 9071 tons, 6404 tons, 2517 tons, 499 tons and 12078 tons, respectively.

   In the literature, there are a growing number of studies that have estimated the effect of climate change on the energy sector (Rubbelke and Vogele 2011; Van Vliet et al. 2012; Olmstead 2014; Li et al. 2019; Zhou et al. 2019; Craig et al. 2019). Most of the existing studies apply non-econometric approaches, including lifecycle assessment (Gao et al. 2019), engineering models (Feeley III 2008; Rubbelke and Vogele 2011), and integrated assessment (Khan et al. 2016; DeNooyer et al. 2016), with a focus on developed regions (Miara et al. 2017; Behrens et al. 2017). With access to well-documented micro-level data, climate variations associated with human adaptation activities are an emerging trend in the studies (Auffhammer and Aroonruengsawat 2011; Barreca...