Introduction

Many environmental problems have accompanied the recent economic growth in China, the most common being accidents leading to surface water pollution (Huang et al. 2013; Xie et al. 2013; Yao et al. 2015). Xue and Zeng (2011) analyzed the distribution of environmental pollution accidents in China and found that surface water pollution accidents accounted for 46.05–63.03 % of the total from 1990 to 2007 (Xue and Zeng 2011). Many of these accidents have had widespread effects and caused great damage to entire communities, including economic loss, environmental deterioration, and negative social impacts. For example, in 2005 over 100 tons of nitrobenzene was released into the Songhua River, forming an 80 km pollution belt that flowed downstream. This accident disrupted the water supply of all 4,000,000 residents of Harbin for 4 days. Moreover, the pollutants eventually flowed into the Heilongjiang River on the Chinese–Russian border, where they threatened the water quality of the city of Khabarovsk, which has more than 600,000 residents (Wo 2009). In another incident, cadmium pollution entered the Beijiang River in December, 2005 (Protection Wogesae 2011) and an algal bloom led to odorous water in Wuxi in May, 2007 (Wo 2007), both of which seriously affected human health and led to large-scale social panic in China. These accidents have resulted in more attention being paid to pollution to prevent surface water accidents.

The available analyses of water pollution accidents in China have primarily focused on particular aspects, such as toxicological effects (Lei et al. 2008; Liu et al. 2008), methodologies for risk evaluation (Wang and Zhang 2007; Zhang et al. 2006), treatment technologies, and environmental emergency response (Li et al. 2009; Wu et al. 2006). Although these research projects have improved and enhanced our understanding of the nature of problems caused by water pollution accidents, none have provided a sound portrait of surface water accidents in China. Accordingly, additional work is needed to identify characteristics of surface water accidents to provide decision makers with valid, plausible suggestions regarding water environmental risk evaluation, prevention, and mitigation.

Historical accidents lead to new policies and laws on risk mitigation, and many researchers have conducted historical analyses of pollution accidents (Darbra and Casal 2004; Hohenemser et al. 1977, 1983; Kang 1999; Montiel et al. 1996; Uth 1999; Vilchez et al. 1995; Yang et al. 2010). Uth (1999) evaluated major industrial accidents in Germany from 1993 to 1996, conducted rooted causes analysis based on these cases, and demonstrated the importance of maintenance, detailed knowledge of chemical properties, human factors, and the role of safety organization (Uth 1999). Kang (1999) classified 93 major industrial accidents that occurred from 1988 to 1997 according to the type of accident, equipment involved, circumstances, and causes, concluding that some plants lacked well trained operators, safety inspection, and safety consciousness (Kang 1999). Darbra and Casal (2004) analyzed 471 accidents occurring in seaports between the beginning of the twentieth century and October 2002. The trends showed a significant increase in the frequency of accidents over time, and 51 % of the accidents were found to originate from loss of containment (Darbra and Casal 2004). Montiel et al. (1996) grasped the main features of accidents occurring in the transportation and distribution of natural gas by analyzing 185 accidents from the MHIDAS database (Montiel et al. 1996). Vilchez et al. (1995) investigated the contribution of different situations, activities, and equipment to the risks associated with hazardous materials based on 5325 accidents from the database MHIDAS (Vilchez et al. 1995). Yang et al. (2010) analyzed the characteristics of 322 road accidents that occurred during road transportation of hazardous materials in China from 2000 to 2008 (Yang et al. 2010).

Historical analysis is commonly applied to determine the probability distribution of related types of accidents during environmental risk analysis (Darbra and Casal 2004; Oggero et al. 2006; Yang et al. 2010), and information implied in historical accidents is fairly important to probability prediction during environmental risk evaluation. In contrast to other probability analysis methods such as even-tree analysis (You and Tonon 2012), mathematical modeling of probability (Thodi et al. 2010), and scenario tree modeling (De Vos et al. 2004), historical analysis is more practical and more easily implemented, especially when there are several types of risk sources.

The present study was conducted to (1) collect representative cases of surface water accidents in China; (2) describe the distribution of these water pollution accidents as a function of time, location, and the materials involved; (3) analyze the accumulated probability of quantities of pollutants in these cases, number of people involved, length of pollution belt, and time for which pollution lasted; (4) summarize the probabilities of occurrences of various types of origin and causes.

Materials and Methods

Accident Collection and Screening

There are no public databases of water pollution accidents in China and no detailed statistical information is available. Accordingly, it is very difficult to gather accident information. For this study, preliminary accident information was obtained from the following sources, which complemented and verified each other:

  1. (1)

    Chinese Newspaper Database on the CNKI academic knowledge network service platform (CNKI 2012b). The database contains important academic and archival information from 500 domestic newspapers published after 2000. Searches were conducted individually or together using keywords such as river pollution accident, lake pollution accident, tap-water suspension incident, large area of fish deaths, hazardous material leakage, etc.

  2. (2)

    China Environmental News This newspaper is the only professional national newspaper pertaining to environmental protection and has been published since 1984. All issues from the first volume to that of December, 2012 were reviewed in their entirety for reports of surface water accidents.

  3. (3)

    Journals and magazines in the field of environmental protection (CNKI 2012a). A database of Chinese academic journals was searched using the same keywords as for the Chinese Newspaper Database.

  4. (4)

    Internet Online data were screened for reliable information using the procedure listed in Fig. 1. Reports were prioritized as follows: official websites (such as that of the local Environmental Protection Bureau, XINHUA.net, Chinanews.net, RENMIN.net), newspaper websites, other dominant and reliable websites (SINA.com, SOHU.com, etc.). For reports from other sources, credibility was determined based on the presence of photos and multiple reports in other websites.

    Fig. 1
    figure 1

    Flow chart of accident screening from the internet

    Full size image

All accidents selected occurred before December, 2012. For each accident, the time, location, pollutants, length of pollution zone (belt), time for which pollution lasted, cause of the accident, loss of the accident eventually, and information source were provided.

Statistical Analysis

Social risk curves, usually referred to as F–N curves (Haastrup and Brockhoff 1990; Kasperson 2014), have been used to analyze accident fatality statistics based on accumulated frequency–number of deaths graphs in many studies (Darbra and Casal 2004; Jonkman et al. 2003; Oggero et al. 2006; Yang et al. 2010). This method relates the number of deaths in a particular accident to the relative probability of there being that number of deaths.

This type of plotting was conducted in the present study to describe the distribution and accumulated probability of the characteristics of surface water accidents. The number of deaths in the F–N curve was then expanded to other features. Similar curves can be drawn based on the frequency and quantities of chemicals involved (F–Q), frequency and number of people influenced (F–P), frequency and number of people affected (F–A), frequency and time for which pollution lasted (F–T), and frequency and length of pollution zone (F–L). These curves were used to establish whether the probability of occurrence of a type of accident with a given severity is proportional to the probability of occurrence of another accident with a different severity, which is the relative probability. For this analysis, the accumulative probability or frequency was calculated using the equation:

$$P(x \ge j) = \frac{{\sum\nolimits_{i = j}^{n} {N_{i} } }}{{\sum\nolimits_{i = 1}^{n} {N_{i} } }},$$

where j refers to the number of occurrences of one type of accident, such as the number of people poisoned, duration of time for which the pollution lasted (abscissa on the curve), etc., n refers to the overall number of categories or ranges, N i refers to the number of entries in category i, and P(x ≥ j) refers to the probability that the number of occurrences in an accident is ≥j (ordinate on the curve).

The accumulated probabilities of one occurrence can be plotted on a log–log chart. The best fit for a curve of type P = x b can provide the slope of the curve(the value b) and the probability of occurrence of a type of accident with a given severity (characterized as x ≥ m) is (m/n)b times greater than the probability of occurrence of another accident with a different severity (characterized as x ≥ n). Thus, the severity and features of the occurrence can be analyzed using this slope.

Results and Discussion

Distribution of Accidents According to Information Sources

A total of 653 surface water accidents were screened as research subjects and stored in the database. The sources of the 653 accidents are listed in Table 1. The numbers of accidents collected from the China Environment Newspaper, other national newspapers, and the internet were almost equal and accounted for 83 % of the total. Accidents from local newspapers only accounted for 10 %, while accidents from national channels comprised about 60 % of the total. The source distribution illustrated that information regarding surface water accidents was primarily released by national sectors, while the proportion of accidents reported by public and local media was relatively low.

Table 1 Date sources of 653 selected accidents
Full size table

Distribution of Accidents According to Time

The variation in the annual number of the 653 accidents observed in this study is shown in Fig. 2. The annual number of accidents is marked above the columns. Before 2000, there were relatively fewer accidents; therefore, they were incorporated into the abscissa.

Fig. 2
figure 2

Distribution of accidents as a function of time

Full size image

The number of accidents was highest in 2006 and 2007, which is inconsistent with the distribution of water pollution accidents officially registered in China (Xue and Zeng 2011). This difference was likely owing to a lack of concern regarding water pollution accidents before 2006. In December 2005, there was an explosion at the Jilin Petrochemical that led to increased awareness of pollution accidents and a greater demand for safety and environmental risk management (Wo 2009).

Distribution of Accident Location

Provincial Distribution

Guangdong, Jiangsu, Hunan, Zhejiang, Guangxi, and Henan Provinces were the top six locations of the 653 accidents (Fig. 3b), with accidents in these areas accounting for 45 % of the total. The provincial distribution of accidents was generally in accordance with the distribution of the total number of surface water pollution accidents officially reported in 1992–2010 (Committee CEYE 1993–2011) (Fig. 3a). Overall, surface water accidents occurred most frequently in relatively developed areas such as the Yangtze River Delta and the Pearl River Delta, illustrating that the environments of these areas has been overloaded and thus have a high probability of accidents. Additionally, the high rate of reporting reflects the stronger environmental awareness of the populations in these regions, likely owing to their higher income and life quality. These conclusions are similar to the perspectives reported (Hou and Zhang 2009; Yang et al. 2010).

Fig. 3
figure 3

Comparison of provincial distribution of accidents. a Presents the provincial distribution of surface water accidents in 1992–2010 and b presents the provincial distribution of cases in this study

Full size image

Pollution Site Description

Polluted sites were described based on the following four details: regional distribution as a function of population density, type of water carrier, whether more than one administrative region was involved, and whether one drinking water source protection area was involved (Table 2). As pollution belts ran across rural areas as well as suburban and urban areas in some cases, the sum of the proportion of this item is more than 100 %.

Table 2 Distribution of accidents’ locations
Full size table

Suburban areas constitute the largest percentage of incidents (59 %), which might reflect a lack of effective water environment management in these interlaced regions, as well as the intensive industrial layout in Chinese suburbs (CMLR 2008). Among river accidents, there were many more entries along main streams than in tributaries, possibly owing to media attention to the main streams. It should also be noted that there was a high proportion of accidents affecting multi-administrative regions and involving drinking water source protection areas. The former is likely related to transboundary pollution, which is now difficult to properly dispose of in China (Guo and Yang 2009; Kun 2011; Xie et al. 2013), while the latter directly influences the lives of local residents and is therefore of great concern. Accordingly, more attention should be given to transboundary and drinking water-related risks in water environmental risk prevention and mitigation programs.

Materials Involved

Types of Materials

Two general types and 14 specific types of materials were included in the 653 selected cases. The distribution among categories is shown in Table 3. The number of accidents involving sewage was approximately equal to that of accidents polluted by chemicals. In sewage pollution accidents, more than one of the considered substances may exist (for example, “high concentration nutrition” and “some other toxic materials” may be present simultaneously); therefore, the sum of the number of the accidents in all specific classifications was higher than that of individual sewage pollution entries. Only 1 % of the sewage pollution accidents involved unknown materials. Accidents involving high concentrations of nutrients and toxic materials such as phenol and nitrobenzene (but not heavy metals) were both dominant, accounting for 23 and 18 % of the total, respectively, while accidents involving heavy metals accounted for 5 %, which was greater than the proportion for the cases involving high concentration suspensions (3 %).

Table 3 Accidents classified as a function of the substance involved
Full size table

In chemical pollution accidents, benzene compounds accounted for 24 % of the total, followed by heavy metals (10 %). Additionally, 6 and 5 % of the cases involved sulfuric acid compounds and pesticides, respectively.

Quantities of Pollutants

Not all quantities of pollutants were available for all the cases because the information was not reported or the incidents involved sewage, which contains compounds in quantities that are difficult to determine. Indeed, only 14 % of the cases reported quantities, all of which were chemical accidents. Among these, accidents in which 100 to 1000 or 30 to 100 tons accounted for 10 and 18 % of the entries, respectively, while the highest proportions of accidents involved between 10 and 30 tons (34 %) or <10 tons (31 %). Only a few cases involved the release of >1000 tons (Fig. 4a).

Fig. 4
figure 4

Proportion (a) and accumulated probability (b) of accidents with Q amount of pollutants

Full size image

Although it is impossible to calculate the quantities in all the accidents owing to there being little available information, it must be assumed that the samples used were representative, which enables a relative frequency to be estimated for accidents in which there were quantities of pollutants above a certain value. Accumulated frequency–quantities of pollutant graphs were plotted to analyze the probability of different amounts of pollutants (Fig. 4b). The best fit for the curve of type P = Q b gave a b of −0.37 for the discrete points. Based on these findings, the probability of one surface water accident involving 10 tons or more pollutants is 2.34 times (10(−0.37)/100(−0.37) = 100.37) greater than that of an accident involving 100 tons or more.

Loss Distribution During Accidents

The loss caused by the accidents was analyzed based on three factors: number of people involved by the accidents, length of pollution zone, and the duration of the pollution. The latter two indices primarily reflected the span and severity of the accidents.

Human Involvement in Accidents

Human impact of the accidents has been considered, such as that on human health and well-being. Human involvements were classified using three variables: number of deaths, number of people poisoned, and number of people affected.

  1. (1)

    Number of deaths

For 83 % (542 cases) of the accidents analyzed, there was no information available regarding fatalities. There were no deaths reported in the majority of the entries, with fatalities occurring in only 16 cases (2.8 %). Of those that did cause fatalities, there were between 1 and 4 deaths. Over half of these cases (9) involved only one death (Fig. 5a).

Fig. 5
figure 5

Proportion (a) and accumulated probability (b) of accidents with N deaths

Full size image

An accumulated frequency–number of deaths graph was used to analyze accident mortality (Fig. 5b). The graph approximately follows a straight line with a gradient of −1.9, indicating that the probability of an accident with one or more deaths was 79.4 times greater than that of an accident with 10 or more deaths. The same representation for accidents in process plants and during transport of hazardous substances results in a probability of −0.84 (Vilchez et al. 1995), while for natural accidents (earthquakes, floods, etc.) values ranging from −0.4 to −0.7 have been reported (Darbra and Casal 2004; Fryer and Griffiths 1979; Hohenemser et al. 1977, 1983). For explosions, fires, and marine and mining accidents occurring between 1946 and 1975, the same authors obtained F-N plots that could be approximated to straight lines with slopes ranging from −1.0 to −1.1 (Fryer and Griffiths 1979).

  1. (2)

    Number of people poisoned

Only 9.0 % (59 cases) of the 653 accidents analyzed in this study contained information regarding the number of people poisoned. Overall, 8738 individuals were poisoned, with 60 % of the accidents that reported poisoning involving between 1 and 100 people. Only two cases involved poisoning of more than 1000 people (Fig. 6a).

Fig. 6
figure 6

Proportion (a) and accumulated probability (b) of accidents with N people poisoned

Full size image

The accumulated frequency–number of people poisoned is plotted in Fig. 6b. The graph approximately follows a straight line with a gradient of −0.51, indicating that the probability of an accident with 10 or more poisoned people is 3.24-times greater than that of an accident involving poisoning of 100 or more people.

  1. (3)

    Number of people affected

The selected reports were also evaluated for the number of people affected by the adverse consequences of the accidents, including drinking water suspension, panic associated with the incidents, and decreased recreational activities. About 21.1 % (138 cases) of the 653 accidents included information regarding the number of people affected, with the highest amount being 10,000,000. In eight of these 138 accidents, between 1 and 1000 people were affected, while between 1000 and 5000 were affected in 15, between 5000 and 100,000 in 38 cases, and between 100,000 and 500,000 in 55 cases. Twelve cases involved between 500,000 and 1,000,000 people, while 10 cases involved more than 1,000,000 people (Fig. 7a).

Fig. 7
figure 7

Proportion (a) and accumulated probability (b) of accidents with N people affected

Full size image

The accumulated frequency–number of people affected is plotted in Fig. 7b. In the figure, the units on the horizontal axis are thousands of people. The graph approximately follows a straight line with a gradient of −0.38, indicating that the probability of an accident involving 10,000 or more people is 2.4-times greater than that of an accident influencing 100,000 or more people.

Length of Pollution Zone

It was only possible to obtain data describing the length of pollution zones for 13.6 % (89 cases) of the total cases investigated. Of these 89 accidents, 38 involved <1 km in rivers, while 21 involved between 1 and 5 km, 14 between 5 and 10 km, and 14 between 10 and 100 km. Only two involved >100 km (Fig. 8a).

Fig. 8
figure 8

Proportion (a) and accumulated probability (b) of accidents with L length of pollution zone

Full size image

The accumulated frequency–length of pollution zones is plotted in Fig. 8b. The graph approximately follows a straight line with a gradient of −0.46, indicating that the probability of an accident involving 1 km or more in a river is 2.88-times greater than that of an accident involving 10 km or more.

Duration of Pollution

Overall, 13.0 % (85 cases) of the cases contained details regarding the duration of pollution. Of these, 24 involved pollution lasting less than 12 h, 18 between 12 and 24 h, 14 between 24 and 48 h, and 13 between 48 and 96 h, while 16 cases involved pollution lasting for more than 96 h (Fig. 9a).

Fig. 9
figure 9

Proportion (a) and accumulated probability (b) of accidents with T duration of pollution

Full size image

The accumulated frequency–time of pollution lasting graph is shown in Fig. 9b. The graph approximately follows a straight line with a gradient of −0.65, indicating that the probability of an accident lasting for 1 h or longer is 4.47-times greater than that of an accident lasting 10 h or more.

Origin of Accidents

Six different classifications of the process or activity were considered as the general origin of the accidents, process plants, storage plants, transportation, agricultural activity, residential activity, and others. Of the 653 surface water accidents, only 1.4 % (9 cases) was of unknown origin. The accidents of the remaining cases classified as a function of origin are provided in Table 4. The sum of the accidents in this table was more than 644 because some accidents had more than one type of origin.

Table 4 Accidents classified as a function of origin
Full size table

The most important contribution to surface water accidents was process plants (49.0 %), followed by transportation (25.0 %). When combined with storage plants (10.4 %) and agricultural activity (10.4 %), these four types of accidents accounted for more than 90 % of the total. Overall, these findings indicate the importance of process plants and transportation to mitigation of surface water accidents.

The specific origin was also analyzed in the second column of Table 4. About 66 % of the accidents in process plants involved chemical production. Considering the scales of plants (Statistics 2011), the largest contribution to accidents is mini enterprise (21.3 % of the total), followed by small-sized enterprises (15.5 %), which together represent more than 36 % of all cases. When compared with other origins, these proportions indicate that more attention should be paid to mini-sized and small-sized chemical production plants to prevent surface water accidents.

Among accidents associated with transportation, the main contributor was road tankers (14.2 %). However, there was little difference among the number of accidents that originated during shipping and those associated with pipelines, while no cases involved rail tankers.

Causes of Accidents

Causes of the accidents were analyzed based on the origin (transportation and others [plant, agriculture, residents]) in Table 5. Five possible causes were considered, leakage during transportation, intentional illegal release, production accidents, natural factors and others (e.g., intensive layout of enterprises). Overall, 636 (97.4 %) accidents had a reported cause. However, some causes were associated with more than one item; thus, the sum of the entries was >100 %.

Table 5 Accidents classified as a function of the cause
Full size table

The most frequent cause was intentional illegal release (49.8 %), followed by production accidents (28.7 %) and leakage during transportation (22 %). The high frequency of illegal releases with intent might be explained by the lagging environmental awareness and environmental risk perception among the population (Ge et al. 2011; Huang et al. 2010; Schmidt and Wei 2006). Such actions might be mostly driven by the greed of the owners and/or top management of the culprit industrial facilities seeking to save money for technology improvement and waste treatment.

Specific causes of leakage during transportation and production accidents were also determined. The biggest factor associated with leakage during transportation was traffic accidents, accounting for approximately 12 % of the total, followed by equipment failure and operation error. Equipment failures were responsible for almost 69 % of the production accidents, and were more common than operation errors.

Conclusions

653 cases of environmental accidents were screened from publically available data. The severity and other features of the accidents, frequency and quantities of chemicals involved, frequency and number of people poisoned, frequency and number of people affected, frequency and time of pollution lasting, and frequency and length of pollution zone were plotted using historical data to relate the occurrence of a type of accident with a given severity in a particular accident to the relative probability of there being that occurrence in surface water accidents. Additionally, the probabilities of occurrences of various origin and causes were summarized based on the 653 observations.

Some general conclusions can be drawn: (1) The high proportion of accidents involving transboundary and drinking water crises indicates that more attention should be paid to environmental risk prevention and mitigation; (2) The proportion of accidents originating from mini and small-sized chemical plants indicates the need for increased consideration of these two types of enterprises during risk management; (3) As the most frequent cause of pollution was intentional illegal discharge, environmental awareness in China should be improved. These understanding on aquatic environmental accidents might be helpful in risk prevention and mitigation decisions.