**4.3. The BMWP index and BMWP water quality classes**

While the WQI scores showed small differences among study sites, the BMWP index showed a wider variation, making evident that the latter has a higher level of sensitivity, as it was able to register fine and important differences between study sites, which were not evidenced by the WQI (**Figure 5**). The sensitivity of the BMWP index is related with the procedure to assign the values for the water quality classes. For this step, the reference sites are indispensable. The BMWP index was able to detect the pristine condition of "Las Truchas," the clean river reference site, located inside an undisturbed oak forest, showing the highest richness score of this study site. Furthermore, this site showed the presence of the family Perlidae, a bioindicator of excellent water quality. Lakew and Moog [30] stated that a reference site must meet both abiotic and biotic requirements; the same authors consider a reference site as the least impaired site characterized by selected physical, chemical and biological characteristics. Las Truchas site reached the higher *Pcq* and WQI scores, the less impairment of adjacent land use and also meet the biological requirements of the higher richness, as well as the higher scores of the BMWP index. In consequence, the site "Las Truchas" meets the requirements of a reference site. The BMWP scores obtained in this study also showed a wide seasonal fluctuation, particularly for Las Truchas site. The BMWP fluctuation provided for Las Truchas showed scores from 69 to 110, these data conformed the set of values for obtaining the median value that correspond to the "Excellent" quality category, from which range values for all water quality classes were assigned. The method used for the allocation of scores to each water quality class was proposed by Pond and McMurray [21]. These procedure displayed a suitable assignation of the water quality categories and consequently reducing the probability of commit errors (Type I and II) [35] for the categorization of the study sites.

The bioindication values of the aquatic macroinvertebrate families do not always match completely from one country to another, which can be due to the variations in taxonomic tolerances of each basin and biogeographic region and to the method of assigning bioindi‐ cation values, which in most cases is unknown, generating some uncertainty in the scores assigned to each family. However, there are families of aquatic macroinvertebrates charac‐ teristic of very healthy environments, as is the case of the Perlidae with a score of 10; or in extremely hostile conditions, the midges and lumbriculids with score values of 1 [14]. In the present study, the wide distribution of families in different intervals of *Pcq*, coupled with the abundance classes, allowed us to generate a bioindication value that is neither overestimated nor underestimated, as was evidenced by the degree of adjustment of the models described above.

Our results show that the BMWP index has a good discriminating capacity; nevertheless, doubtless, any index has to be adjusted, as demonstrated here, to the particular ecological conditions of each region in order to generate a powerful and representative biomonitoring tool.

### **4.4. Index validation and regional extrapolation**

A third step in our procedure included the statistical index validation process, which produced multiple linear regression models for the BMWP index with good results in general. The obtained *R*<sup>2</sup> value in this study for the BMWP (*R*2 = 0.874) indicates that the model included a great proportion of the variance.

For the index validation, we included a procedure with the addition of nine independent sites, validating the regression model as a satisfactory indicator for river water quality in the Apatlaco River for the BMWP index (*r* = 0.67, *p* = 0.048). For the combined nine sites, there was a positive significant correlation. Three out sites of the nine independent sites were outliers for the calculated multiple linear regressions (**Figure 6a** and **b**).

to register fine and important differences between study sites, which were not evidenced by the WQI (**Figure 5**). The sensitivity of the BMWP index is related with the procedure to assign the values for the water quality classes. For this step, the reference sites are indispensable. The BMWP index was able to detect the pristine condition of "Las Truchas," the clean river reference site, located inside an undisturbed oak forest, showing the highest richness score of this study site. Furthermore, this site showed the presence of the family Perlidae, a bioindicator of excellent water quality. Lakew and Moog [30] stated that a reference site must meet both abiotic and biotic requirements; the same authors consider a reference site as the least impaired site characterized by selected physical, chemical and biological characteristics. Las Truchas site reached the higher *Pcq* and WQI scores, the less impairment of adjacent land use and also meet the biological requirements of the higher richness, as well as the higher scores of the BMWP index. In consequence, the site "Las Truchas" meets the requirements of a reference site. The BMWP scores obtained in this study also showed a wide seasonal fluctuation, particularly for Las Truchas site. The BMWP fluctuation provided for Las Truchas showed scores from 69 to 110, these data conformed the set of values for obtaining the median value that correspond to the "Excellent" quality category, from which range values for all water quality classes were assigned. The method used for the allocation of scores to each water quality class was proposed by Pond and McMurray [21]. These procedure displayed a suitable assignation of the water quality categories and consequently reducing the probability of commit errors (Type I and II)

The bioindication values of the aquatic macroinvertebrate families do not always match completely from one country to another, which can be due to the variations in taxonomic tolerances of each basin and biogeographic region and to the method of assigning bioindi‐ cation values, which in most cases is unknown, generating some uncertainty in the scores assigned to each family. However, there are families of aquatic macroinvertebrates charac‐ teristic of very healthy environments, as is the case of the Perlidae with a score of 10; or in extremely hostile conditions, the midges and lumbriculids with score values of 1 [14]. In the present study, the wide distribution of families in different intervals of *Pcq*, coupled with the abundance classes, allowed us to generate a bioindication value that is neither overestimated nor underestimated, as was evidenced by the degree of adjustment of the

Our results show that the BMWP index has a good discriminating capacity; nevertheless, doubtless, any index has to be adjusted, as demonstrated here, to the particular ecological conditions of each region in order to generate a powerful and representative biomonitoring

A third step in our procedure included the statistical index validation process, which produced multiple linear regression models for the BMWP index with good results in general. The

value in this study for the BMWP (*R*2 = 0.874) indicates that the model included a

[35] for the categorization of the study sites.

**4.4. Index validation and regional extrapolation**

models described above.

great proportion of the variance.

tool.

52 Water Quality

obtained *R*<sup>2</sup>

**Figure 6.** Validation and range extension data correlation. (a) Validation BMWP observed *vs* expected. (b) Validation ASP observed *vs* expected. (c) Extension BMWP observed *vs* expected. (d) Extension ASP observed *vs* expected.

The fourth step in our procedure included the range extrapolation analysis, where the multiple linear regression model was extended to study sites from Amacuzac, Cuautla and Yautepec subbasins for BMWP index, in this case we obtained lower correlation and *p* values (*r* = 0.683, *p* = 0.091). The scores accurately reflect the human impacts on the aquatic macroinvertebrates assemblage, with scores from "Bad, polluted" to "Very Bad, Extremely polluted." There were one outlier out of a total of seven sites for the BMWP, thus indicating that the BMWP regression model seemed to represent a more satisfactory geographical extension in this case (**Figure 6c** and **d**).

Therefore, the calibrated BMWP scores and the proposed water quality classes of this study can be used as a tool for the biomonitoring of water quality in the Apatlaco and Chalma‐ Tembembe rivers and even the subbasins Cuautla, Yautepec and Amacuzac. Furthermore, our ranges for water quality class showed also a good fit for the qualification of study sites and the spectrum of the land use conditions [36]. The studied rivers showed that a great portion of the rivers Apatlaco and Chalma‐Tembembe (nine study sites), with agriculture as the main land use, is qualified as: bad polluted to regular, moderated polluted, while another great portion of the Apatlaco River mainly located in urban zones is qualified as bad, very polluted to very bad. The BMWP calibrated and the water quality assignations resulted to be suitable to assess water categories in the studied rivers. Our results make evident that Apatlaco River needs urgently a management and recovery plan.
