Assessment of potential impacts of climate and land use changes on stream flow: a case study of the Nam Xong watershed in Lao PDR

Lao People's Democratic Republic (Lao PDR) is a landlocked country in south-east Asia. It has an abundance of water resources and natural resources (Nam Ngum River Basin Sector Development Project (NNRBSDP) 2008a, 2008b). As a developing country, most of its people rely on natural resources. Farming and shifting cultivation have resulted in land cover changes in the last 30 years, while war and increasing urbanization have damaged the land (Ministry of Forestry & Agriculture (MAF) 2005). Moreover, climate change affects this region as Lao PDR is located in the Lower Mekong Basin, which is almost always hot, often humid, and is classified as tropical monsoonal (Hoanh et al. 2010). In previous decades, climate models were used to investigate different world climates. Nowadays, modern technology (SEA START RC 2006) allows us to make predictions about regional climates and run simulations according to various criteria and scenarios. These results can be applied at a regional level (Intergovernmental Panel on Climate Change (IPCC) 2007).

Land use affects stream flow, especially in developing countries (Takamatsu et al. 2012). The relationship between land change and hydrology is complex. Land cover and use directly affect the amount of evaporation, groundwater infiltration and overland run-off (Bosch & Hewlett 1982), as well as altering water demand, particularly in agricultural areas (Kawasaki et al. 2010; Zawawi et al. 2010). These factors control the water yields of surface streams and groundwater aquifers and, thus, the amount of water available for ecosystem function and human use (Mustard & Fisher 2004). Major changes to the land, such as deforestation and the expansion of agricultural and urban areas, affect local and regional hydrological processes (Sahin & Hall 1996; Takamatsu et al. 2012). Changes in land use and climate, and increasing water demand, are critical factors affecting water resources in watersheds in Lao PDR. These factors have the potential to change stream flows, meaning there is a need for critical water resources planning and management (Lyle & Boualapha 2011).

Changes to land use are influenced by population and economic growth. Over the last 10 years in the Nam Xong watershed in Lao PDR, forested areas have been replaced by urban and agricultural areas. A policy aimed at stopping ‘slash and burn’ practices by 2010 was imposed on the rural population, but studies show that some villages, especially in the highlands, continue to ‘slash and burn’ (Miaillier 2007). In addition, land use change is closely linked to water demand: more land under agriculture and a larger population lead to higher water demand, especially in the dry season (Kawasaki et al. 2010; Takamatsu et al. 2012).

An integrated watershed-scale hydrologic model is a good water resources planning tool. In developing countries, where water resources management is a big challenge with increasing pressure due to human activities and uncertainty of climate change, such a tool can be practically used to assess the hydrological impact of future watershed development scenarios even if data availability is relatively limited. Extensive research has been conducted to study future stream flows under different scenarios. However, those studies have sometimes considered only one factor such as land use change (Parvateesam 1996; Calder 2000; Kawasaki et al. 2010; Ly 2010) or climate change (Jha et al. 2004; Eastham et al. 2008; Chingchanagool 2011; Mango et al. 2011; Lauri et al. 2012). For improved watershed management and planning, all major potential impacts on a watershed should be evaluated. This study used an integrated approach to predict future stream flows considering both the impacts of land use and climate change. Climate data were applied on a local scale. Various scenarios were generated in order to assess changes in stream flow. Statistical techniques and geographic information system (GIS) were used to forecast land use change. This allowed for comprehensive scientific predictions, as opposed to conventional approaches which are based on numerous assumptions. The results of this study can be used to support better planning and decision-making at a local level.

Primary data were collected from relevant organizations such as the Department of Meteorology and Hydrology, Department of Forestry (DoF), the Department of Energy (DoE), the National Agricultural and Forestry Research Institute (NAFRI), and the Mekong River Commission Secretariat (MRCS). Secondary data were obtained from literature, such as the forestry strategy of stopping ‘slash and burn’ practices (Ministry of Forestry & Agriculture (MAF) 2005), rate of water use (Water Supply Authority 2007), and growth rate of population (United Nations 2010). Most data were sourced from the MRCS, with future climate data from SEA-START RC (Hoanh et al. 2010). The detail of data sources is shown in Table 1. Within the study area, data from three weather stations were used: Kasy, Vangvieng and Hinhurb (Figure 1). At each station, rainfall data were recorded for different periods. Temperature data were recorded only at Vangvieng station. Information on other climate parameters recorded at Vangvieng station is in the Soil and Water Assessment Tool (SWAT) database 2009. For each station, data were prepared for the current period (1990–2009) and future period (2011–2030).

This paper applied an integrated approach using statistical analysis of historical trends in land use. Types of land use were compared and contrasted to find out the probability of conversion from one type of use to another (Shahidul & Ahmed 2011). By referring to previous change trends, areas where land use is predicted to change in the future can be determined. Variables such as distance from a stream, distance from a main road, population density, elevation, slope range, and policy (whether it is a local conservation area or concession area, for example) were used to forecast future land use (Verburg et al. 2004, 2005; de Nijs 2009). The ROC (relative operating characteristic) curve was applied to evaluate the accuracy of fit of the logistic regression model (Verburg et al. 2004).

In the study area, water is used mainly for domestic purposes and irrigation. Data on the amount of water used for these purposes were input into the SWAT model. Domestic water use in each sub-watershed was calculated based on population size and rate of water use. In addition, the irrigation water requirement was calculated by CROPWAT 8.0 (software of the Food and Agriculture Organization, FAO), using climate, rainfall, soil type and crop schedule data from the study area.

The SWAT model predicts the impact of land management practices on water, sediment, and agricultural chemical yields in watersheds with varying soils (Neitsch et al. 2011). Climate, land use and water use data measured under various conditions were put into SWAT. Using topographical data taken from a 30 m resolution digital elevation model (DEM), information on sub-watershed boundaries and the stream network was delineated (Zhang et al. 2002; Phomcha et al. 2011). For the purposes of this study, the Nam Xong watershed was divided into homogenous sub-watersheds. The final outlet of the watershed is at Hinhurb station near the mouth of the Nam Xong River, at the junction with the Nam Lik River. Land use and soil type layers in GIS format were used to generate hydrological response units (HRUs) for each sub-watershed. Data inputs were generated from weather stations and gauges in each sub-watershed. Weather data included measurements of rainfall, temperature, solar radiation, relative humidity, and wind speed. However, the limitations of this data meant only precipitation measurements were used. Other data were generated using the SWAT database 2009, modified by MRCS (Hoanh et al. 2010). Some parameters were adjusted at the calibration stage to ensure suitable and best results; the Nash–Sutcliffe model efficiency coefficient (E) and percentage of volume error (VE) were used to assess the accuracy of fit of the simulated result (Cibin et al. 2010; Margaret & Chaubey 2010; Sun et al. 2010). According to the limitations of the SWAT model, there was no diversion node. Therefore, during the calibration stage it was assumed that no water was diverted out of the basin and the observed value used for calibration at the final outlet was a combination of observed values from the Nam Xong diversion and Hinhurb station in 2000–2008. After the calibration stage, water diversion calculations were performed manually for relevant individual outlets.

When comparing annual rainfall for the 1990–2009 and 2011–2030 periods, rainfall will increase at the Kasy and Hinhurb stations. However, at Vanvieng station it will decrease by about 17% and 13% for A2 and B2, respectively. With regard to the Nam Xong watershed, A2 and B2 climate scenarios showed that rainfall will increase by 21% and 27%, respectively, in the upper part, and by 68% and 79%, respectively, in the lower part. However, it will decrease by 17% and 13%, respectively (A2 and B2 scenarios), in the middle part of the watershed.

Table 4 shows that from 1997 to 2008 woodland and shrub land was reduced by 170 km², while forest and agricultural areas increased by 38 km² and 127 km², respectively. The quality of forest was not considered in this study. The long-term rate at which land use changes is an important variable for land use prediction. Changes to land use are highly dependent on the local economy and policies which were not considered in this study. Forest land, which occurs only along the boundary of forested areas, did not increase greatly. No woodland or shrub land was converted to forest between 1997 and 2008. The major change in terms of land use in the Nam Xong watershed was that woodland and shrub land were converted to agricultural land, with agricultural land also converted to woodland and shrub land. This was because of the creation of temporary agricultural areas through slash and burn practices. Agricultural land, therefore, can be divided into two types: permanent paddy fields along with other agricultural lands, and temporary crop fields. In the study area, woodland and shrub land will be replaced by agricultural land at a rate of 4.7% or 8,428 ha per 11 years (a fixed area of 8,428 ha for land use in 2019 and 16,856 ha for land use in 2030). All other types of land use change were not considered due to the small areas involved. This may have caused data digitization errors. However, it did not greatly affect stream flow analysis in the study area.

Domestic water use rates are not constant. They depend on various factors, such as the type and density of the community, economic status and livelihood, sanitation, and access to water supplies (Water Supply Authority 2007). To classify the residential area within the study area, consideration was given to factors such as population density, main roads, facilities and economic conditions, the location of districts, and how the Lao National Statistics Center classes a village. In the study area, residential areas were classed as rural or urban. Average water use rates of 50 L per person per day (L/p/d) and 100 L/p/d (WHO values) were applied to rural and urban areas, respectively. The current population was set according to the 2008 figure. Population estimates for each sub-watershed of the Nam Xong watershed were based on the rate of population change estimated by the United Nations (2010): from 2009 to 2030 the annual rate of population change in Lao PDR will be between 1.5 and 1. The calculation shows that average daily domestic water use will increase from 2,780 to 3,490 m³ between 2008 and 2030.

In the study area, water is used mainly for rice paddy cultivation (Nesbitt 2005). During the wet season paddies are cultivated from early July to the end of November; in the dry season from early January to the end of May. In the middle of the Nam Xong watershed, paddies total 3,127 ha in the dry season and 664 ha in the wet season, while in the lower part of the watershed these figures are 75 ha and 35 ha, respectively (Mekong River Commission 2008). Irrigation is needed on 128 days in the dry season and 65 days in the wet season (Mekong River Commission 2008). The irrigation water requirement per unit area calculation was completed using CROPWAT 8.0. Overall losses from water conveyance, application and storage efficiencies were put at 35% of the total irrigation water requirement.

Analysis indicated that the distribution of land use, soil and slope characteristics within each HRU has the greatest impact on predicted stream flow. As the percentage of land use, slope range and soil threshold increases, actual evapotranspiration decreases due to the elimination of land use classes. Therefore, HRU characteristics are the key factors affecting stream flow. Two hundred and sixty HRUs were generated in this study under the threshold of land use, soil and slope class (0/15/0%).

SWAT model simulations for 1990–2009 show that average daily stream flow at the Nam Xong diversion was 1,206 m³/s, with seasonal daily stream flows of 179 m³/s for the wet season and 22 m³/s for the dry season. The annual water volume was 60,568 Mm³. The highest daily flow (267 m³/s) was in August and the lowest (2 m³/s) in March. The average daily stream flow at the mouth of the Nam Xong River was 1,521 m³/s, with a seasonal daily stream flow of 222 m³/s for the wet season and 32 m³/s for the dry season. The annual water volume was 76,423 million m³. The highest daily flow (330 m³/s) was in August and the lowest (4 m³/s) in March. If water diversion is 60%, it is estimated that the final annual stream flow at the river mouth of the Nam Xong watershed will be 29,232 Mm³, or 50 m³/s per day.

Table 6 shows that future land use (Sc3) will increase stream flow by 0.5% at the point of the Nam Xong diversion, and by 0.5% at the mouth of the Nam Xong River. Flows will vary also by season, with higher flows in the wet season and lower flows in the dry season. Table 7 compares baseline conditions and land use change scenarios for the 1990–2009 period. The results in this table support Bosch & Hewlett (1982) and Parvateesam (1996) in showing that future land use changes (agricultural land incrementally replacing woodland and shrub land) will result in higher surface run-off in the middle part of the Nam Xong watershed, while soil water content and evapotranspiration decrease. In the lower part of the watershed, the amount of surface run-off did not change greatly.

The integration of GIS, statistic regression analysis and regional climate and hydrological models is very complex, but useful for investigating local stream flows. Future scenarios revealed stream flows will be affected significantly by climate and land use changes. Using SEA START RC it was shown that rainfall will increase at Kasy and Hinhurb stations in the upper and lower parts of the Nam Xong watershed, and decrease in the middle part of the watershed (Vangvieng station). Stream flows in the Nam Xong watershed were most affected by climate change under the A2 scenario. Climate analysis results showed a decrease in rainfall at Vangvieng station in the future, a result opposite to that obtained for the other two stations. However, it might be that any decrease in rainfall from 2011 to 2030 will be caused by a 20-year period of climate variability, as occurred between 1969 and 1989. Future land use and water use were estimated in order to analyze stream flow changes in the study area. Agriculture density, population density, and distance to major roads were correlated with land use change. For the period 1997–2008, it was found that the major change in terms of land use involved the conversion of woodland and shrub land to agricultural land, mainly as a result of ‘slash and burn’. Strong local policies relating to forest protection and reforestation will limit this practice in future. Future land use scenarios showed that stream flow will increase only by 0.48% (wet season). Overall, future stream flow in the Nam Xong watershed will change little as a result of climate and land use changes. However, it was found that stream flow will decrease by over 10% in the Nam Xong diversion sub-watershed, and increase in the lower part of the Nam Xong watershed. Having more land under agriculture will not greatly affect stream flow. These results can help development planners, decision-makers and managers plan for the use of water resources in a suitable and sustainable manner. As previously discussed, hydrologic data collection is an essential part of planning for water resources development and management. However, many developing countries such as Lao PDR, which has been facing increasing water resources issues and natural disasters such as floods and droughts, lack hydrologic data due to a limited budget to install and maintain gauges. Therefore, in this study, a modeling approach was used as an alternative way to compensate for data limitations in order to predict future flow patterns under different scenarios. The results from this study can be useful for planners and policy-makers who need a better understanding of a watershed under both current and future conditions in order to establish a basin development plan including climate change adaptation.

This research was partially funded by the Thailand International Development Cooperation Agency (TICA). The authors would like to extend their appreciation to all data providers for sharing the information required for this study.