Aquatic macroinvertebrates as water quality bioindicators in Colombia : A systematic review

The use of aquatic macroinvertebrates to assess water quality has been growing during the last 50 years in Colombia. We conducted a systematic review of 109 references to analyze scientific production related to the evaluation of water quality in Colombian freshwater ecosystems, using aquatic macroinvertebrates. We describe bibliographical, geographical, ecological, and methodological characteristics of these studies. In the last two decades, there was an increase of more than 50% of publications. The Andean region was the most studied, while the other regions had few research. Lotic ecosystems were widely studied, unlike the lentic ones. The main land use management was agricultural. BMWP and diversity indices were the most used in the analyses. Further research is necessary in lentic ecosystems since they are being severely impacted by anthropogenic pressures, but were not assessed yet. Different regions that harbor diverse aquatic ecosystems (e.g. Amazon, Orinoquia) have been little studied. There is a critical need to revisit traditional indices, aiming at avoiding wrong assumptions about the relation between aquatic macroinvertebrates communities and water quality. For this purpose, it is recommended that assumptions and conditions required for using the indices, high taxonomic resolution, and new approaches, such as effective number of species, are taken into account.


Introduction
Water is the essential resource of life on Earth.All living organisms, including humans, depend on water for survival (Nizel Halder and Nazrul Islam, 2015).However, water quantity and water quality (WQ) have been affected in the last few decades by human-induced environmental issues such as climate change (Karl et al., 2009), population growth (Kirbi et al., 2003;Vorosmarty et al., 2000), land cover change, urbanization, industrialization, among others (Vorosmarty et al., 2010).In fact, it is expected an increase of eutrophication in almost all of the water surface in 2030 (UNESCO, 2015).
Water pollution is a serious threat to nature and human population.Its effects have severe impacts on aquatic communities (Zamora, 1998a), human health (Alrumman et al., 2016), food production (Kirby et al., 2003) and economic development (Nizel Halder and Nazrul Islam, 2015).Therefore, assessing WQ is critical for both conservation planning and policy making in order to protect this key resource (Ouyang, 2005).
Traditional evaluation of WQ has been done using physicochemical and microbiological variables (Roldán- Pérez, 2016).Recently, researchers have included macroinvertebrates to assess WQ in freshwater ecosystems.Aquatic macroinvertebrates (AM) are useful bioindicators of WQ due to their broad distribution, high abundance, short life cycles, easy sampling, and rapid response to anthropogenic pressure (Ríos-Touma et al., 2014;Gamboa et al., 2008;Roldán-Pérez, 1999).Hence, the United States of America and many countries of Europe have used these organisms for monitoring lotic and lentic ecosystems, being useful in their recuperation during the last 20 years (Roldán-Pérez, 2016).
In Latin America, the interest on biomonitoring and bioindication studies using AM has been growing and may probably increase in the future (Ramírez and Gutíérrez-Fonseca, 2014).In Colombia, the use of AM in the study of WQ started in the decade of 1970s with investigations that assessed the effect of industrial and domestic pollution on rivers and streams of Medellín (Antioquia) (Roldán-Pérez et al., 1973;Pérez and Roldán-Pérez, 1978).Since then, mathematical indices aiming to establish a relationship between AM and WQ have been widely used in the analysis of WQ in different aquatic ecosystems of the country.These indices include the Biotic index of Hilsenhoff (IBF) (Hilsenhoff, 1988), ETP (Klemm et al., 1990), BMWP-Col (Roldán-Pérez, 2003), ASPT (Roldán-Pérez, 2003) and diversity indices, such as Shannon-Wiener (1949) andSimpson (1949).Other approaches using AM in WQ, as for example, fuzzy logic and neural networks were implemented on waterbodies of Bogotá (Gutiérrez et al., 2002).Recently, indices, such as Biotic Integrity Index of Macroinvertebrates (BIIM) (Martínez-Rodríguez and Pinilla, 2014) and Aquatic Ecological Index based on freshwater (ICERN-MAE) (Forero et al., 2014), were created in order to include more variables for better assessment of WQ using AM in Colombian aquatic ecosystems.
Nowadays, there is a considerable amount of studies of AM as WQ bioindicators in Colombia, but the information is highly dispersed and, often, difficult to find, so that a systematic analysis of this issue is not only relevant but also needed.Gathering research items about this topic in the country will be useful to evaluate the progress of bioindication, define priorities in future research, and identify gaps of scientific knowledge.Hence, we conduct a systematic review to analyze scientific production related to evaluation of WQ in Colombian aquatic ecosystems using AM.We described the principal characteristics of the studies, including geographic information, land use management, ecological information, and methodological analysis.We identified biases and gaps in scientific production for future research on this topic.

Data compilation
This systematic review was carried out in June of 2017.To conduct the most complete possible compilation, literature was collected using the databases ScienceDirect, Redalyc, Web of Science, Scopus, SciELO, EBSCO and Google Scholar without restricting publication year.We searched for references that contained in the title, abstract, and/or keywords the following search terms with all the possible combinations: ("water quality" OR "bioindicators" OR "biological monitoring" OR "environment indicator" OR "biomonitoring" OR "bioindication" OR "biotic index" OR "biological quality" OR "pollution") AND ("macroinvertebrates" OR "aquatic invertebrates" OR "aquatic insects" OR "aquatic entomofauna" OR "benthic fauna") AND ("Colombia").The Google Scholar search produced 50 pages of results, which were completely revised.We also identified additional studies cited within this primary literature.
After eliminating the duplicated documents, we selected original articles, book chapters and theoretical references (i.e. reviews, viewpoints, comments).Gray literature, such as thesis, congress memories, and technical reports were excluded from the analysis.Thus, we found 121 references.We filtered them using two exclusion criteria.First, we excluded checklist, new records, and taxonomic papers because those studies had not the aim of assessing WQ.Secondly, we did not consider studies of diet, temporal distribution, and habitat preference, because those themes are not properly bioindication research.Although these studies are important for ecological knowledge of AM, the objective of those studies was not to assess water quality as such.In total, 109 references were selected for the systematic review.

Data analysis
For quantify data, a Microsoft Excel database was compiled, in which the following attributes were included: (1) Bibliographic information: year, type of publication, journal, institutional affiliation of each author.

Bibliographic information
From 1973 to 2017, 109 references were found (Appendix 1).More than 60% (n=68) were carried out in the last two decades, showing a remarkable increase of bioindication studies using AM in the country (Figure 1c).The type of publication was mainly original articles (n=93), followed by book chapters (n=9) and theoretical references (n=7) (Figure 1a).
Study sites comprise a wide variety of elevation ranges, with a slight tendency to decrease at higher elevations.Most of the study sites were located at elevations between 1000 and 3000 m (n=88, 70%).Fewer studies conducted in lowlands (between 0-1000m) and highlands (above 3000m) were found (n=29, 23% and n=8, 7% respectively).

Ecological information
Most of the studies focused on lotic ecosystems (n=88, Figure 3b), while lentic ecosystems have been notably less studied (n=10, Figure 3c).In relation to the context of the study sites, most of the references reported that land use management around the freshwater ecosystems were agricultural, followed by domestic and industrial.Sites that reported land use associated with mining, dam constructions and wood extraction were less reported (Figure 3a).

Methodological information
Half of the studies had a taxonomic resolution to genera level, followed by family level.However, only 2% of the publications have deepened into specific level, indicating that high taxonomic resolution for AM is scarce (Figure 4b).
Analysis methods were mainly based on biotic indices followed by diversity indices (Figure 4c).Seven biotic indices were found (Figure 4d).From these, BMWP was the most used, followed by ASPT and ETP, while the other four indices were rarely used (Figure 4d).Multivariate statistics, such as principal component analysis, non-metric dimensional scaling, cluster analysis, multiple regressions and correspondence analysis, were quite common.Fuzzy logic and network analysis were rare (Figure 4c).
Finally, other analysis included five main categories (Figure 4a).Within these, physicochemical analyses were widely used paired with AM indices for quantifying WQ.Other approaches included microbiological analysis and to a lesser extent: ichthyological analysis, plankton analysis, and stream channel morphology (Figure 4a).

Publication trends
The last decade showed a noteworthy increase in the studies of AM as bioindicators of WQ in Colombia, re-flecting an expanding research interest in this topic (Figure 1c).This increase is more evident in the period 1997-2000 and is largely related to the formulation and application of biotic indices, specially the BMWP which had adaptations to different departments like Antioquia (Roldán- Pérez, 2003), and Norte de Santander (Sánchez-Herrera, 2005).
In terms of author affiliations, the fundamental role of Colombian universities as producers of knowledge is remarkable (Bucheli et al., 2012;Figure 1b), which in most of the cases is done by Colombian researchers linked to national universities (Salazar-Acosta et al., 2010).These institutions will be crucial in fostering innovation and preparing the next generations to face and overcome the new research challenges (Aldana-Domínguez et al., 2017).

Advances in aquatic ecosystems
Lotic ecosystems, mainly of first order rivers, were the most evaluated, unlike lentic ecosystems such as lakes, wetlands, swamps, and hydroelectric dams (Figure 3B).Even though some physicochemical characteristics have been studied in lentic ecosystems, further research related to the grade of eutrophication and its effect on ecosystem function is lacking (Roldán- Pérez, 2009).AM could be useful for this purpose.Additionally, the evaluation of WQ in flood plain lakes (swamps), or "Ciénagas" as commonly known in Colombia, is important due to its socioeconomic relevance for the ecosystem services provided by fisheries and tourism (Montoya and Aguirre, 2009).Colombia harbors over 1900 swamps (Arias, 1985), but very few studies have evaluated WQ using AM in these ecosystems (Montoya and Aguirre, 2009).
Paramos are some of the least studied ecosystems in the world (Buytaert et al., 2011).Despite some studies assessing community assembly of AM in Paramos lagoons (Alba-Hincapié et al., 2016;Gómez et al., 2016;Posada-García et al., 2008), more research is needed in the evaluation of its WQ.Understand this ecosystem in terms of its hydrology may foster a better management of water resources (Flores-López et al., 2016).Paramos ecosystems are critical for water supply in local communities, regulating water resource during seasonal variations (Alba-Hincapié et al., 2016;Buytaert et al., 2006), and serving as the region's headwaters for downstream users (Flores-López et al., 2016).For instance, this ecosystem provides 95% of water to the capital city (Bogotá) (Buytaert et al., 2011).However, key aspects such as hydrological and biological processes remain poorly understood (Flores-López et al., 2016).
It is important to focus the research on lentic ecosystems because it is unknown how human impact could affect WQ in these ecosystems.Approximately 24% of the total area of Colombian wetland ecosystems have been used in livestock, agriculture, urbanization and mining (Patiño et al., 2016).In addition, Paramos ecosystem for the period 1970-2014 lost 15.9% of its original coverage due to paddock and agricultural land uses (Etter et al., 2015).Nevertheless, there is little data regarding the effect of anthropogenic pressures on the water resource.Further research is needed for evaluating the impact of landscape management on Colombian lentic ecosystems, and AM are great model organisms to do this.

Scientific production in Colombian regions
The hydrological distribution of water resources is not homogenous throughout the national territory, and it is subject to strong variations that determine the availability of the water resource (Roldán- Pérez et al., 2014).Study cases on WQ using AM were focused in one specific region: Colombian Andes (Figure 2).Andean region present the highest population density of the country, and is one of the most important in national economy (Roldán- Pérez et al., 2014;DANE, 2005).Particularly, rivers that belong to Magdalena-Cauca watershed were most investigated.This is not a surprising finding, taking into account that Magdalena-Cauca watershed harbor over 80% of the Colombian population, provides about 80% of the national Gross Domestic Product, and generate 84% of hydroelectric power of the country (Jiménez-Segura et al., 2014).
The other regions had few studies of AM as bioindicators of WQ (Figure 2).Some of the major rivers of the country, such as Amazonas, Caquetá, Vaupés, Guainía, Putumayo (Amazon region), Atrato (Pacific region), and Meta Rivers (Orinoquia region), remain poorly known in terms of their WQ, and are virtually virgins on the study of AM.On the other hand, although some scientific contributions of AM about spatial and temporal patterns have been done in Gorgona island (Gómez-Aguirre et al., 2009), and community assembly studies have been developed in Providence island (Cortés-Guzmán and Ospina- Torres, 2014), Insular region lacks bioindication research.
Most of the studies of WQ using AM were developed in elevations between 1000-3000 meters, but there were also some works above 3000m.Throughout this altitudinal gradient, different Andean ecosystems are found: Subandean forest (1000-2300m), Andean forest (2300-3200m), High Andean forest (3000-3500m), and Paramo ecosystem (over 3200m) (Castaño-Uribe, 2002).Concerning lowlands (0-1000m), few studies were done.It is important to focus further research on lowland aquatic ecosystems to know the effect of water discharges on aquatic ecosystems and its associated biological communities such as AM.

Trends in land use management
Due to the geographical localization, varying topography and climate regime, Colombia has one of the most important water resources of the world (Roldán- Pérez et al., 2014;Roldán-Pérez, 2009).Nonetheless, anthropogenic pressures affect hydrological cycle and WQ (Roldán-Pérez et al., 2014), and it is known that landscape transformation caused by agriculture, has detrimental effects on aquatic ecosystems (Murgueitio and Ibrahim, 2009).Within the studies analyzed, this type of land use was the most frequent (Figure 3a), as it is reported in Latin America (Ramírez and Gutíérrez-Fonseca, 2014).Agriculture may use chemical agents, contributing to water pollution, deforestation of riparian forest, soil erosion, and water flow modification (Chará et al., 2007).In this sense, agricultural activities may affect the structure and function of AM communities (Giraldo et al., 2014).
Urbanization and industrialization were the second land use type reported in the studies (Figure 3b).The urban and industrial expansion is increasing in Colombia (DANE, 2005), and such expansion has consequences in the discharge of wastewater, solid waste, and sediments in the waterbodies (Senhadji-Navarro et al., 2017;Nizel Halder and Nazrul Islam, 2015).
Since the period 2004-2008 mining licenses have been increasing in the country (Grisales-Gonzáles and Insuasty-Rodríguez, 2016;Insuasty-Rodríguez et al., 2013), but it is unknown many of its real effects on the WQ of different Colombian freshwater ecosystems.Carbon mining, for example, affects diversity of AM because of acidification and pollution of heavy metals, which reduce oxygen available for biological processes in the water (Zuñiga, 1986).AM could be useful as bioindicators of water pollution in gold mining and other metals that are highly exploited in the country, as well as the Andean region of South America (Ríos-Touma et al., 2014).
Finally, few studies were reported in hydroelectric dams (Figure 3a), which are important sources of energy in Colombia.In fact, 64.2% of electrical matrix in the country proceed from hydraulic energy (UPME, 2013).Further research using AM as bioindicators should assess the state of WQ altered by dam constructions, considering the government plans to duplicate the development of hydroelectric energy in the next ten years (UPME, 2013), to satisfy the population growth in the country (Jiménez-Segura et al., 2014).

Methodological analysis in bioindication using AM
Many approaches in the analysis of AM as bioindicators have been found.Biotic indices were quite common (Figure 4c); BMWP index was the most used due to its easy application, and frequently was complemented with ASPT index (Roldán-Pérez, 2003, Figure 4d).However, BMWP has many limitations.For instance, BMWP uses only qualitative data (i.e.absence and presence) without taking into account abundance or biomass of AM.This is a major issue, considering that change in abundances (or biomass) is an indicator of species response to environmental gradients.On the other hand, in BMWP the taxonomic resolution is to family level, ignoring the high variation of tolerance ranks within genera and species belonging to the same family; in this index, the scores are based on expert criteria, instead numerical and objective criteria (Forero et al., 2014) and may not accurately reflect WQ (Ríos-Touma et al., 2014).Moreover, although there are some adaptations in different locations (e.g.Sánchez-Herrera, 2005;Zamora, 1999), BMWP needs to be specific to the geomorphology, climate, and ecological attributes of the study region (Roldán-Pérez, 2016;Forero et al., 2014).
Physicochemical and microbiological analyses were the most used complementing AM evaluations (Figure 4a).During 1970During -1990, the main variables used to assess WQ were chemical (e.g.phosphate, nitrate, oxygen, pH), physical (e.g.temperature, conductivity), and microbiological variables (e.g.coliforms).However, these variables are not suitable for the evaluation of long temporal span in waterbodies.Due to their sedentary habits, AM are idoneous for this purpose (Gamboa et al., 2008).The integration of physicochemical and biological analysis may be critical for a better understanding of the WQ.In this sense, it is important to look forward to multimetric indices.Nonetheless, the missing information about environmental tolerance of different taxonomic groups make difficult the application of multimetric indices (Ramírez and Gutíérrez-Fonseca, 2014), but there are some important efforts that can be useful as multimetric indices, such as Aquatic Ecological Index based on freshwater (ICERN-MAE) (Forero et al.,2014).
Diversity indices were highly used for analyzing WQ using AM (Figure 4c).Most of the indices were applied as entropy measure, but not as effective number of species or genera.Furthermore, many studies showed clear misconceptions on the theoretical framework around diversity and its assumptions.It is important to make a scale transformation in diversity indices, such as Simpson and Shannon, for better ecological interpretation (Jost 2006(Jost , 2007;;Jost and Cutid-Medina 2016).In this systematic review, only 2% (n=1) of the publications used the effective number of species (also called "true diversity").Also, in most of the studies misinterpretations of diversity indices were found.Approximately 67% of the studies (n=32) made wrong inferences relating diversity and pollution, in fact, some authors mention a score that corresponds to a level of pollution in Shannon index (e.g.Pino-Chalá et al., 2003;Zamora, 1998b), and Margalef index (e.g.Pérez and Roldán-Pérez, 1978).The 25% (n=4) made numerical comparisons between sites but no statistical comparisons (e.g.t-test, CI 95%), and the 8% made random categories of high and low diversity (e.g.Zamora, 1998a;Longo et al., 2004).
However, some criticisms on the use of diversity indices as a measure of WQ have been done before (Jost and Cutid-Medina, 2016).For instance, Ramírez (2005) argue a spurious relation between pollution and diversity.Diversity indices are only descriptors of community structure, including species richness and distribution of its abundances (Ramírez, 2005).Nevertheless, diversity indices can be useful if are well applied and interpreted.It has been proposed to use the measure of effective number of species or genera, utilizing AM in which high taxonomic resolution is easy to achieve (i.e.species or genera level) (Jost and Cutid-Medina, 2016).
Regarding taxonomic resolution, biotic indices were applied considering the family level (e.g.BMWP, ASPT) or even the order (e.g.ETP) in most studies.Few studies determined their specimens at species level (Figure 4b).A possible explanation is that the Neotropical region has a great diversity of AM within the families, genera and even at intraspecific level, increasing the difficulty in taxonomic resolution.In addition, it is necessary to advance in the taxonomic study of larval stages of AM species (Springer, 2010), to allow accuracy about individual counts and therefore on bioindication of WQ.
With the advances of taxonomic knowledge, it would be important to modify biotic indices or create new indices, using genera level at least to be more accurate in the analysis of WQ and taking into account the new frontiers on the diversity measures (i.e.effective number of species).Other analysis like saprobic approach, which describes the tolerance of organisms in polluted conditions (Kolkwitz and Marsson, 1908), could be applied with additional data to species level, but this information is not available yet in the Neotropical region (Roldán-Pérez, 2016).

Concluding remarks
AM are useful as WQ bioindicators.Despite the remarkable increase of publications in the last years, further research is necessary in different aspects.Additional data is needed in lentic ecosystems; even though lakes, wetlands, swamps, and dams are facing several anthropogenic pressures, WQ of many of them have not been assessed yet.Andean region has most of the publications, but other important regions that harbor diverse and highly important aquatic ecosystems, such as Amazon, Orinoquia and Caribbean, have been little studied.
The main analysis methods were biotic indices, specifically BMWP, ASPT and ETP for its low taxonomic resolution and easy analysis.Diversity indices were quite common, but it is important to consider effective number of species or genera, assumptions of indices and better taxonomic resolution for data analysis.It is important to improve the taxonomic resolution for the application of other analysis, such as saprobic approach and multimetric indices.Nonetheless, high diversity of AM throughout the Neotropical region, as well as scarce knowledge of larval stages, make difficult the taxonomy of these organisms.

Figure 1 .
Figure 1.Bibliographic information related to evaluation of water quality in Colombian freshwater ecosystems, using aquatic macroinvertebrates: (A) type of publication, (B) institutional affiliation of each author, (C) publications in each period.

Figure 2 .
Figure 2. Number of studies related to evaluation of water quality in freshwater ecosystems, using aquatic macroinvertebrates, in each administrative region of Colombia between 1974-2016.

Figure 3 .
Figure 3. Limnological ecosystems and land use management: (A) main land use of the study site, (B) aquatic system (i.e.lotic and lentic), (C) ecosystem.

Figure 4 .
Figure 4. Methodological information in studies related to evaluation of water quality in Colombian freshwater ecosystems, using aquatic macroinvertebrates: (A) physicochemical, microbiological, ictiological, planktonic and stream channel morphology analysis, (B) taxonomic resolution, (C) data analysis, (D) biotic indices.