Macroinvertebrates in Asian and European rivers: a general overview
Taxonomic and ecological knowledge on freshwater macroinvertebrates in Asia, as mentioned earlier, is still limited (Boulton et al. 2008; Boyero et al. 2009). Most research in the tropical Asia is largely restricted to a few geographic regions including the Hong Kong and Peninsular Malaysia (Resh 2007; Boyero et al. 2009; Leung and Dudgeon 2011; Al-Shami et al. 2013). A wider range of studies has also been revealed from the northern subtropical Asian rivers, e.g. Yangtze and the Upper Mekong River or the so-called Lancang River in China. Research topics from these river basins include species distribution, spatio-temporal patterns and species records (Nieser et al. 2005; Shao et al. 2008; Qi et al. 2012). However, most of the research findings are reported in Chinese and are not publically available; only a few are accessible, e.g. the benthic macroinvertebrates as indicators of ecological status in Yangtze River (Pan et al. 2013) and the seasonal variability of metazooplankton (including crustaceans) communities and new mollusc species records from the Upper Mekong Basin (Du et al. 2011; Wu et al. 2014). For the Lower Mekong Basin, more investigations have been recently conducted in Thai streams (Boonsoong et al. 2010; Kudthalang and Thanee 2010; Phaphong and Sangpradub 2012; David and Boonsoong 2014) and recently also some Philippine streams have been investigated (Tampus et al. 2012; Sinco et al. 2014; Fajardo et al. 2015; Magbanua et al. 2015, Forio et al. 2017). Most of these studies are related to species diversity, description of new species and using benthos to assess water quality in river systems (Parnrong et al. 2002; Sangpradub et al. 2002; Flores and Zafaralla 2012). Macroinvertebrates from other geographic areas including Myanmar, Laos, Cambodia and Vietnam, which mainly share the Lower Mekong Basin, remain very scarce.
On the contrary, knowledge on macroinvertebrates from river systems in Europe has been extensively studied (Boyero et al. 2009). Since the adoption of the European Water Framework Directive (WFD) (European Commission 2000), freshwater macroinvertebrates have become the central focus (Pollard and Huxham 1998; Hering et al. 2010). Macroinvertebrates from hundreds of streams have been studied and used to assess water quality (Buffagni et al. 2001; Verdonschot and Nijboer 2004). Within 10 years of the implementation, ~1,900 papers resulted from research projects associated with WFD (Hering et al. 2010). This results in a very well documented knowledge on freshwater macroinvertebrates and their ecological applications for Europe. Furthermore, a diverse assessment methods have been developed (Birk et al. 2012), some of which have applied a predictive modelling framework that is based on macroinvertebrates or use environmental variables to predict future distribution, occurrence and abundance of particular taxa (Goethals et al. 2007; Everaert et al. 2013; Boets et al. 2015).
When macroinvertebrate composition and diversity are related to measured environmental variables, key factors driving spatio-temporal changes have been known to be more or less the same regardless of geographic or climatic regions. For instance, macroinvertebrate communities in river basins from southern China (Pearl, Yangtze and Qiangtang rivers), from northern Portugal (the Olo, Corgo, Pinhao and Tua rivers) and from Susquehanna River (New York, North America) have been reported to be influenced by land use types including anthropogenic disturbance (Allan 2004; Bruns 2005; Cortes et al. 2011; Cortes et al. 2013). Another example can be found from European Mediterranean (Evrotas River, Greece) and Asian streams (Peninsular Malaysia) that stream size (e.g. width and depth), dissolved oxygen and pH were the key factors influencing macroinvertebrate composition and variation (Al-Shami et al. 2013; Salmah et al. 2014; Karaouzas and Płóciennik 2016). These indicate that similar ecological processes can be expected from different ecological systems.
Modelling techniques and applications
Various modelling techniques have been widely and increasingly implemented in ecological systems (Lek et al. 1996; Park et al. 2003; Schröder et al. 2007; Lencioni et al. 2007; Guo et al. 2015). The techniques applied are generally used to explain and predict the relationship between the occurrence or abundance of studied species and environmental variables (Goethals et al. 2007; Boets et al. 2013). Utilization of modelling techniques to combine both explaining and predicting such relationships is also commonly applied (Roura-Pascual et al. 2009; Call et al. 2016). Applications of predictive models have provided knowledge and improved the understanding of the ecology and behaviour of studied taxa, which could be used to support decision making, management and conservation planning. For instance, many previous studies have used predictive models to predict the occurrence and distributional areas of plants, herbs, macroinvertebrates and fish (Thuiller et al. 2005; Roura-Pascual et al. 2009; Vicente et al. 2011; Boets et al. 2013; Chen et al. 2015; Guo et al. 2015).
However, the application of predictive models has been suggested to be carefully taken into account because they can have a wide variation in performance (Segurado and Araujo 2004; Elith et al. 2006; Guisan et al. 2007). Some models even yield contrasting predictions of habitat suitability (e.g. Guisan et al. 2007; Evangelista et al. 2008; Roura-Pascual et al. 2009). Furthermore, predictive models are sensitive to parameterization and selection criteria during the modelling process (Araújo and Guisan 2006; Elith et al. 2006), and thus can result in an uncertainty of current or past/future projections of species distributions (Svenning et al. 2008; Buisson et al. 2010). Due to this fact, when calibrating and validating predictive models, carefully taking into account the data characteristics (e.g. sample size, species prevalence or environmental predictors), parameterization and selection criteria are usually recommended (Luoto et al. 2006; Dormann et al. 2008).
Research problem, aims and objectives
The Lower Mekong Basin (LMB), which includes portions of Thailand, Laos, Cambodia and Vietnam, is characterized by a long and large floodplain (Eastham et al. 2008) and is known for its high biodiversity (Sodhi et al. 2004). However, the knowledge of macroinvertebrates in the LMB is poorly investigated. Given that this river basin is being impacted by various anthropogenic disturbances such as agricultural activities, aquaculture, urbanization and mining (Sodhi et al. 2004; Nhan et al. 2007; Köhler et al. 2012), there is an urgent need to study the patterns of spatial organization, community structure and variations of macroinvertebrates in this basin and their relation to environmental factors. Up to date, only a few studies (except for those conducted in Thailand) have been published on the basin, e.g. community structure and composition of littoral invertebrates in the Mekong delta (Wilby et al. 2006) and the diversity and distribution of crustaceans and molluscs in the Indo-Burma region (Cumberlidge et al. 2011; Köhler et al. 2012). Yet, no attempt has been made to examine the large spatial patterns, community structures, variations (i.e. β diversity) of macroinvertebrate communities and their relation to key environmental variables nor the application of predictive modelling in this hardly studied basin.
On the other hand, river systems in Europe as well as in Flanders suffered from severe water quality degradation in previous decades (e.g. from 1980s to 1990s). During these periods, some native species were reported to disappear (Bernauer and Jansen 2006) and only those that were able to withstand the water quality degradation remained. At the same time, most European river systems have been exposed to a number of alien macroinvertebrate species (Leuven et al. 2009; Boets et al. 2016). From the 2000s until now, the water quality of European rivers has been greatly improved. This water quality improvement does not only promote the occurrence and abundance of native species, but also favours the alien species to spread widely, which consequently may lead to changes in community composition. As such, investigation spatio-temporal changes in community composition, variations and predicting the occurrence of alien species across Flemish rivers, which have been poorly studied, will provide insights into the ecology of overall communities and of studied alien species. Results from this investigation can be used to support management and conservation planning.
Materials and Methods
Case study in the LMB and dataset
The Mekong River Basin is divided into the Upper Mekong Basin (UMB) and the Lower Mekong Basin (LMB). The UMB on the Tibetan plateau in China is composed of narrow, deep gorges and small, short tributaries, whereas the LMB stretches from Yunnan province in South China to the delta in Vietnam and it covers approximately 70% of the total length of the whole basin (Eastham et al. 2008). The LMB consists of a large floodplain and long, broad tributaries and it drains more than 76% of the Mekong basin. The climate of the LMB is dominated by a tropical monsoon rainfall system, which is characterized by a dry (November – April) and a wet (May – October) season generated by the northeast monsoon and the south-west monsoon, respectively. The most intensive rainfall falls from July to September, while the lowest precipitation is observed between January and April (Adamson et al. 2009). The annual rainfall of the LMB varies from 1,000 – 1,600 mm in the driest regions to 2,000 – 3,000 mm in the wettest regions (Hoanh et al. 2003). A higher precipitation is found in the eastern mountainous regions of Laos and in northeast Thailand (Eastham et al. 2008).
The largest floodplain water body of the LMB is the Tonle Sap Lake (TSL) in Cambodia (Adamson et al., 2009), which is the largest freshwater lake in Southeast Asia (Sarkkula et al. 2003). The TSL is connected to the Mekong through the Tonle Sap River, and thus creating an exceptional hydrological cycle. In the wet season, the TSL receives excess water from the Mekong River and expands its surface area from 2,500 km2 to 15,000 km2. In the dry season when the rain ceases and water levels drop in the Mekong, a reverse flow occurs; the drained water from the TSL flows to the Mekong delta (Arias et al. 2011). The Mekong delta is characterized by a number of man-made canals, which are mostly used for domestic and agricultural activities (Kummu et al. 2008).
Data collection and processing
Benthic macroinvertebrates were sampled at 60 sampling sites along the main channel of the LMB and its tributaries by the Mekong River Commission (MRC) (Fig. 2.1). This sampling was carried out once a year in March during the dry season from 2004 to 2008. At each sampling site, macroinvertebrates were sampled from three locations in the benthic zone: near the left and right banks, and in the middle of the rivers. At each location, a minimum of three samples (where inter-sample variability is low, e.g. tributaries) to a maximum of five samples (where inter-sample variability is higher, e.g. the main channel and the delta) were collected using a Petersen grab sampler which has a sampling area of 0.025 m2. With the grab sampler, four sub-samples were taken and pooled to give a single sampling unit covering a total area of 0.1 m2. In total, between nine (3 samples × 3 locations) and fifteen (5 samples × 3 locations) pooled samples were collected at each sampling site. Each pooled sample was rinsed using a sieve (0.3 n mm mesh size). In the field, the samples were sorted and then preserved by adding 10% formaldehyde to obtain a final concentration of about 5%. In the laboratory, they were identified to the lowest taxonomic level possible and counted using a compound microscope (40 – 1,200 magnification) or a dissecting microscope (16 – 56 magnification). Macroinvertebrate abundance data per sampling unit was averaged across all samples (between 9 and 15 samples) collected from each sampling site.
Fig. 2.1 The Lower Mekong Basin (LMB, A) and macroinvertebrate sampling sites (shaded dots, B). Sub-samples and replicates were taken at each sampling site as illustrated in C.
At the sampling site, geographic coordinates and altitude were determined with a GPS (Garmin GPS 12XL). River width was measured in the field using a Newcon Optik LRB 7×50 laser rangefinder. Other physical-chemical variables were measured at the three locations where macroinvertebrates were sampled. River depth was measured using a line metre. With a handheld water quality probe (YSI 556MP5), water temperature, dissolved oxygen, pH and electrical conductivity were measured at the surface (0.1-0.5 m) and at a depth of 3.5 m or at a maximum depth of the river (wherever less than 3.5 m) and then the average value was recorded for each location. Water transparency was measured with a Secchi disc by lowering it into the water and recording the depth at which it was no longer visible. The physical-chemical data of each sampling site was the averaged value across the three sampling locations. Distance from the sea and the surface area of watersheds drained at each sampling site was determined using a Geographic Information System (ArcGIS 10.0, ESRI). Geographic data (ArcGIS shapefiles) about the LMB (river networks, basin boundaries, land covers, and subcatchments derived from topographical maps) was provided by the MRC.
In total, 108 samples were collected from the 60 sampling sites. In 2008, 3 sampling sites were sampled further away from their original sampling coordinates, and thus were considered as different sampling sites (see Appendix T1). Therefore, a total of 63 sampling sites were taken into account in the analyses. Because of unequal sampling efforts (i.e. unequal and different number of samples at each site during the 5-year sampling period) and missing values of environmental variables, we used median values from the collected data to represent each site in the analyses, as suggested by McCluskey and Lalkhen (2007). These median values were used in all of the analyses corresponding to the case study of the LMB.
Case study in Flemish rivers and dataset
Flanders (northern Belgium) is located in Northwest Europe and its Northwestern part is bordered by the North Sea (Fig. 2.2A). Flanders has a total area of 13,522 km2, and is considered as one of the most densely populated regions in Europe (477 inhabitants/km2 in 2015, https://en.wikipedia.org/wiki/). Flanders is classified as a lowland area, which is divided into different rivers basins (Fig. 2.2B). This region is influenced by a temperate oceanic climate, as same as most of northwestern European countries are (e.g. UK, France, Luxembourg, Netherland and Denmark) (Peel et al. 2007). Flanders has a dense watercourse network including navigable canals. Agriculture, industry and residential areas are the main land use types of Flanders and its landscape is characterized by highly fragmented and complex mosaic of land use types (Poelmans and Van Rompaey 2009). This fragmentation and complexity may have put a high pressure on habitat quality and biodiversity in Flanders.
Fig. 2.2 Map of Flanders indicating: (A) the most important watercourses and geographic locations, the polder area (grey) and the three main harbours indicated by rectangles (Boets et al. 2016), (B) different river basins (van Griensven and Vandenberghe 2006) and (C) monitoring sites between 1991-2010, which were used in the present study.
Data collection and processing
The Flemish Environment Agency (VMM) has collected biological and environmental data in Flanders since 1989. The monitoring sites include all types of watercourses from all river basins. Every three year from the beginning, a fixed set of sampling locations was sampled. Most of the sampling locations were only sporadically sampled, and thus results in a large dataset of more than 11,000 biological samples collected at more than 2500 sites spread over different water bodies (Fig. 2.2C). In this monitoring program, the sampling protocol was entirely based on the method as described by Gabriels et al. (2010). Macroinvertebrates were collected using a standard handnet, which is made of a metal frame (0.2 m by 0.3 m) to which a conical net is attached with a mesh size of 300 µm. The kick sampling was made along the watercourses at a stretch of approximately 10-20 m. Each sample was collected for three minutes for small watercourses (less than 2 m wide) or five minutes for larger rivers. At sampling sites where the kick sampling method was not possible, artificial substrates were used. Three replicates of artificial substrates, which consisted of polypropylene nets (5 litres) filled with bricks of different sizes, were left in the water for a period of at least three weeks after which they were retrieved. Leaving this period enables species to colonize the substrates. The different sampling efforts of the two sampling approaches (the kick and artificial substrate sampling) may have repercussion on the diversity of sampled invertebrates. However, according to Gabriels et al. (2010), the two approaches are standardized semi-quantitative methods and are similar in terms of sampled macroinvertebrate abundance. In the laboratory, macroinvertebrates in the VMM database were identified to the level (family or genus) needed for the calculation of the biotic water quality index.
Electrical conductivity (EC), pH and dissolved oxygen (DO) were measured in the field with a hand-held probe (Cond 315i, oxi 330, wtw, Germany and 826 pH mobile, Metrohm, Switzerland). All additional chemical variables, i.e. ammonium (NH4+), chemical oxygen demand (COD), biological oxygen demand (BOD), total phosphorus (Pt), nitrate (NO3-), nitrite (NO2-), Kjeldahl nitrogen, orthophosphate (oPO4), were retrieved from the monitoring dataset compiled by the VMM and which is online accessible (www.vmm.be). Nutrient analysis was performed spectrophotometrically in accordance to ISO 17025. GIS software (version 9.3.1) applied on the Flemish Hydrographic Atlas was used to determine the slope and the sinuosity of a watercourse at a different height in between two points (1000 m apart) and on a stretch of 100 m, respectively.
Table of contents :
1. General Introduction
1.1 Background to the study
1.2 Macroinvertebrates in Asian and European rivers: a general overview
1.3 Modelling techniques and applications
1.4 Research problem, aims and objectives
2. Materials and Methods
2.1 Case study in the LMB and dataset
2.1.1 The LMB
2.1.2 Data collection and processing
2.2 Case study in Flemish rivers and dataset
2.2.2 Data collection and processing
2.3 Statistical analyses and modelling approaches
2.3.1 Communities clustering and diversity measures
2.3.2 Comparative analyses
2.3.3 Regression and Multivariate analyses
2.3.4 Model development, validation and performance
3. Main Results
3.1 Macroinvertebrate communities and diversity patterns in the LMB
3.1.1 Overall macroinvertebrate communities
3.1.2 Spatial community patterns and their relationship with environmental factors
3.1.3 Macroinvertebrate diversity and its relation to environmental factors
3.2 Macroinvertebrate communities and diversity patterns in Flemish rivers
3.2.1 Overall macroinvertebrate communities
3.2.2 Spatio-temporal community composition and environmental factors
3.2.3 Spatio-temporal diversity pattern and its relation to environmental factors
3.3. Modelling and predicting
3.3.1 Performance variation of modelling techniques applied in the LMB
3.3.2 Modelling alien mollusc occurrence and their co-occurrence with native molluscs
3.3.3 Optimizing the prediction of alien mollusc occurrence
4. General Discussion
4.1 Overall community composition and diversity in the two river systems
4.2 Spatio-temporal changes of communities and their relation to environmental factors
4.2.1 The LMB
4.2.2 Flemish rivers
4.3 Model development, performance and predictions
4.3.1 Modelling techniques and their application in the LMB
4.3.2 Predicting alien species occurrence and their co-occurrence with native molluscs
4.3.3 Optimizing the prediction of alien mollusc occurrence
5. General Conclusion and Perspectives
5.1 General conclusion
5.2 Implications for management and restoration