Developing and applying a macroinvertebrate‐based multimetric index for urban rivers in the Niger Delta, Nigeria

Abstract Urban pollution of riverine ecosystem is a serious concern in the Niger Delta region of Nigeria. No biomonitoring tool exists for the routine monitoring of effects of urban pollution on riverine systems within the region. Therefore, the aim of this study was to develop and apply a macroinvertebrate‐based multimetric index for assessing water quality condition of impacted urban river systems in the Niger Delta region of Nigeria. Macroinvertebrate and physicochemical samples were collected from 11 stations in eight river systems. Based on the physicochemical variables, the stations were categorized into three impact categories namely least impacted stations (LIS), moderately impacted stations (MIS) and heavily impacted stations (HIS). Seventy‐seven (77) candidate metrics were tested and only five: Hemiptera abundance, %Coleoptera + Hemiptera, %Chironomidae + Oligochaeta, Evenness index and Logarithm of relative abundance of very large body size (>40–80 mm) were retained and integrated into the final Niger Delta urban multimetric index (MINDU). The validation dataset showed a correspondence of 83.3% between the index result and the physicochemically‐based classification for the LIS and a 75% correspondence for the MIS. A performance of 22.2% was recorded for the HIS. The newly developed MINDU proved useful as a biomonitoring tool in the Niger Delta region of Nigeria and can thus be used by environmental managers and government officials for routine monitoring of rivers and streams subjected to urban pollution.

exception, as the majority of urban rivers in the region are seriously impacted (Arimoro & Ikomi, 2008). Despite the growing urban pollution in the Niger Delta region, no biomonitoring tool exists for assessing and monitoring the extent of the effects of urban pollution on riverine ecosystems. The development of an appropriate biomonitoring tool can contribute to managing pollution through effectively monitoring and assessing urban pollution effects on riverine biota.
Macroinvertebrates are particularly useful for index development because they occupy an important position as consumers, can easily be collected, have high diversity, and are differentially sensitive to a gradient of pollution (Bonada et al., 2006;Odume et al., 2012).
While the majority of macroinvertebrate-based multimetric indices are developed for general water quality (Pešić et al., 2019;Petriki et al., 2017;Stevenson et al., 2013), the intention in this study is to develop a pollution type-specific multimetric index for assessing urban rivers water quality impairment in Nigeria. The significance of developing an index specific for urban pollution is based on the realization that Nigeria is urbanizing rapidly, and rivers in the Niger Delta region, in particular, suffer from serious urban pollution effects. Therefore, the aim of this study is to develop and apply a macroinvertebrate-based multimetric index suitable for assessing and monitoring ecological impairments of urban rivers in the Niger Delta region of Nigeria. This study is the first regional macroinvertebrate-based multimetric index in Nigeria, where studies on biomonitoring methods development are still scanty. The present study thus adds to the few existing studies on macroinvertebrates multimetric indices for biomonitoring of freshwater ecosystems in sub-Saharan Africa (e.g., Aura, Kimani, Musa, Kundu, & Njiru, 2017;Chirwa & Chilima, 2017;Edegbene et al., 2019;Lakew & Moog, 2015;Mereta et al., 2013;Odume et al., 2012).

| The study area
The Niger Delta occupies an area of approximately 70,000 km 2 in the southern tip of Nigeria. The area is characterized by mangrove swamps, wetlands and inland waters (Umoh, 2008). Biodiversity within the region is high (Adekola & Mitchell, 2011). The region supports a wide range of subsistence inland fisheries and wood logging (Zabbey, Erondu, & Hart, 2010). There are two main seasons: the wet and dry season within the Niger Delta Edegbene & Arimoro, 2012). The wet season is characterized by extensive and intensive rainfall, which begins in April and ends in September. The dry season is characterized by high temperature, usually between 25°C and 35°C. The dry season starts in October and ends in March. The region is known for oil exploration and exploitation. Drainage system in urban cities within the region is poor, and rivers are often impacted by untreated wastewater, storm water return flow, and run-offs from informal settlements. All of these imply that urban rivers and streams within the regions are being impacted at an alarming rate.

| Study river systems
Eight river systems draining urban landscape in Edo and Delta States within the Niger Delta Region were selected for the study.

| Macroinvertebrates and physicochemical sampling
Macroinvertebrates and physicochemical data collected from 2008 to 2012 (five years) were used for the development and validation of the index. Samples were collected monthly in the 11 stations for two seasons, wet season (April-September) and dry season (October-March). Macroinvertebrate data collected from 2008 to 2010 were used for the development of the multimetric index, and those from 2011 to 2012 were used for its validation.
Macroinvertebrates samples were collected at each sampling station for a period of 3 min per biotope. Samples of macroinvertebrate collected from vegetation, sand, silt, mud, and stones were grouped as composite samples and thereafter preserved in 70% alcohol for onward transfer to the laboratory for sorting, identification, and enumeration. Macroinvertebrates were identified to the family level under a stereoscopic microscope at ×10 magnification.
Physicochemical data were also collected alongside the biological data throughout the sampling period. Physicochemical parameters analyzed for this study were as follows: water temperature, depth, flow velocity, electrical conductivity (EC), pH, dissolved oxygen (DO), five-day biochemical oxygen demand (BOD 5 ), nitrate, and phosphate. A calibrated stick was used in determining the depth of the water in meter. Flow velocity was measured according to Gordon, McMahon, and Finlayson (1994) method. Dissolved oxygen (DO) was measured using dissolved oxygen meter (YSI 55 dissolved meter), while water temperature, pH, and EC were determined using a portable HANNA HI 9913001/1 instrument. Nitrate, phosphate, and BOD 5 were determined in the laboratory using APHA (1995) methods.

| Delineation of stations along an urban impact gradient
The 11 stations in the eight river sampled were delineated along an urban impact gradient into three impact categories namely least impacted stations (LIS), moderately impacted stations (MIS), and heavily impacted stations (HIS; Table 1). This was achieved by correlating the physicochemical data with the selected river stations using principal component analysis (PCA; Figure A1). Stations strongly correlated with physicochemical indicators of urban pollution such as high nutrients, BOD 5 , and high EC were deemed heavily impacted, and those posi-

| Metrics selection for multimetric index development
Seventy-seven (77) candidate metrics were compiled (Table A1), which takes into account various community structure of macroinvertebrates including measures of absolute abundance, composition, richness, diversity, and traits (Baptista et al., 2007;Edegbene et al., 2019;Fierro, Arismendi, Hughes, Valdovinos, & Jara-Flores, 2018;Mereta et al., 2013;Odume et al., 2012). Trait information was obtained from Krynak and Yates (2018) and Odume, Ntokolo, Akamagwuna, Dallas, and Barber-James (2018). A fuzzy coding system of 0-3 affinity scores was used to award trait information to macroinvertebrate taxa (Chevenet, Dolédec, & Chessel, 1994). A score of 0 was awarded to a taxon if the taxon has no affinity to the trait attribute, 1 was awarded if the affinity was low, 2 if the affinity was moderate, and 3 if it was high (Chevenet et al., 1994). Measures of abundance was included as part of candidate metrics to be tested in order to represent all component of macroinvertebrate community structures.

| Index development
Five steps were followed in developing the index, and these include subjecting all candidate metrics to (a) sensitivity test, (b) seasonality test, (c) redundancy test, (d) integration of selected metrics into the multimetric index, and (e) index validation.

| Sensitivity test
Candidate metrics were tested for their potential to discriminate between the LIS from the MIS and HIS. Box plots were used to visualize the metrics. Two levels of discrimination were considered satisfactory. First, a metric was deemed sensitive if there was an overlap between the interquartile ranges (IQRs) of the MIS and HIS, and those of the LIS, but the medians are outside of the interquartile ranges (Edegbene et al., 2019;Odume et al., 2012). Second, a metric was considered sensitivity if the IQR of the LIS do not overlap with those of the MIS and HIS (Edegbene et al., 2019;Odume et al., 2012).
Metrics that met all or any of the criterion were selected for further testing.
Selected metrics based on the box plot visualization were further tested for significant differences using the Mann-Whitney (U) test.

Mann-Whitney (U) test was used because Kolmogorov-Smirnov test
indicated that metrics were non-normally distributed. Metrics exhibiting a significant difference between the LIS, and the MIS and HIS at p < .05 were retained for further analysis (Barbour et al., 1996). Box plots were done using Statistica version 13.4.14 (TIBCO Software

| Seasonality test
Metrics that were deemed sensitive after confirmation with Mann-Whitney test were further subjected to seasonality test for seasonal stability. Box plots were used to visualize metrics' seasonal stability, TA B L E 1 Categorization of stations into potential impact categories along the gradient of increasing urban pollution and the Kruskal-Wallis test was further used to confirm seasonally stable metric (Baptista et al., 2007). Only metric data from the least impacted stations were used for seasonality test to avoid the confounding effect of pollution on seasonal variation of metrics (Edegbene et al., 2019;Odume et al., 2012).

| Redundancy test
Redundant metrics convey the same or similar information (Odume et al., 2012). Spearman's rank correlation coefficient (r) was performed on the seasonally stable metrics to explore co-linearity between the metrics. Metrics with correlation values (Spearman's r ≥ .78, p < .05) were considered redundant (Edegbene et al., 2019).
Non-redundant metrics were selected for integration to the multimetric index. Where two or more metrics were redundant, only one of such metric was selected for inclusion in the multimetric index (Edegbene et al., 2019).

| Integration of the metrics into a multimetric index
Prior to integration, selected metrics were standardized by using the minimum value, lower quartile (25%), mid-quartile (50%), upper quartile (75%), and maximum value of each metric datasets according to the method described in Baptista et al. (2007). Lower, mid, and upper quartiles were computed with Microsoft Excel, 2010 version.
Metrics that were predicted to increase with increasing urban pollution were assigned a score of 5 if the metric value was below the upper quartile (75%) of the LIS, a score of 3 was awarded, if metric value is above the 75%, and a score of 1 is awarded, if the metric value is above the maximum value of the LIS. On the other hand, for metrics that were predicted to decrease with increasing urban pollution, a score of 5 was awarded if metric value of LIS is greater than or equal to lower quartile (25%), a score of 3 was assigned, if the metric value was between the minimum value and <25% of the LIS, while score of 1 is assigned, if the metric value is lower than the minimum value of LIS.

| Sensitivity and seasonal stability tests
Of the 77 candidate metrics, only 26 metrics satisfactorily discriminated between the LIS, and the MIS and HIS (Table A2). In all, after subsequent analysis, only five metrics were integrated into the final index, and their discrimination potential are visualized in Figure 2.

| Redundancy test
Apart from the trait measure: very large body size (log VeL), all other sensitive and seasonally stable metrics were found to be redundant with one another (Table A3). However, given that only 15 metrics have been retained thus far and 14 were redundant, and they represent different measures, four of the 14 redundant metrics were retained in addition to log VeL. The four metrics selected in addition to log VeL were Hemiptera abundance, %Coleptera + Hemiptera, %Chironomidae + Oligochaeta and Evenness index (Table A3).

| Development of the multimetric index
To develop the multimetric index, the minimum value, lower quartile (25%), mid-quartile (50%), upper quartile (75%), and maximum value of each metric for the least impacted stations (LIS) metric assemblages values were used as thresholds for calculating the metric scores ( Table 2). The multimetric index was computed by summing the scores of the five metrics component, and the index value range (5-25) since five metrics were used (5 × 5 = 25). The index value range then reflect five water quality categories as shown in Table 3.

| Validation of the multimetric index
The index validation results showed that 25% of the times, stations designated as LIS had very good water quality, and 58.3% of the times, stations designated as LIS had good water quality (Figure 4).
Since none of the station could be said to be pristine, the agreement

| Relating the selected metrics to physicochemical variables
The first RDA axis explained 86.98% of the ordination plot, while the second axis explained 13.02%. The Eigen value of the first axis was higher, 6.409 compared to the 0.40918 Eigen value of the second axis. There was no significant difference in the two RDA axes correlation with metrics and the physicochemical variables (p > .05) as revealed by the Monte Carlo test at 999 permutation.
Dissolved oxygen strongly correlated with Evenness index and % Coleoptera + Hemiptera ( Figure 6). Logarithm of relative abundance of very large body size was positioned at the centre of the RDA triplot and was correlated with depth. Five-day biochemical oxygen demand and EC were strongly correlated to % Chironomidae + Oligochaeta at the HIS. Hemiptera abundance was correlated to water temperature and flow velocity at LIS (Figure 6).  unstable because of the difficulty of disentangling variation occasioned by natural seasonal dynamics from those occasioned by anthropogenic activities.

| D ISCUSS I ON
One of the five metrics integrated into the final MINDU was trait measure, that is, very large body size (>40-80 mm). Organisms with body size ranging between >40 and 80 mm have proved highly sensitively to urban pollution and was non-redundant with the rest of the taxonomic metrics. Abundances of very large-bodied macroinvertebrates have been hypothesized to decrease in response to environmental stress because they are often associated with long reproductive cycle and fewer offspring per reductive event compared to small bodied individuals, which often reproduce rapidly (Castro, Dolédec, & Callisto, 2018;Serra et al., 2017;Townsend & Hildrew, 1994). Studies testing metrics for integration into multimetric indices have often ended up with one or two trait-based metrics in the final indices, indicating that the present study, which found only a single trait to be highly sensitive and non-redundant was in accordance with most other studies (e.g., Baptista et al., 2007;Fierro et al., 2018;Gieswein et al., 2019;Ntislidou et al., 2018). The validation of the performance of the developed MINDU with separate datasets revealed that the index performed better for LIS and MIS compared with the HIS. The relatively good performance of the index for the LIS and MIS stations indicates that using the index may not lead to under or over protection, whereas the poor performance of the index for the HIS could be that pollution at these stations are seasonally mediated such that macroinvertebrate recovery and recolonization are rapid, reducing the cumulative effects of pollution. Even though seasonal stability was tested for during the selection of the metrics, it appears that a "flushing effect" aggravated the effects of urban pollution during the wet season. During the wet season, water quality at the HIS was generally poor, compared with the dry season. It is postulated that increased urban storm water run-off, as well as run-off from settlements, carrying pollutants may have led to the poor water quality during the wet season at the HIS.
Similar findings have been reported by Speak, Rothwell, Lindley, and Smith (2013) that increased urban run-off due to increased precipitation led to increased pollution of riverine ecosystems. Water quality at the HIS seems to recover during the dry season and thus mediating the overall performance of the developed index. The implication therefore is that monitoring need to be structured to take account of seasonality, and data interpreted taking into account the season-mediating effects of urban pollution.
The developed MINDU performed better in the dry season than in the wet season except for the LIS. The reason for the better performance of the index during the dry season could be attributed to reduced urban run-off during the season. Urban run-off is one of the major factors influencing water quality of rivers in the Niger Delta.
In addition, heavy rains have impact on water quality because debris and other pollutants are carried into urban river systems during down pour. In contrast to the findings in the present study, is the work of  Edegbene Ovie is also acknowledged for her technical assistance.