The environment comes later: when and how environmental considerations are included in transboundary water agreements

ABSTRACT

Until recently, most transboundary water agreements largely focused on utilization, frequently leading to environmental impacts. We examine under which circumstances are environmental considerations included in transboundary water agreements, how are they included and whether they address environmental effects of previous agreements. We analyse 75 treaties, many of which occur within (sub)basins with previously established water agreements. Human water stress, upstream large dams, OECD membership and previous water treaties (particularly those concerning utilization) are the most prevalent conditions where environmental degradation is considered. Our results also suggest that aridity incentivizes negotiating relationships to develop, thereby facilitating environmental water cooperation.

KEYWORDS:

• transboundary river basins
• water resources
• international water treaties
• environmental agreements

View correction statement:

Correction

Introduction

Nations may have an incentive to overexploit transboundary water resources because the benefits of use accrue to one nation while the cost due to the use may be passed onto other riparian nations (hereafter referred to as ‘riparians’; Giordano, 2003). Without scale-appropriate, effective governance institutions, natural resources and the environment are in jeopardy (Dietz et al., 2003). Institutions help to mitigate the over-exploitation of common-pool resources such as water (Anderies & Janssen, 2013).

While institutions, in the form of international agreements and river basin organizations, might help to protect the over-exploitation of resources, the level of protection depends on the content of the agreements. Historically, the focus of water agreements has tended to concern the utilization of water resources, such as water allocation and the construction of large-scale water infrastructure (e.g., dams) (Petersen-Perlman & Wolf, 2016). Until recently water quality and the environment were rarely considered in international water agreements (Giordano et al., 2014). Still, many basins lack any type of institution governing water quality or environmental flows. In Giordano’s (2003) survey of 227 international water agreements, only 62 (27%) of the agreements reviewed were found to contain references to water quality, and only 38 international river basins (or parts of those basins) were covered by a water quality agreement. Most international water institutions have started to consider water quality in recent decades, particularly in North America and Europe (De Stefano et al., 2010), with water quality provisions historically only playing a minor role in co-riparian agreements (Giordano & Wolf, 2003). Even in basins where institutions considering water quality or the environment exist, the lack of specific, substantive language and full participation among basin riparian countries have limited their effectiveness (Shmueli, 1999). This might prove to be troublesome for hydropolitical relations as water quality issues may lead to future water-related conflict (Haftendorn, 2000; Wolf et al., 2003).

This focus on utilization can lead to externalities affecting other basin characteristics such as groundwater replenishment, water quality, aquatic ecosystems and environmental flow provisions (Zeitoun et al., 2020). It has been suggested that present methods of water allocation, dam and turbine operation, and inflexible water delivery priorities, many of which have been historically present in international water agreements, might worsen water quality problems (Gleick, 1993), with widespread ecological and societal ramifications (Kaneti, 2020). Questions of spatial scale of agreements can be manipulated (Feitelson & Fischhendler, 2009), in that agreements can be signed between two parties that may adversely affect a third party. An example of this would be the agreement between the Lao government and Thai company CH over construction of the Xayaburi Dam, which may negatively impact fisheries and agriculture (Hensengerth, 2015).

Analyses of treaties/agreements related to water quality have been performed before (e.g., Giordano, 2003; Giordano et al., 2014; Shmueli, 1999). But there is a lack of research that explains why and how water quality and environmental aspects are incorporated into some international water agreements and not in others, and whether agreements that do not consider (or regulate) water quality and environmental issues lead to further agreements or modifications to manage environmental effects brought forth. As environmental aspects gain greater prominence in many countries, this paper asks to what extent a similar trend can be recognized in international water agreements.

Specifically, we address three questions:

• Under which conditions are water quality and the environment considered in water agreements and river basin organizations, and in which cases is the environment an initial driving force?

• Under which circumstances do agreements have adverse environmental effects that lead to amendments designed to address such effects?

• How are water quality and environmental issues addressed within these transboundary water agreements?

What brings countries to consider the environment? A brief literature review

It is difficult for parties to enter into international agreements concerning water – parties must overcome perceived risks that cooperation may bring, jeopardizing something of value or the political future of individual policymakers (Subramanian et al., 2012). Part of the problem is that transboundary waters is a ‘fugitive’ common-property resource – the benefits of using the resource (e.g., irrigation, hydropower) may accrue to one party while the costs (e.g., river flow reductions or alterations) are passed onto other riparian parties (Giordano, 2003). This trait may create disincentives for cooperation. Water quality issues are therefore often relegated to the sub-basin or national scale due to high political transaction costs of reaching transboundary agreements (Ashkenazi, 2004). The political intricacies of negotiation over the use of shared resources may prove to be too difficult to overcome in international basins, as riparians may develop or alter shared waters unilaterally (Wolf, 2007), resulting in detrimental ecological impacts. This was evident when dams built in South Africa’s portion of the Incomati River Basin reduced freshwater flows in the Incomati estuary, thereby increasing salinity (Wolf et al., 2005). Free riding may thus substantially increase pollution in international rivers compared with those of domestic rivers (Sigman, 2002).

The presence of institutions between riparian parties may offset the incentives for unilaterally exploiting shared water and foster better distribution of benefits. International environmental agreements may be designed to internalize negative externalities and reduce transaction costs associated with cooperation (Fischhendler et al., 2011). These agreements may lead to positive-sum outcomes for all parties involved, allowing for maximum use of the resource (Drieschova et al., 2009). Yet, such seemingly ‘win–win’ agreements may lead to a ‘win–win–lose’ dynamic, with the environment paying a price for this increasing utilization of the water resources. This outcome manifests when agreements to utilize most or all of the water resources leave little to no water for environmental flows, thereby leading to basin closures, such as in the cases of the Colorado (Molle, 2008) and Jordan (Venot et al., 2008) rivers.

Over time, the content of international agreements governing international waters has evolved as the number of agreements (and areal coverage) increased. Certain basin characteristics may allow parties to negotiate agreements more successfully (with ‘success’ being defined as signing an agreement). While past international water agreements tended to focus on allocation and hydropower (Wolf, 1999), water quality and the environment have become a more common foci in recent agreements (Giordano et al., 2014). Language concerning the environment was not explicitly mentioned in international water treaties until the 1960s. But environmental concerns became much more recognized in international governance frameworks in general starting in the 1970s and 1980s, including in the water field (Figure 1) (Roberts et al., 2004).

Figure 1. Water treaties that have an environmental component by decade through original agreements (blue), and amendments or protocols to original agreements (red).

Source: Oregon State University Transboundary Freshwater Dispute Database (TFDD) (2009).

In the 20th century, international water law scholars recognized the need for environmental protection in legal frameworks. While most early laws governing international watercourses focused on navigational uses until the late 19th century, the focus of international water law shifted to utilization as water users increasingly relied upon lakes and rivers (Salman, 2007). The 1966 International Law Association’s Helsinki Rules established the principle of equitable and reasonable utilization of international rivers as a central tenet of international law. In addition to the inclusion of equitable and reasonable utilization, the 1992 United Nations Economic Commission for Europe Convention on the Protection and Use of Transboundary Watercourses and International Lakes also had objectives that focused on protecting and managing transboundary waters in an ecologically sound manner and reducing, preventing, and controlling transboundary pollution (McIntyre, 2004). The 1997 United Nations Convention on the Law of the Non-Navigational Uses of International Watercourses, which entered into force in 2014, also includes a provision where all States should take ‘all appropriate measures to prevent significant harm’ to other basin states (Paisley, 2018).

To motivate parties to enter into environmental agreements, certain incentives must be present where these political intricacies seem more surmountable. In a study of seven selected international basin accords (Shmueli, 1999) four major forces were identified that influenced the extent to which pollution issues were encompassed: (1) flashpoints (e.g., spills, fish die-offs, disease outbreak, drought/flooding); (2) financial capacities (the ability of governments to finance water quality projects in their own and neighbouring basins); (3) globalization/regionalization of industry (where offshore firms increase pressure on water sources); and (4) policy windows (when advocates have a chance to push or draw more attention to problems; Kingdon, 2003). Other motivators that may appeal to parties in an agreement include non-environmental costs, such as breaking social norms and conventions or being excluded from trade agreements (Hoel & Schneider, 1997).

Previous work has analysed environmental provisions in international water agreements, including how water quality is addressed in selected international basin accords (Shmueli, 1999), and how agreements address water quality (Giordano, 2003; Giordano et al., 2014). The content of agreements addressing water quality issues is quite variable. Nearly half of the ‘water quality’ treaties identified by Giordano (2003) had indefinite, non-specific commitments. Water quality and the environment are now the most common foci in international water agreements. Since 1990 three quarters of all agreements mention the environment in some form (Giordano et al., 2014). Yet it is unclear whether the aforementioned forces are present globally.

It appears that the inclusion of the environment and the content of environmental provisions in transboundary water agreements varies as a function of the precipitation regime and basin utilization ratio (Shmueli, 1999). In closed or closing¹ basins increasing water withdrawals degrade water quality and the environment, in some cases rendering water unusable for downstream uses. Such overdevelopment outstrips basin resources and ecosystem resilience (Molle et al., 2010), possibly leading to an agreement to manage the basin in a more environmentally friendly manner, or to modifications of existing agreements. In more humid basins, however, agricultural runoff, industry and hydroelectricity generated from dams are the main factors that lead to water quality degradation (Diamantini et al., 2018). Thereby in such regions the environment may be more likely to be the focal point of initial agreements.

Several variables may influence the inclusion and content of environmental clauses and agreements. The presence of water-related institutions appears to be strongly influential in opening the door to possibly addressing environmental degradation within international frameworks (Drieschova et al., 2009; Fischhendler et al., 2011). Following the environmental Kuznets curve rationale, a certain level of economic prosperity can also be hypothesized to be an important factor for considering the environment in international treaties. Other variables that can influence whether and how environmental clauses and agreements are considered include the level of human, environmental, and agricultural water stress (Glenn et al., 2017), as well as forms of previous utilization of the basin through, for example, the construction of dams, industrial use and exploitation of fisheries.

Conceptual framework

Based on the brief review four processes can be hypothesized for the evolution of environmental facets in transboundary water agreements, differentiated by the state of the basin, which in turn is a function of the climatic zone. In humid climates, where extractive uses are limited relative to the river flow and hence basins are not closed, two sequences can be identified:

• Degradation due to pollution or the damming of a river leads to an environmental agreement. One such example is the International Commission for the Protection of Lake Geneva Waters, created in 1960 to maintain or restore water quality (Ferrier, 2000). Another example is the 1990 agreement between Germany, Czechoslovakia and the European Economic Community addressing pollution in the Elbe River Basin (Lindemann, 2006).

• An agreement leads to water utilization, such as the building of a dam (or dams), or utilization leads to an agreement that results in deleterious ecological effects, which in turn spurs an environmental agreement, or modification of the original agreement, to address the environmental externalities. As an example, Rhine River Basin countries agreed to develop and implement the ‘Salmon 2000’ programme to help the salmon population rebound after dam construction within the basin (Mostert, 2009).

In drier climates where much of the water is extracted and hence basins are stressed or closed, there appears to be a more complicated sequence:

• Water extraction leads to water stress, possibly even basin closure, creating deleterious effects on the environment. This leads to an environmental agreement. OR

• Water extraction leads to an agreement focusing on water quantity. The agreement results in deleterious water quality and ecological effects, ultimately leading to an agreement or an environmental amendment to the agreement. This situation is evident in the North American Colorado River Basin, where the basin was over-allocated in previous intranational and international agreements, and both countries agreed on a release of water to mimic seasonal flooding in 2014 (Bark et al., 2016).

However, these sequences are mediated by other variables. The power of the environmental lobbies within the different countries, and environmental awareness allow environmental considerations to become more prominent in the internal discourse that structures the parties’ stance in the international fora (Putnam, 1988). These positions are also a function of the relations between the parties on other issues (upper politics) and the water relations between the parties. In particular, the existence of transboundary water institutions, established in prior utilization agreements, is likely to advance the inclusion of environmental considerations.

These different paths to environmental clauses in existing transboundary water agreements, or new environmental agreements to address the externalities generated by transboundary water agreements are presented in Figure 2.

Figure 2. Paths to environmental clauses or environmental agreements in stressed and unstressed basins.

Methodology

We analyse the introduction of environmental considerations in international water agreements using Oregon State University’s Transboundary Freshwater Dispute Database (TFDD) (2009), which contains information on more than 600 international, freshwater-related agreements from the period 1820–2007. Treaties included in this study are primary agreements, replacements of primary agreements, and protocols and/or amendments to a primary agreement. Documents categorized as ‘not a treaty’, ‘semi-international treaty (i.e., an agreement between a nation and a sub-national government)’, or documents that do not deal with water as a consumable resource were excluded. Treaties that were global or regional in their geographical scope were also excluded as they only provided general concepts for management of shared water resources rather than specific rules for action.

From this subset we identified the treaties that considered environmental or water quality issues (see Table S1 in the supplemental data online for a description of issues covered). Treaties whose only environmental content was general language about protecting the environment were eliminated. We also eliminated treaties that did not focus on freshwater issues and treaties that were between colonial powers. Treaties were also eliminated if they were part of the same treaty lineage as identified by Giordano et al. (2014). A treaty is part of a lineage if the treaty was coded by Giordano et al. as an amendment or protocol to a previous treaty with environmental content. If treaty lineages contained a treaty originally focused on utilization followed by later amendments with environmental facets, that lineage was included.

For the 75 treaty lineages that were identified as including a significant environmental component (see Table S1 in the supplemental data online), countries that signed the treaty were then coded by physical and social characteristics described below.

All treaties were categorized by three physical characteristics using data from the Transboundary Waters Assessment Programme: environmental water stress, human water stress and agricultural water stress. In this context, environmental water stress is defined as alterations of flow regimes from natural flow conditions. Human water stress is classified as water availability per capita and water use compared with the available amount of water, and agricultural water stress is the imbalance between water use and water availability (UNEP-DHI & UNEP, 2016). The metrics were calculated by the Transboundary Waters Assessment Programme for each basin country unit in a transboundary river basin. A basin country unit is a country’s drainage area within a transboundary basin. For example, the Tagus River Basin has two basin country units – the portion of Spain in the Tagus drainage area, and the portion of Portugal within the Tagus drainage area.

Each basin country unit was rated on a level of stress from 1 (very low) to 5 (very high) on environmental, human, and agricultural water stress. In this analysis, average scores were calculated for each unit that participated in each treaty for all three physical characteristics. For instance, Argentina’s La Plata basin country unit score for environmental water stress was averaged with Uruguay’s La Plata basin country unit score when calculating the environmental water stress score for the 1961 treaty between Argentina and Uruguay concerning the Uruguay River in the La Plata River Basin. If a treaty covered multiple river basins, then all units that were signatories to the treaty were included in the average score for the treaty for each characteristic. Certain basin country units did not have scores assigned for the three physical characteristics, mostly due to their small size. In that case the unit was excluded from calculating the average treaty score.

The presence of large dams upstream of one of the riparian countries that entered into a treaty was also examined. We used the Global Reservoir and Dam Database (GRanD) for locations and dates in which dams were constructed. We then coded for whether a dam had been built upstream to one of the riparian countries that had signed a treaty (Lehner et al., 2011).

Events that were considered conflictual could lead parties to agree on a treaty. Crises before a treaty was signed were analysed using a list of events from Oregon State University’s TFDD. The International Water Events Database provides over 6400 historical events from international water relations from 1948 to 2008 and codes events based on level of conflict on a 15-point scale, from −7 representing the most negative events to +7 representing the most positive events. Events were considered as worthy of inclusion if it was coded as −3 or beyond, therefore including diplomatic–economic hostile actions, political–military hostile actions, small scale military acts, extensive war acts causing deaths, dislocation or high strategic cost, and formal declarations of war. Crises were also coded for whether the events themselves were related to water quality and/or the environment. Events were not included if there was no information about why the event occurred or if the event dealt with navigation or border issues (TFDD, 2009).

Each treaty with environmental components was examined to determine whether water treaties with environmental components were more likely to occur if previous treaties were signed over water between riparians. Using data from the TFDD, each treaty was coded for whether the same riparians signed a previous treaty over the same basin(s) (or sub-basin(s)) involving water allocation, hydropower/water infrastructure, or any other issue.

Finally, each treaty with environmental components was examined to determine whether the countries belong to the Organization for Economic Co-operation and Development (OECD). This was done in part to determine whether the environmental Kuznets curve applied within these cases.

An additional group of 25 treaties in the TFDD that did not include environmental provisions were then selected at random using a random number generator to serve as a reference group in the analysis, and thus allowed us to compare the group of treaties that do include environmental provisions to the group that has the potential to include future environmental provisions but do not.

Transboundary water treaties can have several key parameters: scope, substantive rules governing rights and obligations, procedural rules, institutional rules, dispute settlement mechanisms, and implementation and compliance (Wouters et al., 2005). For the treaties with environmental components, the scope, rules, procedural and institutional requirements, and implementation and compliance was examined. This fell into five categories of cooperation: (1) an expressed intention to prevent environmental degradation (e.g., ‘Parties will take measures …’; ‘Parties will take steps …’; and/or “Parties will establish a joint programme); (2) a commitment to inform other parties of potential environmental degradation (e.g., ‘Parties agree to submit projects that may potentially harm flora and fauna’); (3) establishing a commission to address environmental issues (e.g., ‘Pollution shall be dealt with by a commission …’, ‘A commission will create rules …’, ‘A commission will harmonize rules and regulations’); (4) parties agree to execute joint studies, monitoring and/or will craft recommendations; and (5) specific rules or limits of contamination. Treaty language within the 75 treaty lineage groups was examined to accomplish this.

The analysis of treaty language was also used for determining whether environmental issues were a driving force for the treaty. Other factors considered were whether an environmental issue was named and addressed in the treaty title, specific rules or limits of contamination were listed within the treaty, whether the environmental issue was listed in the beginning of the treaty, and whether the treaty created a commission listed to tackle environmental degradation.

Data analysis

The qualitative comparative analysis method was used for analysis. The method allows for analysing multiple conditions without assigning weight to each of them and allowing to calibrate partial membership in sets of each condition without dismissing core set principles (Ragin, 2009). Qualitative comparative analysis is an approach that aims to reveal the minimum conditions, or combinations thereof, that bring about a particular outcome. With more cases, qualitative comparative analysis moves towards identifying multiple conditions of causes, as opposed to the causes of effects, aiming for understanding how outcomes come about (Vis, 2012). It should also be noted that qualitative comparative analysis stresses the concept of equifinality, meaning that multiple configurations can be equally effective in achieving an outcome (Fiss, 2007). Three analyses were completed using qualitative comparative analysis: (1) conditions necessary for environmental components within a treaty; (2) differences in conditions within arid versus humid basins; and (3) differences in types of environmental content within arid versus humid basins.

The crisp set qualitative comparative analysis method requires defining the outcome and the causal conditions into crisp set values of 0 or 1 (Ragin, 2009). The value of 1 indicates full membership within a set, while 0 means no set membership. Consistent calibration rules were applied to each of the conditions, where data were assigned membership scores. Three different anchors need to be set for calibration: a value for full non-membership, full-membership, and a crossover point (Ragin, 2009). The crossover point represents cases that that would be neither members nor non-members, that is, the point of maximum ambiguity (Stockemer, 2013).

In the first analysis the first three conditions (climate type, human water stress and basin closure) were calculated as fuzzy set values. The averages for the basins (as described in the previous section) were divided by 5 to fit the 0 to 1 scale required for crisp set qualitative comparative analysis. Therefore, averages falling below 0.5 were classified as non-members, that is, low stress basins, for those conditions, and averages at or above 0.5 were classified as members, that is, high stress basins. The remaining conditions for all analyses were coded as 0 for those cases with non-membership and 1 for cases with membership for each condition.

After calibration, the conditions were analysed to determine whether they are necessary for causing the outcome. In other words, we examined each condition to determine its presence or absence in all cases in which the outcome was present or absent (Ragin, 2009). Any condition is considered as necessary if the consistency score exceeds the score of 0.9, meaning that if the condition is present or absent in 90% or more of the cases, that condition is identified as necessary. The consistency score increases as more cases meet this condition.

To determine sufficient conditions, that is, condition(s) that are enough but are not required for the outcome to occur, a truth table must be constructed, prepared and analysed (Ragin, 2009). The first step, constructing the truth table, consists of all logically possible causal combinations of the conditions, with each row consisting of a unique combination of conditions (Tóth et al., 2015), not all of which are observed in the dataset. After this, Boolean minimization occurs, which reduces the number of conditions based on whether said conditions appear irrelevant to the desired logical outcome (Rihoux & De Meur, 2009).

The next step involves reducing the truth table’s rows to the ones identified as relevant through selecting a frequency threshold for how many cases appear within each row. In this study, a frequency threshold of three was chosen, meaning configurations with two or fewer observations were classified as remainders. For the rows that still exist, a minimum level of consistency score of 0.85 was set for both analyses, thereby establishing configurations as either sufficient or not sufficient for achieving the outcome. The solution coverage is then determined to the proportion of the configurations that explain the presence of a treaty with environmental characteristics (Rihoux & De Meur, 2009).

Results

In this section the results of the analysis of which conditions lead to water quality or environmental provisions in treaties are first presented. This is followed by an analysis of which cases the environment was a driving force and which cases considered the environment as a result of a past transboundary agreement. The next subsection analyses the environmental content of the identified treaty lineages. The section concludes with an identification of the circumstances under which agreements have adverse environmental effects, leading to amendments to address those effects.

Under which conditions are water quality and the environment considered in water institutions (i.e., agreements and river basin organizations), and in which cases is the environment an initial driving force?

Physical water stress

Table 1 compares the scores for environmental, human, and agricultural water stress in our subset of basins containing the 75 environmental agreements (see Table S2 in the supplemental data online) with all basins where a score for environmental, human and agricultural water stress was calculated in UNEP-DHI & UNEP (2016), and our dataset of 25 randomly selected treaties (referred to as the counterfactual dataset). When comparing the basins in our dataset with the average score for all transboundary basins, the scores reveal that basins with higher water stress (more so for human and environmental water stress) are more likely to have treaties with environmental components. It should be noted that few basins with high water stress in all categories (the exceptions including the Colorado, Indus, Jordan and Tigris–Euphrates) were represented in the dataset of 75 environmental agreements.

Table 1. Physical water stress in basins with treaties containing significant environmental components compared with physical water stress in basins in the counterfactual dataset and the average score for all transboundary basins where a water stress score was computed.

	Basins in our dataset	Basins in the counterfactual dataset	Average score for all transboundary basins
Environmental water stress	2.223	2.109	1.948
Human water stress	2.517	2.204	1.968
Agricultural water stress	2.346	2.424	2.133

Other variables

Almost every treaty lineage in our dataset had a large upstream dam that was built before a treaty was present, at a much higher rate than within the counterfactual dataset. Only in one basin, the Danube, had a crisis that was directly about the environment (the health of the Drava River) that was caused by externalities imposed by one country onto another, thereby leading to a treaty with significant environmental components. Other basins with crises leading to environmental components within treaties had prior treaties focusing on water allocation, diversion, or dam construction.

The treaty lineages in our dataset were more likely to have previous utilization treaties signed than in the counterfactual dataset, suggesting that previous cooperation leads to a higher likelihood of future environmental cooperation. Parties in at least 11 treaty basins agreed to future environmental provisions after their initial environmental cooperation.

Analysis of necessary conditions

Table 2 shows that none of the consistency or coverage scores for a condition’s presence or absence were above 0.90. Therefore, no condition was deemed necessary for environmental degradation to be addressed within a treaty. The ‘presence of upstream dams’ condition had the closest consistency score with 0.881. Next, we tested for sufficient configurations of conditions through reducing the truth table.

Table 2. Consistency and coverage scores for the conditions analysed to be necessary for the outcome of having a treaty with environmental components.

	Environmental clauses/agreement		No environmental clauses
Condition	Consistency	Coverage	Consistency	Coverage
Environmental Water Stress	0.184	0.667	0.28	0.333
~Environmental Water Stress	0.816	0.775	0.72	0.225
Human Water Stress	0.553	0.792	0.44	0.208
~Human Water Stress	0.447	0.708	0.56	0.252
Agricultural Water Stress	0.224	0.654	0.36	0.346
~Agricultural Water Stress	0.776	0.787	0.64	0.213
Water Stress	0.605	0.767	0.56	0.233
~Water Stress	0.395	0.732	0.44	0.268
Upstream Large Dams	0.881	0.848	0.48	0.152
~Upstream Large Dams	0.118	0.409	0.52	0.591
Previous Utilization Treaty	0.342	0.788	0.28	0.212
~Previous Utilization Treaty	0.658	0.735	0.72	0.265
Crises	0.171	0.813	0.12	0.188
~Crises	0.829	0.741	0.88	0.259
OECD	0.461	0.761	0.44	0.239
~OECD	0.539	0.745	0.56	0.255
Previous Water Treaty	0.671	0.785	0.56	0.215
~Previous Water Treaty	0.329	0.694	0.44	0.306

Seven configurations of conditions were identified as leading to a water treaty (see Table S3 in the supplemental data online). Each of the configurations listed in Table S3 online had the presence of upstream large dams. Other conditions were present in at least three of the listed configurations.

For the counterfactual set of non-environmental treaties, no configuration had a raw consistency above 0.85. This is partly explained with the relative lack of counterfactual cases as compared with the environmental treaty set (25 versus 75), but could also be explained by the fact that the conditions tested were specifically chosen for analysing the presence of environmental components of treaties, not the presence of treaties alone.

The truth table for humid treaty basins shows combinations of configurations that appeared most frequently (see Table S4 in the supplemental data online). Overall, not many strong patterns emerged. There was not a high enough frequency of cases that shared a combination of configurations in arid basins that met our criteria for including it in the truth table.

Most environmental treaties have been signed since 1970, with 1991–2000 being the decade with the highest number of environmental treaties. The largest number of environmental treaties have been signed in Northern Europe (i.e., excluding basins flowing into the Mediterranean Sea), more than twice as many as any other region (North America, Sub-Saharan Africa, South America, Mediterranean and Middle East/North Africa, Asia, and Oceania) which correlates with that region’s higher standard of living and high frequency of previous water treaties signed between riparians.

Our analysis suggests that environmental considerations are rarely the main driver of creating these treaties. Many treaties include some passing references to the environment and do not contain the regulatory teeth necessary for substantial environmental protection. Of the 75 treaty lineages, 32 were identified as having environmental issues as one of the (not the) driving forces within the initial treaties addressing environmental degradation. In four other treaty lineages the initial treaty addressed environmental degradation, but the environment was not identified as a driving force for subsequent amendments.

Differences between arid and humid basins

Our environmental treaties dataset was then analysed to determine appreciable differences in the presence or absence of certain conditions, or the combinations of conditions, between arid (environmentally water stressed) and humid (not environmentally water stressed) basins. Table 3 shows the consistency and coverage scores of the variables. As one might expect, agricultural water stress is closely coupled with environmentally water stressed basins (consistency > 0.9), and the lack of agricultural water stress is closely coupled with basins with no environmental water stress (consistency > 0.9; coverage > 0.9). The presence of upstream large dams was a necessary condition in arid basins (consistency > 0.9; coverage < 0.9) but not in humid basins.

Table 3. Consistency and coverage scores for the conditions analysed to be necessary within humid and arid basins.

	Environmental water stress (‘arid’ basins)		No environmental water stress (‘humid’ basins)
Condition	Consistency	Coverage	Consistency	Coverage
Human Water Stress	0.846	0.268	0.484	0.732
~Human Water Stress	0.154	0.059	0.516	0.941
Agricultural Water Stress	1	0.813	0.048	0.188
~Agricultural Water Stress	0	0	0.952	1
Upstream Large Dams	0.923	0.182	0.871	0.818
~Upstream Large Dams	0.077	0.111	0.129	0.889
Previous Utilization Treaty	0.538	0.280	0.290	0.720
~Previous Utilization Treaty	0.462	0.120	0.710	0.880
Crises	0.308	0.333	0.129	0.667
~Crises	0.692	0.143	0.871	0.857
OECD	0.385	0.143	0.484	0.857
~OECD	0.615	0.200	0.516	0.800
Previous Water Treaty	0.769	0.200	0.645	0.800
~Previous Water Treaty	0.231	0.120	0.355	0.880

The truth table for humid treaty basins shows combinations of configurations that appeared most frequently (see Table S4 in the supplemental data online). Overall, not many strong patterns emerged. There was not a high enough frequency of cases that shared a combination of configurations in arid basins that met our criteria for including it in the truth table. Treaties in arid basins had higher consistency scores for previous water treaties and previous utilization treaties than treaties in humid basins.

Under which circumstances do agreements have adverse environmental effects that lead to amendments designed to address such effects?

Amendments to agreements that address adverse environmental effects appear to be rare – only three instances were found in our dataset. This includes agreements such as Minute 241 in the Colorado River Basin, which addresses water quality issues related to depleted water quantities. The quality issues arguably result from previous water allocation-centred agreements. The other two amendments, which amended treaties on the Mekong and Rio Grande basins, address environmental effects due to previously agreed-upon water infrastructure projects.

How are water quality and environmental issues addressed within these transboundary water agreements?

The type of environmental content within the 75 treaty lineages varied in specificity (Figure 3). Only 18 of the treaties had language specifically addressing an environmental issue. Nineteen treaties had language establishing a commission that would collaboratively address environmental issues. Many more treaties had more general language intending to prevent degradation, informing other parties about environmental degradation, and/or committing to studies, monitoring, and/or recommendations. Treaties in (sub-)basins with and without environmental stress had negligible differences in the breakdown of environmental content, with a slightly greater frequency of studies and specifics appearing in the language of treaties in environmentally water-stressed basins.

Figure 3. Types of environmental content within our treaty dataset.

It should be mentioned that these categories are non-exclusive; some treaty lineages exhibited multiple types of content. It was also observed that some treaty lineages increased the specificity of environmental action through subsequent amendments. An example of this is the 1956 treaty, ‘Convention between the Federal Republic of Germany, the French Republic, and the Grand-Duchy of Luxembourg on the subject of Moselle canalization’, that stated that the states would take measures to protect its waters from pollution, and then established an international commission to protect the Moselle against pollution in 1961.

An analysis of the consistency and coverage scores shows that not including studies, monitoring, and/or recommendations within treaties is a necessary condition for treaties in humid basins. Further, not including a treaty clause where parties inform each other on possible degradation is a necessary condition for treaties in both humid and arid basins (consistency > 0.85).

Discussion

Our conceptual framework suggested four scenarios that would result in environmental considerations in transboundary water agreements. For the first scenario for humid basins (degradation due to pollution or the damming of a river leads to an environmental agreement), the results provided mixed evidence. The presence of upstream large dams was a necessary condition for the presence of environmental considerations within an international water treaty in both humid and arid basins, meaning that there is a strong correlation between a basin’s utilization and environmental treaty considerations. Pollution in the form of crisis or ‘flashpoint’ events was not found to be a necessary condition, which suggests that it’s likely that treaties were formulated largely to address more chronic environmental problems.

The second scenario (‘An agreement leads to water utilization …’) was found to be more likely in arid basins than in humid basins. This could be due to greater competition over water in arid basins, incentivizing the parties to reach agreements over water utilization and water quantity to fully utilize the scarce water resources. Indeed, cooperation between riparians over shared waters is much more likely than conflict, even in arid regions (De Stefano et al., 2010).

For arid (sub-)basins, no cases were found where high stress in the basins directly led to an environmental agreement (third scenario). Rather, it appears that water stress led to water utilization agreements, in which institutions were put in place that in some cases served to formulate later environmental provisions (fourth scenario).

Consequently, arid basins had higher consistency scores for previous utilization treaties, previous crises, and previous water treaties. This suggests that aridity of basins provided an incentive for negotiating relationships to develop, thereby facilitating environmental water cooperation.

Several different configurations of conditions appear to result in environmental treaty components. Examining each condition individually indicates that human water stress, upstream large dams, OECD membership, and previous water treaties (particularly over-utilization) are conditions that are the most prevalent for environmental degradation to be considered in international water treaties. Utilization through the construction of large dams is something that is present in almost all basins in our dataset that contain treaty lineages with environmental degradation language; this is in clear contrast to the counterfactual dataset.

Most of our cases (62 of 75) occur in basins without environmental water stress (i.e., humid basins), supporting our initial hypothesis. However, the majority (42 of 75) of treaty lineages in our dataset take place in basins where human water stress exists, suggesting a linkage between utilization and environmental considerations. Also, only 35 of the 75 treaty lineages are in basins where the majority of state treaty parties are OECD members. The lower than expected OECD membership result can be partially explained by the higher rate of environmental treaties in the last half of the 20th century, corresponding with Stern’s (2004) finding that basins within the developing world are addressing environmental issues in a relatively shorter time lag than in the developed world.

Only three treaties mentioned adverse effects of past utilization treaties: the Agreement on Paraná River Projects, Protocol VI to the treaty on the Lesotho Highlands Water Project, and the Joint Declaration of Principles for Utilization of the Waters of the Lower Mekong Basin. While there are few cases where previous treaties were explicitly mentioned, it is likely that other treaties were indirectly inspired to address environmental issues caused by past utilization both agreed upon and unaddressed within previous treaties based on treaty language (e.g., Minute 241, an interpretation of the 1944 US–Mexico treaty).

Also, surprisingly, there were few water crises listed within the TFDD Water Conflict Event Database that appear to trigger environmental water agreements, contrary to Shmueli’s (1999) findings which were based on seven international basin accords. One important caveat is that some flashpoints, such as droughts or floods, would not register as conflictive events within the TFDD, as they may inspire joint action. Our study did not test for political windows of opportunity, as we did not test for the timing of agreements within their political context. Another factor that might explain the results is that environmental degradation does not trigger treaties to address the issue by itself because environmental issues are perceived as difficult to generate political will (Falkenmark & Molden, 2008). Indeed, most treaty lineages state the intention to prevent environmental degradation and are coupled with other transboundary water issues, while less than a third of the total treaty lineages state specifics on how to address environmental degradation and/or establish a commission focused on environmental degradation (Figure 3).

Over half of the treaty lineages (40 of 75) govern (sub-)basins in Northern Europe and North America, despite having less than half of identified transboundary river basins. This could be partly explained by a strong environmental regulatory emphasis present in these regions, particularly Northern Europe (e.g., the 1992 Convention on the Protection and Use of Transboundary Watercourses and International Lakes). However, this trend is shifting over time with more environmental water agreements in the rest of the world in the latter part of our analysis time period. The trend of basins with environmental water stress has also been shifting, where 12 out of 13 of the treaty lineages with environmentally water stress in our dataset have signed environmental treaty provisions in or after 1972.

The absence of a treaty does not imply a lack of institutional relationship. For instance, Jordan and Israel were not coded as partners to a previous treaty but had a working relationship since the so-called Picnic Table talks (Sosland, 2007). Also, countries could have a working relationship on other water bodies not covered by a formal treaty, and a presence of a treaty does not necessarily translate to improved water quality (Renner et al., 2018).

Some basins and sub-basins appeared more than once in the dataset. Often this was because the treaty lineages dealt with separate issues. For example, the Rio Grande/Rio Bravo Basin is represented multiple times within the dataset because each treaty addressed separate environmental issues – one treaty focused on the environment while another focused on transboundary wastewater issues.

Conclusions

This study analysed the evolution of the inclusion of clauses addressing environmental degradation within transboundary water agreements. We examined how transboundary water agreements addressed environmental degradation in several ways. The first analysis examined which conditions are needed for the consideration of environmental degradation in water institutions. While none of the conditions were determined to be necessary for environmental clauses to be included within transboundary water frameworks, a few combinations of conditions seem to be conducive to reach this outcome.

The hypothesis that states in water-stressed basins/sub-basins were less likely to consider the environment, is substantiated in the results. However, this is true only for direct impacts that lead directly to environmental agreements. Despite the smaller likelihood, arid basins appear to be increasingly addressing environmental issues, modifying previous agreements which facilitated the utilization (largely abstraction) of water within those basins. Previous institutional relationships also correlated with a higher likelihood of environmental provisions. Thus, in arid basins prior agreements that established institutional relationships for water utilization serve to facilitate the later incorporation of environmental provisions. Finally, our hypothesis of a greater likelihood for high-income economies to include environmental provisions was true historically. However, in recent decades lower income economies are signing more treaties with environmental clauses.

Our results suggest that there is no magic recipe for creating environmental provisions. However, we did find some general patterns. Slightly less than half of the cases examined are where the environment was a (or the) driving force in the agreement. Most of the environmentally driven agreements were stronger in addressing environmental degradation through specific steps or through establishing a commission. Very few treaties addressed degradation that resulted from previous treaties.

Our study has limitations in its approach. For one, the case studies are not examined in-depth, making it more difficult to pinpoint the circumstances that led to environmental considerations being incorporated within treaties. Our study has also not assessed how effective these treaties have been in preserving, protecting, and/or improving the environment. Also, our data focused on treaties signed until 2008, excluding treaties signed since that date.

Addressing transboundary environmental degradation is a complicated task. The trend in transboundary water agreements shows that more countries are paying attention to it, and thus environmental provisions are increasingly included in water treaties. However, there are still many transboundary environmental problems that are not (or not adequately) being addressed by any framework, international or otherwise. Even when environmental provisions are included, they tend to be declaratory rather than specific. The question when and under which circumstances specific clauses are included, and to what extent do such treaties indeed lead to a better environment require further study.

Acknowledgements

The authors thank their anonymous reviewers for their feedback and suggestions.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplementary data for this article can be accessed at https://doi.org/10.1080/02508060.2022.2072663.

Notes

1. In this context, basin ‘closure’ refers to when basins have no water left to be mobilized and used (Venot et al., 2008).