Changes to anthropogenic pressures on reach-scale rivers in South and Southeast Asia from 1990 to 2014

We use the MERIT Basins to represent river networks. This new dataset created by Lin et al (2019) reconstructs the global river networks from MERIT Hydro at a 3-arc-second resolution (∼90 m at the equator) (Yamazaki et al 2019). A 25 km² drainage area threshold was selected for channels to derive river flowlines as per Lin et al (2019). Although there are currently several river network datasets (Vörösmarty et al 2000a, Allen and Pavelsky 2018), we use MERIT Basins because it includes numerous low-order small rivers. The bankfull width of low-order small rivers ranges from 32.7 to 2444.5 m (Lin et al 2020). We clip MERIT Basins using the study area boundary to obtain a total of 155 443 river reaches, yielding a drainage density of 0.2 km⁻¹ and an average reach length of 9.6 km (figure A1).

We use the GHS built-up area dataset to estimate anthropogenic pressures from human usage (both industrial and domestic use). Built-up area describes the distribution of industry and human settlement (i.e. human presence), and represents human usage on river water (Vörösmarty et al 2010). Developed by the European Commission's Joint Research Centre, the GHS dataset depicts 30 m global multi-temporal built-up area derived from Landsat imagery (Pesaresi et al 2013, Martino et al 2016, Kemper et al 2017). This dataset has the highest accuracy (89.0%) of all freely accessible global long-term built-up area datasets (Liu et al 2019). We extract built-up area in South and Southeast Asia for the years 1990, 2000 and 2014.

To model irrigation, we use the irrigated area dataset from the ESA-CCI. The ESA-CCI irrigated area dataset is derived from the ESA-CCI Land Cover product, which consists of 300 m annual global land cover with 36 classes from 1992 to 2018 (Bontemps et al 2013). We use the land cover class most closely related to irrigation, 'Cropland, irrigated or post-flooding,' to represent irrigated area in 1992 (as data for 1990 is unavailable), 2000, and 2014.

We use the Digital Global Map of Irrigation Areas (GMIA) to obtain the contribution of surface water to total irrigation water demand (Siebert et al 2013). GMIA categorizes the sources of irrigation water into surface water, groundwater, and non-conventional water at a 5-minute scale. We use the map of irrigation water taken from surface water to represent irrigation water from rivers.

We propose a simple and straightforward approach to estimate dynamic anthropogenic pressures on river reaches (figure 2). Human usage and irrigation were selected as our two essential indicators of anthropogenic pressures on river sustainability. We calculated the pressures induced by each of our two indicators and summed them to obtain the total anthropogenic pressures on river reaches (figure 2).

Zoom In Zoom Out Reset image size

First, the study area was divided into 5 × 5 km grids. Next, the built-up and irrigated area within each grid were calculated. Then, we calculated the distance of each grid centroid to its nearest river reach using the Near tool in ArcGIS 10.3. Each grid was associated with one river reach, which was based on the assumption that humans are more likely to use water from the river reach closest to them (Kummu et al 2011, Fang et al 2018, Fang and Jawitz 2019). The distance reflects the capacity of anthropogenic pressures on rivers, for the closer anthropogenic activities are to a river reach, the more anthropogenic pressures it bears (Kummu et al 2011, Fang and Jawitz 2019). As such, the pressures induced by each indicator on each river reach were calculated as follows:

$$\begin{equation}{P_j} = \mathop \sum \limits_{i = 1}^n \frac{{{A_i}}}{{{D_i}}} \times \lambda \end{equation} \tag{ 1 }$$

where P_j indicates the pressures induced by each indicator on river reach j. A_i is the area (km²) of each indicator in grid i. D_i is the distance (km) of grid i to its nearest river reach j. λ represents the proportion of water for human usage or irrigation sourced from rivers. We set λ= 1 to estimate anthropogenic pressure from human usage, and obtain λ for estimating irrigation-induced pressure from the GMIA dataset. N represents the number of all grids attached to river reach j. To calculate N, we first obtained the river reach ID closest to each grid. Then, taking each river reach as a unit, we counted all grids associated with the river reach.

Next, the anthropogenic pressures across all 155 443 river reaches from each indicator were standardized and normalized. To compress extremely high values, we standardized our results using the Cumulative Distribution Function (CDF). CDF standardizes the pressure of each river reach by calculating its percentile relative to the frequency distribution of pressure across all reaches (Vörösmarty et al 2010). The standardized values were then normalized to a continuous 0–1 scale by subtracting the minimum anthropogenic pressures and dividing by the pressure range across all river reaches and years.

Finally, the total anthropogenic pressures on each river reach were obtained by summing the values of the two indicators, with total sums ranging from 0 to 2. We used the Strahler stream orders of 1–3 and >6 to identify small and large river reaches (Vannote et al 1980, Allen et al 2018), respectively, and analyzed corresponding differences in anthropogenic pressures. The growth rate of anthropogenic pressures was calculated as the ratio of the difference value in a given period to the anthropogenic pressures in the starting year. The dominant pressure on river reaches was determined to be the higher of the two indicators. If anthropogenic pressures caused by the two indicators were equal, the dominant pressure indicator was determined to be the combined impacts of human usage and irrigation.

In 2014, the average anthropogenic pressure on the 155 443 river reaches in our study area was 0.4, representing an increase of 14.8% since 1990. Our finding of increasing anthropogenic pressures is consistent with Ceola et al (2019). Only 41.2% of total river reaches are 'healthy' in 2014, or unaffected by anthropogenic pressures from human usage or irrigation. This marks a decrease of 12.5% (from 73 164 to 63 985) from 1990 (figure 3). Healthy river reaches are mainly concentrated in the southern areas of the Tibetan Plateau, whose steep, high-altitude mountains are unsuitable for human settlement and irrigation (Chen et al 2013). Different areas also exhibit heterogeneity in spatial patterns of anthropogenic pressures on river reaches (figure 3). For example, anthropogenic pressures on river reaches near the southern Himalayas have remained low since 1990. In contrast, reaches near eastern Pakistan, northern India, Bangladesh, southern Thailand, and southeast Vietnam are consistent hotspots of anthropogenic pressures throughout our study period (figure A2), indicating stable patterns of human settlement and irrigation over time. Finally, in places such as India, Pakistan, and Bangladesh where irrigation relies heavily on groundwater, its depletion over time may intensify anthropogenic pressures on river reaches (Wada et al 2010, 2012) (appendix A4).

Zoom In Zoom Out Reset image size

River reaches in eastern Pakistan and northern India exhibit similar trends in anthropogenic pressures, which increase more rapidly than in other parts of South and Southeast Asia from 1990 to 2014 (figure 4). These two densely populated areas have the greatest spatial concentration in increases in anthropogenic pressures (figure A2). The 1724 river reaches that experienced a decrease in anthropogenic pressures were scattered across a much wider area (figure 4). These decreases are possibly due to the observed transformation of irrigated land into other non-built-up land cover types (e.g. grassland and forest) (Li et al 2019), which may have helped reduce anthropogenic pressures on river reaches.

Zoom In Zoom Out Reset image size

Anthropogenic pressures on river reaches are found to increase with stream order, with larger (higher-order, >6) rivers facing more intense anthropogenic pressures than smaller (lower-order, 1–3) rivers (figure 5(a)). The average anthropogenic pressures were 0.4 on reaches within small rivers and 0.6 on reaches within large rivers, increasing by 50.0% between small and large rivers (table 1). The average anthropogenic pressures on first-order rivers were 0.4 in contrast to 1.0 on eighth-order rivers, increasing by 150.0% between first-order and eighth-order rivers (figure A3). This difference may be due to the fact that areas near large rivers tend to run through large floodplains, which are more suitable for human activities than areas adjacent to small rivers (Di Baldassarre et al 2013). Anthropogenic pressures on small rivers, however, increased more rapidly than those on large rivers (figure 5(a)), and in a statistically significant manner, too (table 1). The average anthropogenic pressures on first-order rivers increased by 15.8% from 1990 to 2014, in contrast to 13.3% on eighth-order rivers. Low-order small rivers are expected to bear much more anthropogenic pressures than high-order large rivers if groundwater overexploitation persists (appendix A4). Notably, the opposite results were obtained when estimating water scarcity, which includes water availability. Since high-order large rivers have greater discharge than low-order small rivers, water scarcity poses less of an issue for them (0.01 for large rivers vs 0.38 for small rivers, appendix A5).

Zoom In Zoom Out Reset image size

The number of healthy small river reaches decreased more rapidly than the number of healthy large river reaches (figure 5(b)). The proportion of healthy river reaches from orders 1 to 7 decreased from 1990 to 2014, within which the healthy first-order rivers had the largest decrease of 7.3%. The proportion of healthy first-order river reaches is lower than that of reaches from orders 2 to 7 (figure 5(b)). The is because first-order river reaches are both densely and widely distributed across the study area, with more than twice as many first-order as second-order river reaches (78 807 vs 36 893). Therefore, a large number of first-order river reaches are linked with a grid, meaning they are subject to pressure from nearby surrounding anthropogenic activities. The proportion of healthy eighth-order river reaches (21.1%) remained unchanged over the 25-year period (figure A4). The results reveal that although large rivers initially did and still do experience the greatest anthropogenic pressures, built-up and irrigated areas are rapidly encroaching upon small rivers, whose sustainability is increasingly at risk.

We further investigated the dominant pressure on river reaches (figure 6). In 2014, irrigation accounted for the dominant pressure on 33.3% of total river reaches, while human usage accounted for the dominant pressure on 24.0% of river reaches. These latter river reaches were concentrated in the Malay Peninsula, where irrigated area is sparse due to the presence of steep hills and mountains (Cavendish 2007). Human usage also predominantly influenced a number of river reaches in India and Bangladesh due to the combined factors of significant built-up area and the presence of sufficient local groundwater sources to help relieve pressure on irrigation water (appendix A5). The combined pressures from human usage and irrigation account for 1.6% of total river reaches, mainly in eastern Pakistan, southern Thailand, and Bangladesh (figure 6). Large rivers appear more susceptible to the combined pressures of human usage and irrigation, while small rivers tend to be dominated by a single pressure (table 2). For example, although human usage is the main pressure on 29.9% of first-order rivers, this is only the case for 14.8% of eighth-order rivers. In contrast, the combined impacts of human usage and irrigation affect 4.5% of eighth-order rivers, compared to 1.5% of first-order rivers.

Zoom In Zoom Out Reset image size

Since 1990, anthropogenic pressures in the Ganges-Brahmaputra river basin have consistently remained the highest (∼0.5) of all four transboundary river basins examined (figure 7). This river basin experiences the lowest anthropogenic pressure increment (3.8% from 1990 to 2000 and 4.9% from 2000 to 2014), indicating a relatively stable speed of urbanization and intensification of irrigation. The Indus river basin exhibited the next highest level of anthropogenic pressures (∼0.4), although its value increased the most rapidly of the four river basins (by 10.4% from 1990 to 2000 and 10.8% from 2000 to 2014). The finding suggests that the Indus river basin exhibits a higher speed of urbanization and intensified irrigation than the other three transboundary river basins. Compared to South Asia, anthropogenic pressures in Southeast Asia were relatively low (∼0.2) from 1990 to 2014 (figure A3), with an average increase of 8.5% and 6.9% in the Mekong and Salween-Irrawaddy river basins, respectively.

Zoom In Zoom Out Reset image size

Although anthropogenic pressures exhibited an overall increasing trend, further downstream, different river basins exhibited variations in longitudinal anthropogenic pressures (figure 8). Within the highly impacted Indus river basin, which mostly lies within Pakistan and India, hotspots are located near the Pakistan-India border (figure 8(a)). This concentration is likely due to a certain path dependency: the Pakistan-India border runs through fertile plains that have been home to agricultural activity for thousands of years (Biemans et al 2019), which over time gave rise to major conurbations located near the border such as Lahore in Pakistan's Punjab province and Amritsar in India. River reaches flowing through Pakistan's Punjab province have underwent the fastest increases within the Indus river basin in anthropogenic pressures from 1990 to 2014 (figure 8(a)). Unlike the Indus river basin, in the other three river basins, the greatest anthropogenic pressures all occur near river outlets, where rivers meet heavily populated coastlines (figures 8(b)–(d)). There are evident spikes in anthropogenic pressures in upstream river reaches flowing across China in these three river basins, which suggests high levels of human activities. These intense demands on rivers may affect downstream water usage since such areas are highly dependent on upstream river discharge (Gleason and Hamdan 2017, Munia et al 2020, Viviroli et al 2020). River reaches under intense anthropogenic pressures in the Salween-Irrawaddy river basin are mainly located near the cities of Yangon and Mandalay (Myanmar), Chamdo (China), and along the China-Myanmar border (figure 8(c)). Although the Mekong river extends across 6 countries, hotspots are concentrated in China's Yunnan province, along the Thailand-Laos border, and near the Mekong Delta's river outlets in the South China Sea. Given relatively consistent levels of anthropogenic activities over time, there is little difference in changes in longitudinal anthropogenic pressure dynamics in the Salween-Irrawaddy and Mekong river basins from 1990 to 2014 (figures 8(c) and (d)).

Zoom In Zoom Out Reset image size

Our study represents a preliminary attempt to investigate spatiotemporal variations in anthropogenic pressures on river reaches of different sizes. We find that anthropogenic pressures on river reaches with different stream orders are heterogeneous: Small rivers experience a faster increase in anthropogenic pressures and a greater decrease in the number of healthy river reaches than large rivers (figure A4). This is consistent with previous studies, which find that more small rivers than large rivers have disappeared in China, the U.S., and Canada due to rapid urbanization (Elmore and Kaushal 2008, Stammler et al 2013, Han et al 2020). Anthropogenic pressures on large rivers remain high but stable, reflecting the fact that they have long been disturbed by humans due to their conduciveness to settlement and agriculture. These two factors are important reasons why urban areas initially developed near large rivers during the Neolithic Revolution (Di Baldassarre et al 2013, Li et al 2015). With cities along large rivers facing severe water problems such as water scarcity (Li et al 2015), however, built-up and irrigated area are expanding towards small rivers, leading to rapid increases in anthropogenic pressures. Furthermore, due to small rivers' low river discharge levels, increasing anthropogenic pressures will aggravate water scarcity in small rivers (appendix A4). More attention to these little-studied small rivers is critical, for they comprise 88.1% of total river length in our study area and are more sensitive to anthropogenic pressures than large rivers (Kaushal et al 2006, Han et al 2020). Small rivers also perform essential functions, reducing pressures on large rivers by dissolving nutrients that are exported to them and providing clean water (Peterson et al 2001, Meyer et al 2007, Elmore and Kaushal 2008). Increasing anthropogenic pressures on small rivers may thus not only decrease their surface area, discharge, riparian erosion, and stream burial (Burns et al 2005, Napieralski and Carvalhaes 2016, Taniguchi et al 2018), but also aggravate issues facing large rivers such as biodiversity loss, river contamination, and nutrient and sediment increment (Meyer et al 2007).

While the four transboundary rivers in our study area all originate in the Tibetan Plateau (Brookfield 1998, Clift 2002, Ziv et al 2012), their river reaches show distinct divergences in anthropogenic pressures due to regional and national variations in levels of economic development and irrigation (OECD 2019). Contrary to Ceola et al (2019), we find that anthropogenic pressures increased more rapidly in the Indus river basin than in the Mekong and Salween-Irrawaddy river basins (figure 7). This discrepancy may be due to the fact that the nighttime lights data used within their study is affected by saturation, which means that the digital number (DN) of illuminated pixels cannot be measured beyond a certain level (63) (Levin et al 2020). Given the already high levels of human activities in the Indus river basin, DN values there are saturated and do not vary interannually, meaning that anthropogenic pressure dynamics may be underestimated.

We also identify several 'peaks' of intense anthropogenic pressures on midstream and upstream river reaches in the four transboundary river basins, such as in Punjab (Pakistan) along the Indus river and in Chamdo (China) adjacent to the Salween-Irrawaddy river (figure 8). Intense upstream anthropogenic pressures will impact downstream communities and their river water use because downstream regions are highly dependent on upstream water, possibly sparking tensions and conflicts (Zawahri 2009, Petersen-Perlman et al 2017, Viviroli et al 2020). Greater efforts are required to balance anthropogenic pressures along transboundary rivers' entire watercourses and ensure that the needs of downstream populations can be met. Furthermore, some of the largest conurbations within South and Southeast Asia are located around river outlets, such as Yangon in Myanmar and Dhaka in Bangladesh (Hedley et al 2010). High levels of human activities near river outlets can lead to extremely high anthropogenic pressures (figures 8(b)–(d)), potentially affecting delta morphology and increasing flood risks (Syvitski and Saito 2007, Syvitski et al 2009, Tessler et al 2015). Understanding the sustainability of these vulnerable urbanized deltas both in and beyond Asia will require interdisciplinary research into rivers and how they interact with social, political, and economic systems.

Our study develops a simple, straightforward approach to estimate river sustainability by focusing on two fundamental indicators: human usage and irrigation. Although this approach captures the main drivers of anthropogenic pressures on rivers, it has some limitations. For instance, we do not include comprehensive indicators (e.g. dams and water pollution) (Vörösmarty et al 2010), nor do we build complex hydrological models (e.g. WaterGAP, PCR-GLOBWB 2.0) (Yassin et al 2019, Doll et al 2003, Sutanudjaja et al 2018) to assess anthropogenic pressures on rivers. In addition, although reservoir and dam operations play an important role in impacting hydrological processes—especially in South and Southeast Asia, where many have been built to take advantage of significant hydropower potential in recent decades—our study does not include them because we aim to separate the impacts of anthropogenic activities on river reaches (i.e. potential river water demand). In contrast, reservoir operations control water availability by adjusting river discharge (Hanasaki et al 2006, Mateo et al 2014, Zhang et al 2014, Akbas et al 2020, Getirana et al 2020). Furthermore, while we consider the effects of human usage and irrigation to be equal, from the perspective of water use, irrigation exerts a greater effect than human usage (Islam et al 2015, Karthikeyan et al 2020). Lastly, we calculate the anthropogenic pressures of each grid on its nearest river reach and do not consider the anthropogenic pressures on other river reaches due to major water transfer projects such as China's South–North Water Transfer Project (Moore 2014). These issues may all be worth considering in future research.