The SASS, NASS, TARISS, and ZISS sampling protocols are all applied in wadeable (<1 m depth, <20 m wide) rivers, while the ZISS protocol also includes a method for larger non-wadeable rivers (>20 m wide). These protocols are intended for use in low to moderate flows and are not to be used during high flow (flood) events, or in lentic systems such as wetlands and impoundments, or in estuaries or ephemeral rivers (Dickens and Graham, 2002). The sampling protocol of these four RBPS are similar since they were all derived from SASS. An overview of the sampling protocols is provided in Table 1. Importantly, sampling is undertaken per biotope, namely stones (stones in current and stone out of current), vegetation (marginal and aquatic), and gravel/sand/mud, with effort for each biotope standardized by time or area (Table 1). The OKASS sampling protocol was adapted from SASS and is applied in non-wadeable, deltaic aquatic biotopes, with sampling undertaken per biotope including marginal vegetation in current (channel), marginal vegetation out of current (lagoon), floating vegetation, submerged vegetation, and seasonally-inundated floodplain.

For existing RBPs, a taxon's tolerance to water quality impairment is reflected in its sensitivity weighting, with highly sensitive taxa assigned a weighting of 15, while highly tolerant taxa have a weighting of one. For all RBPs in Africa, sampled macroinvertebrates are identified to family level (or for a few taxa, e.g., Porifera, to a higher level; or a lower level, e.g., Baetidae and Hydropsychidae) on site using a magnifying glass. The total number of taxa in each RBP are 99 (SASS5), 93 (NASS2), 96 (TARISS1), 90 (ZISS1), and 58 (OKASS1). The number after the RBP name denotes the version, with SASS already in version 5, NASS in version 2, and the others in version 1. When the four RBPs developed for rivers are examined, 81 taxa are common to all (Table 2). Regional endemics such as those common in the south-western Cape, South Africa, and tropical taxa resulted in some taxa being absent from one or more RBPs.

SASS sensitivity weightings were initially derived using expert opinion and subsequently validated through field testing and correlation with impact. SASS underwent five revisions (SASS1 to SASS version 5) following extensive testing in several regions within South Africa by a number of river ecologists (Dallas, 1995, 1997; Chutter, 1998; Dickens and Graham, 2002) with more than 1,000 SASS assessments undertaken during this testing phase. This facilitated an iterative approach to the development of SASS, with new taxa added and sensitivity weightings adjusted based on observed data. Since then, more than 12,162 SASS assessments have been undertaken in South Africa (Freshwater Biodiversity Information System (FBIS), 2020) and it forms the backbone of river health assessment in South Africa. In comparison, NASS comprised approximately 50 assessments on the perennial rivers in Namibia (pers comm. R Palmer), TARISS comprised 101 assessments on 85 rivers in four freshwater ecoregions of Tanzania (Kaaya et al., 2015), ZISS comprised 151 assessments on 95 rivers across the country (Dallas et al., 2018), while OKASS comprised 103 assessments on 54 deltaic sites in three regions of the Okavango delta (Dallas, 2009). For all of these RBPs further use of the respective indices has taken place and will most likely be used to further improve the index. When an RBP is used in a new region, as part of the validation process, sensitivity weightings of existing taxa need to be checked and sensitivity weightings of new taxa assigned. Palmer and Taylor (2004) assigned sensitivity weightings to new taxa based on observed data and similarity with existing taxa. Canonical analysis of principal coordinates has been used to predict sensitivity weightings of new macroinvertebrate taxa along the disturbance gradient (Kaaya et al., 2015), as well as correlation of occurrence to impact ratings, evaluation of closely related SASS families, known life-history modes, and anatomical adaptations (Dallas et al., 2018).

To evaluate the efficacy of the biotic index to demonstrate changes in river water quality, data on anthropogenic activities and ecosystem disturbance need to be collected at each site by assessing catchment, channel and habitat impacts. Both TARISS (Kaaya et al., 2015) and ZISS (Dallas et al., 2018) utilized versions of the site characterization protocol (Dallas, 2005) and Index of Habitat Integrity (Kleynhans, 1996), developed as part of the SASS testing phase in South Africa. This protocol evaluates the quantity and severity of anthropogenic impacts at a site and integrates potential impacts into an index of habitat integrity; in addition to assessing water quality impacts, modification to the channel, condition of the local catchment, and land-use (Dallas, 2005). For TARISS, levels of human disturbance across sites were derived by evaluating local catchment disturbance, instream and riparian habitat integrity (Kaaya et al., 2015). In Dallas et al. (2018) distinction between “impacted” and minimally impacted (or “unimpacted”) reference sites were used to generate thresholds of impact for each of the groups of impact. For OKASS, potential anthropogenic disturbances were used to calculate a Human Disturbance Score for each site (Dallas, 2009). To evaluate the ability of the biotic index to detect impacts and disturbance, metrics are correlated with the disturbance gradient to determine how each metric responds to the disturbance gradient, and to test differences among metrics from sites classified as impacted or reference.

Dallas (2004a), Kaaya (2015), and Dallas et al. (2018) demonstrated that macroinvertebrate assemblages differed among river types and recommended the inclusion of a hierarchical spatial framework within which bioassessment data is interpreted. SASS, TARISS, and ZISS, all use two-level hierarchical spatial frameworks to offer geographic partitions within which macroinvertebrate assemblages are expected to be similar, thereby assisting in interpretation of bioassessment data (Table 3). Level I relate to broad geographic regions, while level II relates to longitudinal zonation of river systems. In South Africa, for Level I, 31 ecoregions were derived from vegetation and terrain, with inclusion of geology, soil, altitude, rainfall, air temperature, and runoff variability (Kleynhans et al., 2005). For Level II, simplified geomorphological zonation was used to differentiate rivers into Upland or Lowland (Dallas, 2007a), based on research that demonstrated macroinvertebrate assemblages are typically divided into upland and lowland assemblages, with little differentiation at the finer level, e.g., Mountain Stream vs. Upper Foothill (Dallas, 2004a). In Tanzania, for Level I, Kaaya (2015) used freshwater ecoregions of the world (Abell et al., 2008) and five climatic zones (Indeje et al., 2000) to divide the country into 12 ecoregions based on hydrological (catchment) boundaries and climatic characteristics. For the Level II, 12 landform features and three slope classes were used to generate a geomorphologic classification (Kaaya, 2015). In Zambia, for Level I, freshwater ecoregions of the world (Abell et al., 2008) were used, and for Level II, stream order (“high” = stream orders 7–9 and “low” = stream order 3–6) was used (Dallas et al., 2018). Further partitioning of spatial variability is also included at biotope level, with all RBPs undertaking sampling for each biotope separately, and in some cases also interpreting metrics per biotope (e.g., ZISS). In comparison, examination of spatial variation of aquatic macroinvertebrate assemblages in the Okavango Delta (Dallas and Mosepele, 2007, 2020) showed that assemblages did not vary significantly amongst different region of the Delta, although macroinvertebrate assemblages differed amongst aquatic biotopes (Table 3). On this basis, Dallas (2009) proposed preliminary data interpretation guidelines based on two dominant deltaic habitats, namely marginal vegetation in current, and marginal vegetation out of current.

Temporal variability was only examined for SASS (Dallas, 2004b) and OKASS (Dallas, 2009), with the latter focusing on high-water level period (July) and the low-water period (October), which is more appropriate for the deltaic biotopes. The influence of sampling season on macroinvertebrate assemblages, taxon occurrence, and SASS Scores was investigated and seasonal variability was shown to be more prevalent in some regions of South Africa and some biotopes (Dallas, 2004b), with SASS Scores in the stones in current biotope varying seasonally, inter-annually and in response to wet and dry cycles, as well as antecedent flow events, water temperature and abundance of benthic algae (Palmer, 1997). For OKASS, although more taxa were generally recorded during October, the low-water period, compared to July, the high-water period, this was not statistically significant (Dallas, 2009).

SASS data interpretation guidelines are most advanced for South Africa, with guidelines provided for each “ecoregion-geomorphological zone” using the relationship between SASS Score and ASPT (Dallas, 2007a; Dallas and Day, 2007). Specifically, the number of biotopes sampled was positively correlated with SASS Score and number of taxa, while ASPT was negatively correlated with the number of biotopes (Dallas, 2007b). Data interpretation guidelines were not developed for NASS (Palmer and Taylor, 2004) as Namibia only has five perennial rivers, namely the Orange, Kunene, Okavango, Zambezi, and Chobe, while NASS has only been used in the north-eastern region. Data interpretation guidelines have not yet been developed TARISS (Kaaya et al., 2015). Provisional data interpretation guidelines have been produced for ZISS, although these require further testing and validation (Dallas et al., 2018). ZISS guidelines are based on metric scores (ZISS Score, Number of Taxa and ASPT) generated for reference sites within each “ecoregion-stream order-biotope” combination (e.g., Zambezian Headwaters-low order-stones). Dallas (2009) proposed preliminary data interpretation guidelines based OKASS Score for two deltaic biotopes, namely marginal vegetation in current, and marginal vegetation out of current.

The rationale for sampling biotopes separately is to ensure more accurate data interpretation since differences in availability of biotopes at a site may affect macroinvertebrate assemblages, given biotope preferences of some macroinvertebrate taxa (Dallas, 2007b). The availability of biotopes typically varies longitudinally down a river in response to broad geomorphological characteristics, with upper reaches dominated by stones, while lower reaches may only have vegetation and sand (Dallas, 2004a). When an RBP is being considered for use in a region outside the country where it was developed, it is important to consider the river systems prevalent in the region, including the variety of aquatic biotopes across rivers in the region. Biotopes sampled for all RBPs are dependent on which biotopes are present at the site. Where biotopes of the region resemble those of existing RBPs, then the sampling protocol can be adapted without much revision. For example, for ZISS, Dallas et al. (2018) concluded that stones and vegetation biotopes were more reliable than gravel/sand/mud in differentiating impacted from reference sites. Similarly, Dallas (2007b) noted that gravel/sand/mud biotope added very little to the SASS Scores or number of taxa in SASS. The earlier RBPs were limited to wadeable rivers, but since the development of OKASS and ZISS, non-wadeable, larger rivers, and deltaic biotopes may also be sampled. It should be emphasized that if the sampling protocol of an RBP is not followed (as summarized in Table 1), then it is not legitimate to assign sensitivity weightings to taxa recorded at a site to generate metrics for interpreting the impact of water quality impairment or deriving an ecological condition for the site. In particular, a result generated from a chemically preserved sample is not a legitimate SASS/NASS/TARISS/ZISS/OKASS result. Typical examples of where a protocol has not been adhered to include laboratory identification of preserved taxa instead of time controlled field-based identification, use of surber or box sampling instead of kick sampling, and only sampling one biotope at a site. The strict adherence to established sampling protocols is to ensure quality control and standardization so that results may be compared (Dickens and Graham, 2002).

Variation in geology and climate may influence physico-chemistry of river water, which may affect the distribution and sensitivity of macroinvertebrate assemblages (Day and King, 1995). When calibrating and validating a biotic index for a new region, it is important to undertake extensive sampling of macroinvertebrate assemblages across a range of sites in the region so that the full variety of taxa present in the region may be established. This is often done iteratively as a community of biomonitoring practitioners is formed within a region. Normally, when developing a RBP for a new region, macroinvertebrate samples collected from each biotope using the RBP would be preserved, and the identification of each taxon confirmed in the laboratory. This also facilitates the creation of a reference collection of aquatic macroinvertebrates for the region, which is a particularly useful resource for new biomonitoring programmes. Extensive sampling and taxonomic confirmation enable the detection of new taxa, not previously included in existing RBPs, including regionally endemic taxa, and provides evidence of the absence of other taxa in a region. Unfortunately, the development and adaptation RBPs and biotic indices is often hindered by poorly known taxonomy (e.g., tropical East Africa, Elias et al., 2014a, Ochieng et al., 2019). Fortunately, the RBPs described in this paper are mostly family level, which has greater taxonomic confidence than genus or species. Indeed, the use of predominantly family-level taxonomy in RBPs is a prerequisite since they are field-based.

The full range of taxa present may differ within a region, depending on latitudinal differences and regional endemism. The majority of taxa in the four RBPs developed for rivers were common to all (81 taxa), with a few regional endemics and less common taxa comprising the balance. Some taxa (Dicercomyzidae, Ephemerythidae, Machadorythidae, Sisyridae, Curculionidae, Sciomyzidae, Stratiomyidae, Ampulariidae, Bithyniidae, and Mutelidae) included in NASS, TARISS, and ZISS should ideally be included in future versions of SASS as these taxa have been recorded in the South Africa, and a further revision of SASS is likely. In addition, a recent study in Uganda noted five new taxa to be included in a modified TARISS, namely Chordodidae, Ptilodactylidae, Aspidytidae, Leptopodidae, and Paraecnomidae, which were not included in the TARISS (Tumusiime et al., 2019), although the latter is not a recognized family and is included in Ecnomidae. Importantly, prior to any new taxon being included in a RBP developed for a new region, it is recommended that all identifications be confirmed by a recognized institute and their taxonomic classification be verified on the taxonomic backbone of the Global Biodiversity Information Facility. Taxa excluded from the list of taxa during the validation process for a new region, are normally because they were not sampled in the new region. In some instances, this may be due to limited sampling or because the taxon is rare, or in other cases it may be because the taxon does indeed not occur in the region, such as those taxa endemic to the south-western Cape, South Africa. This emphasizes the importance of undertaking extensive sampling across the region to ensure that all potential taxa are sampled and included in the updated list of taxa for a new region. It is also useful to calculate the frequency with which each taxon is recorded in a region to provide an indication of taxon rarity.

When a RBP and biotic index are developed for a new region, attention should be given to validating the sensitivity weightings assigned to each macroinvertebrate family in the region. Existing indices apply different weightings for two families (Baetidae and Hydropsychidae) based on the number of species within each family sampled and recorded at a site. Families that have many species may exhibit a wider range of within-family tolerance compared to families with few species (Bonada et al., 2006). As part of the validation process, sensitivity weightings of existing taxa need to be checked and sensitivity weightings of new taxa assigned based on correlation with disturbance gradients, sensitivity weightings of closely related families, known life-history modes, anatomical adaptations, and expert knowledge. It is recommended that the RBP be assigned a version to facilitate keeping track of revisions and that an iterative approach be adopted when developing a RBP for a new region.

Biotic indices should be able to detect changes in water quality and ecological condition of the aquatic ecosystem they are designed to be used in. As part of the regional validation process it is necessary for the index to be applied across a range of sites exhibiting a gradient of impacts. During the testing of the index, and undertaken concurrently with sampling of macroinvertebrates, data on anthropogenic activities, and ecosystem disturbance need to be collected at each site by assessing catchment, channel, and habitat impacts. Strong correlations between the metrics and human disturbance provides confidence in the biotic indices to detect impacts and disturbance. The relationships between metrics and disturbance can be evaluated for each biotope (Dallas et al., 2018) and for the site as a whole (Kaaya et al., 2015). Recently, Tumusiime et al. (2019) showed that macroinvertebrate assemblages differed between test (= impacted) and reference sites, which provided confidence in the ability of the TARISS to distinguish impacted and reference sites, thus validating the efficacy of TARISS in Uganda.

Lotic systems are intrinsically heterogeneous with spatial and temporal variability occurring at multiple scales. Spatial variability may result from catchment-scale factors such as altitude, geology, channel slope; site-scale factors such as canopy cover, stream width, stream depth; and habitat-scale factors such as substratum, biotope availability, hydraulics (Dallas, 2007c). These factors may effect macroinvertebrate assemblage structure and an understanding of spatial variation in macroinvertebrate assemblages and the use of biotic indices based on these assemblages, is thus needed. Spatial frameworks are often used to overcome this intrinsic spatial variability and improve the reliability of data interpretation. The hierarchical spatial frameworks developed for SASS, TARISS, and ZISS, facilitated partitioning of intrinsic spatial variability, although further testing has been recommended for TARISS (Kaaya, 2015). The inclusion of biotope level sampling ensures that substrate, which has been identified as an important predictor for classification of macroinvertebrates in rivers (Dallas, 2007c), is considered. During the initial validation of the SASS protocol, the issue arose as to whether to combine samples based on substrate similarities (e.g., all stone, all vegetation) or hydraulic similarities (e.g., all in-current samples), with the former selected for SASS. Subsequent comparison of two RBPs, SASS, and the Iberian Peninsula (IB-protocol), which combined hydraulic biotopes (Prat et al., 2000), confirmed the similarity of the bioassessment results and the ability of each RBP to detect water quality impacts even though the RBPs used different biotope combinations (substrate vs. hydraulic) and had different sampling equipment (mesh diameter), sampling and laboratory processing methods (Bonada et al., 2006).

Temporal variability in macroinvertebrate assemblages may occur in response to seasonal variability in factors such as water temperature (Dallas, 2008), biotope availability (Armitage et al., 1995), and stream flow (McElravy et al., 1989). Given that seasonal variability was prevalent in some regions of South Africa and some biotopes (Dallas, 2004b), Dickens and Graham (2002) recommended that season be factored into SASS data interpretation, as some natural intra- and inter-annual variation in macroinvertebrate assemblages is likely. Kaaya et al. (2015) recognized the potential influence of seasonal variability on TARISS and recommended that it be examined for each river type in Tanzania.

Developing appropriate data interpretation guidelines is perhaps the most challenging aspect when developing RBPs and biotic indices for a new region. However, it is critical to ensure that impacts reflected in the metrics are real and not merely a consequence of intrinsic spatial and temporal variability of macroinvertebrate assemblages, as this may affect our ability to interpret bioassessment data. An understanding of variability is important to facilitate the establishment of reference conditions, including expected macroinvertebrate assemblages (Dallas, 2004a,b; Dallas and Day, 2007). Reference conditions represent the natural or least-impacted condition for a particular type of river, and are used as a measure with which impacted sites are compared. Highly variable systems may lead to patchiness of taxa, which need to be considered when developing the mechanism for interpreting bioassessment data (Dallas, 2004a,b; Dallas, 2007b; Dallas and Day, 2007). The basis for interpretation of metrics for a particular sample is understanding the “natural” variability of metrics from the site or similar sites (i.e., variability of reference sites) and whether the metric for a test site falls within (unimpacted = reference) or outside (impacted) that variation.

Normally metrics are used for assessing ecological condition, either as tables or graphs showing different ecological categories. Embedding data interpretation within a spatial framework, as for SASS, provides a robust, easy to use system for evaluating change in water quality and ecological condition. Spatial and temporal variability may be accounted for by defining the reference condition as a band (Dallas and Day, 2007). Further, by utilizing the relationship between SASS Score and ASPT, between-site variation in the availability of biotopes is taken into account (Dallas and Day, 2007). The validity of this relationship for metrics derived from other RBPs and biotic indices in other regions, however, would need to be tested. Kaaya et al. (2015), while demonstrating that the validated TARISS technique is a dependable method for rapid bioassessment of rivers in Tanzania, advised that interpretation guidelines for each river type still need to be developed, using the “river type specific reference condition” approach. Since TARISS was only validated in two Tanzanian ecoregions, further validation in the other ecoregions is needed before it qualifies as a national biotic index.