Conflict or cooperation? How does precipitation change affect transboundary hydropolitics?

Global climate change affects hydrology and ecology, and aggravates the contradiction between water resources supply and demand, thus leading to transboundary water conflict and cooperation attracting increasing attention. This paper uses the precipitation data sourced from the Global Precipitation Climatology Centre, hydropolitical data collected from the Transboundary Freshwater Disputes Database and, for approximately half a century of socioeconomic indicator for countries, to discuss the relationship between precipitation change and transboundary hydropolitics. As demonstrated by the panel regression results, lower precipitation would lead to more water conflicts and more significant change of precipitation would lead to more water hydropolitical events. This result remains robust after adjustment being made to the defined thresholds of conflict and cooperation. The findings suggest that the countries in a transboundary river ought to avoid conflict and seek more cooperation, considering the uncertain prospect of precipitation changes.


INTRODUCTION
Human society is always accompanied by conflict in the competition for resources. When conflict turns violent, it can have significant consequences for the well-being of humans. Over the last decades, mortality resulting from war and interpersonal violence has reached 0.5-1 million deaths annually (Gleditsch et al. ; Mathers et al. ).
Understanding the causes and solution of conflict has been a major project in the field of science.
Under the context of global environmental change, climate has received increasing recognition as a significant cause of human conflicts (Carleton & Hsiang ). As early as in the enlightenment period, Montesquieu started to give consideration to the relationship between climate and conflict, primarily from the perspective of climate influence on human character (Dell et al. ). In the most recent years, plenty of literature has started to issue warnings against diminishing resources and potential social conflicts arising from climate change, which has a real possibility to alter the supply of a resource and result in disputes over the allocation of it. In some cases, climatic conditions are possible that make people seek violence or cooperation to reach some pre-set goals.
The existing observations and climatological studies have demonstrated that one of the most noticeable impacts by climate change is manifested in the hydrological system (Bates et al. ). Water resources are definitely the primary resources closely associated with the survival of life and the sustainability of civilization. In the past, water gave birth to the civilization history of all countries across the world. Nevertheless, the unbalanced distribution of water resources caused conflicts on a frequent basis (Dimitrov ). For thousands of years, drought was viewed as a significant reason for nomadic tribes to attack farming country (Bai & Kung ). In a modern world, where water remains vitally important, transboundary watershed management has posed a challenge to all countries.
Unbalanced allocation of water resources frequently causes water conflicts. Particularly after the Second World War, many new countries were founded due to decolonization, which fueled the national consciousness, but traditional boundaries have been changed so that the invisible danger of transboundary water conflicts increased rapidly.
A new term named 'Transboundary hydropolitics' is used to describe the political events related to water. Transboundary hydropolitics have been made the focal point of existing academic research, which is particularly significant at a time when global precipitation is changing. As revealed by the Fifth Assessment report of the Intergovernmental Panel on Climate Change (IPCC), the global average surface temperature (GMST) has risen since the late 19th century, and the period 1983-2012 is likely to be the warmest 30 years in the past 800 years (IPCC a, b). Global warming has contributed to a string of events either indirectly or directly, for instance, the changes in precipitation levels, temporal and spatial distribution, melting of mountain glaciers and rising sea levels. Moreover, climate can cause change in precipitation patterns, which in turn, can change runoff in many rivers.
Meanwhile, climate change is likely to stimulate the demand for river water, as the more frequent occurrence of droughts puts more strain on alternative sources of water (Vijay et al. ). The resultant pressure on transboundary basins could increase international tension and heighten the risk of military conflicts. Therefore, climate change has attracted increasing attention in the field of water conflict.
Other creatures intensify intraspecific competition to adapt climate change (Best et al. ; Aitken et al. ), although the same is true of humans in some cases; human society is capable of making a response to crises in a cooperative way. The long-standing civilization created in human society, especially the establishment of the international order over the last century, has opened up the possibility for human beings to address the challenge as a community when encountering climate change (Petersen-Perlman et al. ). There is limited research conducted into collaboration as compared with the plentiful literature on climate and conflict. However, cooperation is definitely a possible and extensive solution to climate change. Therefore, the purpose of this paper is to achieve an empirical establishment of the causal path from precipitation change to transboundary hydropolitics and to investigate whether precipitation change could lead to more conflicts or cooperation. With over 2,000 national-year sample data based on 50 years of climate data and water events used as reference, this paper explores whether absolute and relative changes in precipitation have an effect on the probability of a country being involved in water conflict or water cooperation.
The remainder of this paper consists of four parts. The second section conducts the review and the hypothesis.
Review of the existing literature and hypothesis are mainly about the relationship precipitation change and transboundary hydropolitics. The third section introduces the data and models in the research. The fourth section shows the empirical results.
In the fifth section, the conclusion is drawn and discussed.

REVIEW AND HYPOTHESES
As mentioned in the Introduction, discussions about the association between climate and conflict have a long-standing history, and a fast-paced expansion of literature in this area has occurred over the past few years (Hsiang et al. In recent years, with the increasingly apparent changes in global precipitation and more extreme precipitation events, climate changes have also drawn a great deal of attention from water scholars (Bergholt & Päivi ). Climatic environments have direct and widespread influence on the changes in water resources. For instance, the risk of persistent drought in the Euphrates-Tigris river basin has caused water shortage. Compounded by the impact of resource constraints, growing population, economic growth and diplomatic disputes between coastal countries, global climate change has contributed to persistent drought, which in turn, has exacerbated previous droughts, thus leading to various water conflicts (Caruso ). When water resources are in shortage and insufficient to satisfy the general demand, water conflicts arise (Oftadeh et al. ). The reasons of water conflicts include resource scarcity, transboundary management failure, competition for economic benefits, social and cultural conflicts, and political interests (Burgess et al. ; Bhmelt et al. ), but the effects of climate change are rarely defined.
Precipitation is a hot spot for climate change. There has been a long-standing debate among researchers working across different disciplines around the extent to which precipitation change is blamed for conflict, violence, or political volatility. A variety of different pathways associating climate with these outcomes has been suggested by Hsiang The correlation between precipitation change and water conflicts shows ambiguity. Some evidence suggests that climate factors are only weakly associated with water conflicts and such a correlation is largely reliant on socioeconomic conditions (Wolf et al. a, b). Nevertheless, some of the prior hydropolitics studies have revealed that water shortage is beneficial to the formulation of transboundary river treaties (Tir & Ackerman ). There is also evidence suggesting that the reduction of precipitation could prompt water cooperation among countries (Dinar et al. , ). In fact, the impact of climate change on precipitation can also be a driving force for water cooperation in human societies. For example, Salehyan & Hendrix () argued that water scarcity might have a pacifying effect on armed conflict. Lautenberger & Norris () discussed the initiative to designate international water law (IWL) to be discussed in the context of climate change.
Hemati & Abrishamchi () demonstrated that an integrated, sustainable and efficient water allocation considering changes in water resources due to climate change and change of users' demands is necessary. Fostering cooperation and managing conflict thus become fundamental in transboundary water management (Mgquba & Majozi ). As well, a new study performed by Dinar et al. () indicates that a low variability of precipitation is conducive to enhancing hydropolitical relationship between different countries.
Overall, the issue of water conflict caused by climate change has been covered in the existing literature, but the relationship between precipitation and hydropolitics, including water conflict and water cooperation, has rarely been clearly quantitatively demonstrated. On the one hand, the decrease of precipitation may lead to more water conflicts, which is recognized by most researchers (Haug et  Therefore, we seek to elaborate the relationship existing between precipitation and transboundary hydropolitics, and how precipitation change is examined, involving the absolute amount of precipitation and the annual relative anomaly of precipitation that have effects on international water conflicts and water cooperation. Based on the summary of the scattered achievement in the existing literature and objectives set for this research, the following hypotheses are raised for empirical study.
Hypothesis 1a: There is a certain relationship between absolute change in precipitation and hydropolitics: less precipitation can result in more conflicts.
Hypothesis 1b: There is a certain relationship between absolute change in precipitation and hydropolitics: less precipitation can result in more cooperation.
'Absolute change in precipitation is to present absolute change of annual precipitation, referring to the absolute amount of precipitation'.
Hypothesis 2: There is a certain relationship between relative change in precipitation and hydropolitics: more frequent changes in precipitation can increase hydropolitical events.
'Relative change in precipitation refers to the relative variability of annual precipitation, the absolute value of the difference between the actual precipitation and the multi-year average precipitation, reflecting the stability and frequency of precipitation'

Climate data
The research focuses on the discussion of global climate change and human society on rainfall and hydropolitics.
The precipitation dataset is collected from GPCC (Global Precipitation Climatology Centre) (Ziese ), with the latest version of 'Total Full V2018 (0.5 × 0.5)', gathered from about 67,200 stations around the world that feature a record duration of over one decade (Schneider et al. ).
This dataset provides monthly precipitation statistics from 1891 to 2016 covering the whole globe with 0.5 latitude × 0.5 longitude resolution. Some other relevant studies also cite these data in different versions (Sun et al. ).
Considering that the size of some countries is excessively small, with Python scripts and ArcGIS software 10.2, each 0.5 × 0.5 grid is distributed to its country with a 30 km window. Subsequently, annual precipitation of each grid is calculated, and the annual precipitation (the mean of grids' annual precipitation) is aggregated at the country level.

Conflict and cooperation data
The hydropolitics dataset is sourced from the Transbound- The field 'BAR_SCALE' in the TFDD dataset presents information on whether an event is regarded as a conflict or a cooperation quantitatively. Scores in the field range from À7 to þ7, with À7 indicating the most negative events (the highest level of conflicts, war). For instance, 0 denotes neutral events and þ7 signifies the most positive events (two countries merge to one country voluntarily and peacefully). The threshold in our research design is ±3. Being equal to or more than þ3 indicates that there is a material agreement or support, which is viewed as a cooperation.
On the contrary, being equal to or less than À3 suggests that there is a material hostile action, which is treated as a conflict. Event scores from À2 to þ2 are merely verbal-level issues, as shown in Supplementary Material, Table A1.
Based on the threshold set, two outcome variables are identified as the number of conflict events or cooperation events a country is implicated in for a specific year. As well, concerning robust tests, some researcher have tried to modify the threshold. For more details about the TFDD dataset and the scores we can refer to Yoffe ().

Socioeconomic and other data
With climate factors being excluded, the issue of water source availability is closely related to the constant-rising demand driven by population, agriculture, and economics (Falkenmark  Table 1 indicates the descriptive statistics of these variables.

Model design
The main challenge for the regression is to ensure that variables are relevant and valid (Murray ). Rather than presuming that all confounders are accounted for in a cross-sectional regression, the bulk of recent studies estimate the effect of climate on conflict by using time series variation for identification, usually in a panel data context (Burke et al. ). In our research design, we subdivided the explanatory variable, namely, precipitation index, into absolute change and relative change. However, according to the existing literature, hydropolitics is quite complicated. The first problem is how to quantify the explanatory variable of this paper, that is, hydropolitics (including water conflict and water cooperation); we use the classification standard of Oregon State University to divide hydropolitics into water cooperation and water conflict by setting a threshold. In addition, due to the complex background of hydropolitics, economic, agricultural, political, military and other factors may lead to water conflicts and water cooperation. Therefore, in order to achieve the goal of this paper, the above non-climatic factors must be controlled.  To make the estimation results more robust, hierarchical regression analysis is adopted. First, add various variables continuously, and finally add all variables. Each column from Tables 2-6 is used to represent the regression results when

Descriptive statistics and evidence
In order to verify the hypothesis that precipitation changes will affect hydropolitics, on the basis of maintaining data integrity and reliability, we constructed the above-mentioned national panel data for almost half a century from 1960 to 2008. Table 1 indicates the descriptive statistics of each variable. Then, multi-year (1960-2008) average of annual precipitation, frequency of conflicts or cooperation, and spatial visualization is calculated using ArcGIS, as shown in Figures 1 and 2. Darker shading denotes more precipitation in this country, and larger size of a circle indicates a higher frequency of hydropolitics events (conflict or cooperation) in this country.
As shown in Figures 1 and 2, water conflicts tend to arise in arid areas surrounding the Tropic of Cancer. There is an Note: The p value is in parenthesis (the same as below).
*p < 0.1, **p < 0.05, ***p < 0.01. apparent distribution pattern in Figure 1 that lower precipitation is accompanied by more conflict events. Beyond direct visual findings, statistical evidence suggests this sort of space-coupling phenomenon. Using data from 1948 to 2008, it shows that the Pearson correlation coefficient between precipitation and conflict was À0.13, which was significant at the 0.001 level. It demonstrates that annual precipitation has a significant negative association with how frequently conflicts arise. In Figure 2, nevertheless, the pattern is more confusing and the distribution of precipitation and cooperation events seems to be randomized.

Baseline results
To strictly and reliably determine further the causality, we car-  Note: The p value is in parenthesis (the same as below).

model are indicated in
As shown in Table 2, under different control variables, the coefficients of LNPrec are significantly negative at all times, mostly at p < 0.001 level, suggesting that less precipitation tends to trigger more conflict hydropolitical events for a country. Therefore, Hypothesis 1a is supported by empirical results.
According to column 6, which contains all control variables, one standard deviation reduction of LNPrec will increase conflict by an extra 0.39 times, which is about Note: L. means the variable is first-order lagged. The p value is in parentheses (the same as below).
*p < 0.1, **p < 0.05, ***p < 0.01. 86.7% of one standard deviation of Conf3. In general, the results shown in Table 2  As revealed in Table 3, the coefficients of LNPrec are invariably negative but insignificant. It is thus insufficient to suggest that precipitation has exerted influence on the transboundary water cooperation. Therefore, Hypothesis 1b is not held by empirical results.
Then, rooted in Chinese classical political wisdom, the effects of relative changes of precipitation on transboundary hydropolitics are investigated in this paper, under the insights of Hendrix & Salehyan (). In Table 4, RVPrec replaces LNPrec as the independent variable, and other controlled sets are the same as in Tables 2 and 3. implying that more extreme deviations in rainfall will prompt more cooperation. As shown in column 3, one standard deviation increase in RVPrec from the mean value is related to a quarterly increase of cooperation in the year.
The estimation of control variables are excluded from Table 4, but there is more information worth reporting.
Former socialist countries remain positive and significant at p < 0.05 level, suggesting that they have a tendency to have more cooperation than other countries. Moreover, the agriculture ratio produces a significant positive effect on cooperation. A higher agricultural land use ratio is more likely for a country to reach more water cooperation.
In contrast, the efficiency of military ratio turns out to be significantly negative, indicating that transboundary water cooperation is even less likely to be reached in militaryfocused countries. For Conf3, the estimated coefficients are positive and significant, especially in columns 6 and 7, which both contain a majority of the control variables, indicating that more extreme deviations in rainfall will result in more conflicts. Therefore, Hypothesis 2 is initially verified.
In general, a RVPrec increase of one standard deviation from the mean value is associated with a 28.9% increase of conflicts in the year. Military ratio and life expectation are also significant factors that cause conflicts, the effects of which show similarity to the results shown in Table 2.

Robust test
In studies of causal inference, the main consideration is given to the endogeneity. In this research, on the one hand, panel data are competent to mitigate the issue of missing variables, to some extent. On the other hand, the association between precipitation and human factors is complicated and difficult to completely understand at present (Schaller et al. ). According to the IPCC AR5 report (IPCC a, b), human activities are speculated to have an effect on precipitation change in the second half of the 20th century. However, the corresponding direct impact relationship remains unclear, and the lack of clarity remains apparent (Hegerl et al. ; IPCC a, b). Therefore, the model also seeks to remove the mutual causal concerns over the core explanatory variables and the dependent variables. Therefore, Hypothesis 2 is verified.

Hysteresis effect
The results of the general model with no lag contradict the null hypothesis that absolute or relative changes in precipi- Regression results with lagged independent variable are indicated in Table 6. Columns 1-4 represent the association between LNPrec of one year lagged and hydropolitics, conflict or cooperation. Columns 5-8 show the independent changes to be lagged for RVPrec. Likewise, all control variables are contained in any column as in column 6 of Table 6.
As indicated in Overall, we believe that this study will provoke more interest in the relationship between climate change and hydropolitics. As the IPCC warns, human demand for water is growing, and water usage has been on the increase worldwide by about 1% annually since the 1980s. However, the amount of freshwater placed at the disposal of everyone around the world has been in decline. Water scarcity continues to deteriorate and water conflicts are manifested in various forms due to population growth, exacerbated water pollution, lack of planning and management of transboundary and other shared water, and inefficiency in water supply and distribution systems (Dinar et al. ; Carleton & Hsiang ). Encountering uncertain climate change prospects, countries ought to pay more attention to water cooperation and take the proper approach to resolve disputes among themselves and prevent conflicts from escalation. The key to sustainable development lies in the establishment of fair and effective water resources allocation mechanisms through cooperation in various forms.

ACKNOWLEDGEMENTS
The study was funded by the Chinese Academy of Sciences