Data and Methods Differences and Problems

The available data on sediment flow on the Lancang-Mekong river varies in time, quality, methods, sampling frequency, and measuring statements throughout the upper and lower basin. The data most frequently derives from measuring stations located in the following places (descending in order from north to south): Jinghong (China), Yunjinghong

(China), Chiang Saen (Thailand), Luang Prabrang (Lao PDR), Nong Khai (Thailand), Mukdahan (Thailand), Pakse (Lao PDR) (see map #2, below). At several of these stations, measurements began in the 1960s, but the available records are discontinuous and frequently limited in the number of available samples. While researching, the author found several key datasets used across a number of studies:

Sources Timescale Sampling

  57   Chart 1: Available datasets on the Mekong’s sediment load

The Lower Mekong Project, began in 1962, was a US Agency for International Development effort to build a basin-wide database. While originally functioning through the Harza Engineering Company, it is continued today by local national agencies

(Walling 2009). The next dataset comes from either sparing releases of information by the Chinese government, academic articles from scholars given access to and permission to publish work on the information, and through a compilation of secondary sources by Walling (2009) from 1963, and 1965-1990. Walling’s compiled dataset lacked

information on the sampling regime, but given existing information suggests it likely took frequent and daily measurements with standard sampling tools. The final primary dataset derives from the MRC’s water quality monitoring network measuring total suspended solids (TSS).

  58  

Map 2: Upper and Lower Mekong measuring stations (Source: Walling 2009, 116)

As described by Walling (2009), and reiterated in an interview³, “the reliability of the results obtained will depend heavily upon the nature of the sediment-sampling or monitoring program and the accuracy of the resulting load estimates” (Walling 2009).

3  Interview  01:  Fu  Kaidao,  senior  researcher  at  the  AIRC.  

of the basin occupied by steep mountainous headwaters, characterized by high sediment yields, and more lowland areas, which are characterized by lower sediment yields and fre-quently represent significant sediment sinks.

Like the Yangtze, the Pearl River, and the Indus, the Mekong has its headwaters in steep mountainous areas, but a considerable propor-tion of its basin is occupied by areas of interme-diate and low relief, which contribute less

FIGURE 6.1 The Mekong basin.

116 6. SEDIMENT LOAD

  59  

For example, the more standard isokinetic samplers used in the Lower Mekong Data project and similar tools used in China, allow for more accurate samples by allowing a more accurate representative sample of silt of varying coarseness to be measured. The Water Quality Measuring network dataset used the technique of capturing the silt in plain bottle and then measuring the dry weight later, which leaves the measurements made from this data open to inaccuracies (Walling 2008a, 2009).

5.1.2 Uncertainty

As nearly all articles claim, the impacts of the dam cascade on Lower Mekong basin (LMB) are contentious, in large part because of the lack of long-term and accurate sediment measurements in the Lower Mekong River. There are three primary means of measuring sediment volume: total suspended solids (TSS), suspended sediment

concentration (SSC), and a variation on the former, depth-integrated SSC. As Walling (2008a, 2009) describes, however, this measurement system involves collecting samples near the surface (0.3m depth) with a bottle rather than a true sampler, which then later measures the dry weight of the silt sampled. This process, however, does not integrate sediment at various depths, which in turn does not factor in heavier sediment flow in middle depths, nor the coarser sediment found close to or on the riverbed. As such, samples likely have inaccuracies from the initial sampling site, and inaccuracies from the instrument itself (a bottle) (Walling 2009). Depth-integrated SSC makes up for these deficiencies through measuring multi-depths with better equipment at one measuring site.

A number of these well-regarded and oft-cited studies are forced to use reconstructed datasets. As Wang et al. (2009) pointed out, the more accurate depth-integrated SSC measurements are traditionally time-consuming, relatively expensive, and in the Lower Mekong river, sporadic. Indeed, among the five measuring stations covering 42 years from 1962-2003, depth-integrated SSC measurements are only available for 28 years at Chiang Saen, 16 years at Luang Prabang, 30 years at Nong Khai, 42 years at Mukdahan, and 25 years at Khon Chiam (Wang et al 2009). During several of those years, less than 10 SSC measurements were taken. Excluding those years, they drop to 25, 11, 29, 38, and 22 years respectively (Wang et al 2009), greatly hindering sediment

  60  

load estimations (Walling 2006, 2009). Studies that did try to make estimates for the years without significant data used a number of methods to reconstruct the data, many using the rating curve method (e.g. Walling 2005, Lu and Siew 2006, Kummu and Varis 2007). This particular method left the period 1962-1980, sediment loads at Luang Prabang only existed for 1 year, and none at all at Pakse. Therefore, comparisons of annual sediment loads following pre- and post-Manwan and Dachaoshan dam may require more information. Data comparison among stations in the Lower Mekong conducting relatively reliable analysis can also be difficult due to differing laboratory procedures. For example, comparing SSC data between Thailand and Laos is difficult as the labs used differing procedures, and Laos additionally lost some organic matter during sample processing (Walling 2006; Wang et al 2009).

  61  

Figure 3: The availability of annual sediment load estimates (x10⁶ t) in Thailand and Laos, and equivalent data for the Upper Mekong at Jinghong from 1983-1990.

(Source: Walling 2009, 126)

5.1.3 Measuring Frequency, Timescales, and Methodology

Sampling frequency strongly correlates with the confidence of a given dataset, as it impacts the ability to make a “best estimate” from the given data. One of the primary difficulties facing accurate data within the basin falls on infrequent and/or inconsistent

FIGURE 6.4 The available estimates of annual sediment load (!10⁶ t) for the five designated sites on the Mekong River in Thailand and Lao PDR and equivalent data for the period 1983-1990 for the Upper Mekong or Lancang River at Jinghong, China.

126 6. SEDIMENT LOAD

  62  

sampling. In a review of available basin-wide datasets, Walling (2009) attempted to find a “best estimate” baseline through a comparison of nine measuring station’s

measurements with years holding relatively high rates of suspended sediment

concentration data. Walling found that a decrease in sampling frequency considerably increased the uncertainty of estimates. From the given data (1960-2003), measurements taken every 14 days allowed for + - 10% error rate at the 95% confidence level, while at every 28 days the error rate dropped to + - 20-25% at a given station.

Data from measuring stations only represents values of sediment concentration on the occasion of its sampling. Samples taken on a monthly basis, then, will vary widely during the flood season and offer a less accurate sampling of sediment flow trends.

Therefore, the TSS based water quality network dataset, measured monthly, has serious limitations on the data as a source of representative information, especially during the flood season. While comparisons of data from two different sampling methods is not possible, the error rate of surface dip (TSS) samples can be found through a comparison of equivalent concentration values. Through a comparison of equivalent data at Chiang Saen for the years 1994-2001, Walling (2009) found that TSS concentrations ranged from 1000 mg 1^-1 to 1500 mg 1^-1, but not exceeding 1500. The sediment-sampling program indicated concentrations exceeding 1000 mg 1^-1 for extended period with some samples exceeding 2500 mg. TSS, therefore, likely underestimates actual sediment flow (Walling 2009). SSC, on the other hand, increases with greater depth. As such, datasets like the MRC’s water quality monitoring network underestimate the true mean concentration as they offer only measurements of shallow river depths, while also varying in during the wet and dry season.

  63  

Figure 4: A comparison of TSS concentrations from the water quality monitoring program at Chiang Saen (A) versus the SSC from the sediment-sampling program at the same location (Source: Walling 2009, 124)

Walling (2009) findings put serious doubt on the use of TSS data. This is an important claim as a number of important studies (Fu et al 2006; Kummu and Varis 2007; Lu and Siew 2006) used by opponents of the cascade used these very datasets to claim large already existing sediment flow impacts from the Manwan and Dachaoshan dams