Methods described in this section are used to make decisions about classification attainment. The models are constructed to sequentially amass evidence concerning the highest level of classification criteria that a test community attains, using quantitative predictor variables defined in Section 3(C). The pertinent question, in terms of the classification attainment, is whether or not a test community is attaining at least its statutory classification. The methods described in this rule may also be used to determine if a given waterbody attains a higher class and therefore may be subject to statutory antidegradation provisions or considered for water quality reclassification. The methods may also be used, where appropriate, for other purposes including assessment of pre-impact baseline conditions or site-specific impact evaluations.
A.General provisions for aquatic life standards. Except as otherwise provided in Section 3(G)(3), Professional judgement, of this chapter, all samples of benthic macroinvertebrates that are collected for the purpose of classification attainment evaluation using the linear discriminant model described in the following section, whether collected by the department or by any person submitting data to the department, must be collected, processed and identified in conformance with "Methods for Biological Sampling and Analysis of Maine's Rivers and Streams" (DE P LW 0387 -B2002). Selection of an appropriate sampling site must also conform to criteria set forth in "Methods for Biological Sampling and Analysis of Maine's Rivers and Streams" (DE P LW 0387 -B2002). Quantitative analysis of the sample must conform to the requirements set forth in Sections 3(B) through 3(F) of this chapter and must include a quality assurance plan approved by the department, as specified in "Methods for Biological Sampling and Analysis of Maine's Rivers and Streams" (DE P LW 0387 -B2002). Samples must be identified to the genus level, where practicable. Computation of indices and measures of community structure required for the linear discriminant models must be adjusted to the genus level of taxonomy (see Section 3(C), V ariable number 2, Generic Richness). Minimum Provisions. Samples that have been properly collected and analyzed but fail to meet either of the following criteria are unsuitable for further analysis through the linear discriminant models:
(1) Total mean abundance (Section 3(C) Variable number 1) must be at least 50 individuals (average per basket/cone/bag); and(2) Generic richness (Section 3(C) Variable number 2) for three replicate basket/cone/bag samplers must be at least 15. Samples not attaining these criteria may be evaluated according to Section 3(G) of this chapter, Professional judgment.
B.Aquatic life statistical decision models. The following statistical decision models consist of linear discriminant functions developed to use quantitative ecological attributes of the macroinvertebrate community (see Section 3(C) through 3(E)) to determine the strength of the association of a test community to any of the water quality classes (Appendix 1). The coefficients or weights (see Section 3(F)) are calculated using a linear optimization algorithm to minimize the distance, in multivariate space, between sites within a class, and to maximize the distance between sites between classes. The linear discriminant function has the form:
Z = C + W1X1 + W2X2 + ...WnXn
Where: Z = discriminant score
C = constant
Wi = the coefficients or weights (from Section 3(F))
Xi = the predictor variable values (from Section 3(C))
Association values are computed, using variable values from a test sample, for each classification by employing one four-way model and three two-way models. The four-way model uses nine variables pertinent to the evaluation of all classes and provides four initial probabilities that a given site attains one of three classes (AA/A, B, or C), or is in nonattainment (NA) of the minimum criteria for any class. Class AA and Class A have the same aquatic life standards and, therefore, are treated as the same aquatic life class. These probabilities have a possible range from 0.0 to 1.0, and are used, after transformation, as variables in each of the three subsequent final decision models. The final decision models (the three, two-way models) are designed to distinguish between a given class and any higher classes as one group and any lower classes as the other group (e.g., Classes AA/A+B+C vs. NA; Classes AA/A+B vs. Class C+NA; Class AA/A vs. Classes B+C+NA). The equations for the final decision models use the predictor variables relevant to the class being tested (Section 3(F)). The resultant discriminant scores are known as the Mahalonobis Distance where:
Mahalonobis Distance = Zt (sample x) = g1 (x,t) + g2 (t)
Where: Zt = discriminant score for sample x, class t
g1 (x,t) = (x-mt)' S-1 (x-mt)
g2 (t) = -2 loge (qt) = 0 (if prior probabilities are equal)
Where: x = a vector containing all the values of all the variables for a given linear discriminant function, for a given sample, of class t
mt = a vector, as for x, but containing the means of all predictor variables in the given linear discriminant function, for the given sample, of class t
S = pooled covariance matrix (the variance of the multivariate observation)
qt = value of the prior probability that a given sample is Class A, B, C, or NA.
The probability (association value) of a sample x, belonging to a particular class t, is:
Click here to view Image
Where: Pt(x) = the probability that sample x belongs to class t (for Classes A, B, C, NA)
e = the exponential function
-0.5 = a standardization constant from the normal distribution
Zt = the discriminant score or Mahalonobis Distance for class t (Classes A, B, C, NA)
C.Methods for the calculation of indices and measures used in the linear discriminant models Variables (1) to (30) are as follows.(1) Total mean abundance. Count all individuals in all replicate samplers from a site and divide by the number of replicates to yield the mean number of individuals per sampler.(2) Generic richness. Count the number of different genera found in all replicate samplers from one site. Counting rules for generic richness:
(a) Species-level counts. All population counts at the species level are aggregated to the generic level.(b) Family-level counts, no more than one genus. A family level identification that includes no more than one taxon identified to the generic level is counted as a separate taxon in generic richness counts.(c) Family-level counts, more than one genus. A family level identification with more than one taxon identified to generic level is not counted toward generic richness. Counts are divided proportionately among the genera that are present.(d) Phylum, Class, or Order counts. A higher level taxonomic identification (Phylum, Class, Order) is not counted toward generic richness unless it is the only representative.(e) Pupae. Pupae are ignored in all calculations.(3) Plecoptera mean abundance. Count all individuals from the order Plecoptera in all replicate samplers from one site and divide by the number of replicates to yield mean number of Plecopteran individuals per sampler.(4) Ephemeroptera mean abundance. Count all individuals from the order Ephemeroptera in all replicate samplers from one site and divide by the number of replicates to yield the mean number of Ephemeropteran individuals per sampler.(5) Shannon-Wiener Generic Diversity. Shannon-Wiener generic diversity is computed after adjusting all counts to genus, as described under paragraph (2) above. Click here to view Image
where: Click here to view Image= Shannon-Wiener Diversity
c = 3.321928 (converts base 10 log to base 2)
N = Total abundance of individuals
ni = Total abundance of individuals in the ith taxon
(6) Hilsenhoff Biotic Index. HBI is calculated using all taxa in the sample that have assigned tolerance values. Tolerance values are provided in Hilsenhoff, William. 1987. An Improved Biotic Index of Organic Stream Pollution, The Great Lakes Entomologist 20:31-39. Click here to view Image
Where: HBI = Hilsenhoff Biotic Index
Ni = number of individuals in the ith taxon
aI = tolerance value assigned to that taxon
N = total number of individuals in sample with tolerance values
(7) Relative Chironomidae abundance. Calculate the mean number of individuals of the family Chironomidae, following the counting rules in Variable 4, and divide by total abundance (Variable 1).(8) Relative Diptera richness. Count the number of genera of the Order Diptera, following counting rules in Variable 2, and divide by generic richness (Variable 2).(9)Hydropsyche abundance. Count all the individuals from the genus Hydropsyche in all replicate samplers from a site, and divide by the number of replicates to yield mean number of Hydropsyche individuals per sampler.(10) Probability (A+B+C) from first stage model. The sum of probabilities for Classes A, B, and C from first stage model.(11)Cheumatopsyche abundance. Count all individuals from the genus Cheumatopsyche in all replicate samplers from one site and divide by the number of replicates to yield mean number of Cheumatopsyche individuals per sampler.(12) EPT-Diptera richness ratio. Divide EPT generic richness (Variable 19) by the number of genera from the order Diptera, following counting rules in Variable 2. If the number of genera of Diptera in the sample is 0, a value of 1 is assigned to the denominator.(13) Relative Oligochaeta abundance. Calculate the mean number of individuals of the class Oligochaeta, following counting rules in Variable 4, and divide by total abundance (Variable 1).(14) Probability (A+B) from first stage model. The sum of probabilities for Classes A and B from first stage model.(15) Perlidae mean abundance. Count all individuals from the family Perlidae (Section 3(E)) in all replicate samplers from one site and divide by the number of replicates to yield mean number of Perlidae per sampler.(16) Tanypodinae mean abundance. Count all individuals from the subfamily Tanypodinae (Section 3(E)) in all replicate samplers from one site and divide by the number of replicates to yield mean number of Tanypodinae per sampler.(17) Chironomini mean abundance. Count all individuals from the tribe Chironomini (Section 3(E)) in all replicate samplers from one site and divide by the number of replicates to yield mean number of Chironomini per sampler.(18) Relative Ephemeroptera abundance. Variable 4 divided by Variable 1.(19) EPT generic richness. Count the number of different genera from the order Ephemeroptera (E), Plecoptera (P), and Trichoptera (T) in all replicate samplers, according to counting rules in Variable 2, generic richness.(21) Sum of mean abundance of Dicrotendipes & Micropsectra & Parachironomus & Helobdella. Sum the abundance of the 4 genera and divide by the number of replicates (as performed in Variable 4).(22) Probability of Class A from first stage model.(23) Relative Plecoptera richness. Count number of genera of Order Plecoptera, following counting rules in Variable 2, and divide by generic richness (Variable 2).(25) Sum of mean abundance of Cheumatopsyche & Cricotopus & Tanytarsus & Ablabesmyia. Sum the number of individuals in each genus in all replicate samplers and divide by the number of replicates (as performed in Variable 4).(26) Sum of mean abundance of Acroneuria & Stenonema. Sum the number of individuals in each genus in all replicate samplers and divide by the number of replicates (as in Variable 4).(28) Ratio of EP generic richness. Count the number of different genera from the order Ephemeroptera (E), and Plecoptera (P) in all replicate samplers, following counting rules in Variable 2, and divide by 14 (maximum expected for Class A).(30) Ratio of Class A indicator taxa. Count the number of Class A indicator taxa as listed in Section 3(D) that are present in the community and divide by 7 (total possible number).D. Indicator taxa for Class ABrachycentrus --- (Trichoptera: Brachycentridae)
Serratella -------- (Ephemeroptera: Ephemerellidae)
Leucrocuta ------- (Ephemeroptera: Heptageniidae)
Glossosoma ------ (Trichoptera: Glossosomatidae)
Paragnetina ----- (Plecoptera: Perlidae)
Eurylophella ------ (Ephemeroptera: Ephemerellidae)
Psilotreta -------- (Trichoptera: Odontoceridae)
E. Family functional groupsPLECOPTERA
Perlidae
Acroneuria Agnetina
Attaneuria Beloneuria
Eccoptura Neoperla
Paragnetina Perlesta
Perlinella
CHIRONOMIDAE
Tanypodinae
Ablabesmyia Clinotanypus
Coelotanypus Conchapelopia
Djalmabatista Guttipelopia
Hudsonimyia Labrundinia
Larsia Meropelopia
Natarsia Nilotanypus
Paramerina Pentaneura
Procladius Psectrotanypus
Rheopelopia Tanypus
Telopelopia Thienemannimyia
Trissopelopia Zavrelimyia
Chironomini
Pseudochironomus Axarus
Chironomus Cladopelma
Cryptochironomus Cryptotendipes
Demicryptochironomus Dicrotendipes
Einfeldia Endochironomus
Glyptotendipes Goeldichironomus
Harnischia Kiefferulus
Lauterborniella Microchironomus
Microtendipes Nilothauma
Pagastiella Parachironomus
Paracladopelma Paralauterborniella
Paratendipes Phaenopsectra
Polypedilum Robackia
Stelechomyia Stenochironomus
Stictochironomus Tribelos
Xenochironomus
F.Model coefficientsFirst Stage Model
Coefficients
| Variable number | Transformation | Class A | Class B | Class C | Nonattainment |
| Constant | -99.95508 | -105.70948 | -112.67581 | -107.74283 |
| 1 | nLog (value +0.001) | 10.77061 | 11.46981 | 11.80888 | 11.26793 |
| 2 | -0.38619 | -0.43340 | -0.50051 | -0.48822 |
| 3 | nLog (value +0.001) | 0.23940 | 0.03946 | -0.60923 | -0.95480 |
| 4 | nLog (value +0.001) | -0.59970 | -0.55500 | -0.67722 | -1.79032 |
| 5 | 21.22732 | 20.91256 | 21.07602 | 19.46547 |
| 6 | 8.01620 | 9.12163 | 10.31492 | 10.72746 |
| 7 | nLog (value +0.001) | -11.70298 | -11.52650 | -11.49414 | -11.66371 |
| 8 | 70.77937 | 71.09637 | 72.46514 | 70.22517 |
| 9 | -0.00535 | -0.00398 | -0.00152 | 0.00007 |
Final Classification Models
Class C or better model
Coefficients
| Variable number | Transformation | Class A-B-C | Nonattainment |
| Constant | -25.70020 | -8.55844 |
| 10 | Arcsin | 19.98470 | 3.36032 |
| 11 | nLog (value +0.001) | -0.26001 | -0.43781 |
| 12 | Sq. root | 5.57672 | 5.92732 |
| 13 | nLog (value +0.001) | -2.33229 | -1.89945 |
Class B or better model
Coefficients
| Variable number | Transformation | Class A-B | Class C-nonattainment |
| Constant | -17.81016 | -6.93836 |
| 14 | Arcsin | 12.04826 | 3.63707 |
| 15 | nLog (value +0.001) | -1.11091 | -1.03934 |
| 16 | nLog (value +0.001) | -0.10582 | 0.01978 |
| 17 | nLog (value +0.001) | 0.17813 | 0.10825 |
| 18 | 4.03202 | -1.14508 |
| 19 | 0.87400 | 0.63310 |
| 21 | nLog (value +0.001) | -0.69360 | -0.53194 |
Class A model
Coefficients
| Variable number | Transformation | Class A | Class B-C-nonattainment |
| Constant | -9.59254 | -4.08552 |
| 22 | Arcsin | 8.34341 | 1.52080 |
| 23 | 3.78999 | 4.27447 |
| 25 | nLog (value +0.001) | 0.53110 | 0.77851 |
| 26 | nLog (value +0.001) | -0.55838 | -0.51448 |
| 28 | 12.32529 | 9.81592 |
| 30 | 6.94828 | -0.67475 |
G.Professional judgment. Where there is documented evidence of conditions that could result in uncharacteristic findings, allowances may be made to account for those situations by adjusting the classification attainment decision through use of professional judgement, as provided in this section, paragraphs 3(G)(1) to 3(G)(3). The department may make adjustments to the classification attainment decision based on analytical, biological, and habitat information or may require that additional monitoring of affected waters be conducted prior to issuing a classification attainment decision. (1) Sampling procedures and minimum provisions conform but other confounding factors exist. When samples of test communities conform to criteria provided in "Methods for Biological Sampling and Analysis of Maine's Rivers and Streams" (DE P LW 0387 -B2002) and Section 3(A) of this chapter, they are suitable to be analyzed by the linear discriminant models for classification attainment evaluation. These models are not suitable for use in areas of impoundments that thermally stratify or in areas where there is a net annual deposition of fine sediment. Professional judgement may be utilized when conditions are found that are atypical to the derivation of the linear discriminant model, as provided in Section 3(B-F). Factors that may allow adjustments to the model outcome include but are not limited to: habitat factors, including lake outlets from waters classified GPA, unusual naturally-caused substrate character, tidal effects, cataclysmic events, oligotrophic conditions; sampling factors, including disturbed samples, unusual taxa assemblages, and documented human error in sampling; and sample processing factors, including subsample vs. whole sample analysis and documented human error in processing. The following adjustments may be made to correct for these conditions: (a) Raise the finding. On the basis of documented evidence of specific conditions such as those defined above, the department may decide: (i) To raise the classification attainment outcome predicted by the model from nonattainment of any class to indeterminate or to attainment of Class C; or(ii) To raise the classification attainment outcome predicted by the model from attainment in one class to attainment in the next higher class; or(iii) To determine that a sample with an indeterminate outcome for a given class attains that class.(b) Lower the finding. On the basis of documented, substantive evidence that the narrative aquatic life criteria for the assigned class are not met, the department may decide to lower the classification attainment finding.(c) Indeterminate. Where the department cannot make a finding as described in 3(G)(1)(a-b), additional monitoring of the test community may be required before a determination of class attainment can be made.(2) Minimum provisions do not conform. For classification evaluation of test communities that do not conform to criteria provided in Section 3(A) of this chapter, minimum provisions, professional judgement may be used as follows: (a) Determination of nonattainment. Those samples having any of the ecological attributes not attaining the minimum provisions (Section 3(A)), and where there is no evidence of confounding factors that could result in uncharacteristic findings as defined above (Section 3(G)(1)), must be determined to be in nonattainment of the minimum provisions of the aquatic life criteria for any class.(b) Determination of attainment when minimum provisions are not met. Where there is evidence of factors that could result in minimum provisions not being met, professional judgment may be used to make a professional finding of attainment of the aquatic life criteria for any class. Such decisions will be provisional until appropriate resampling is carried out.(3) Standard sampling procedures are not feasible or appropriate. For classification attainment evaluation of test communities that do not conform to criteria provided in "Methods for Biological Sampling and Analysis of Maine's Rivers and Streams" (DE P LW 0387 -B2002), the department may make an assessment of classification attainment or aquatic life impact in accordance with the following procedures: (a) Approved assessment plan. A quantitative sampling and data analysis plan must be developed in accordance with methods established in the scientific literature on water pollution biology, and the department must approve the plan.(b) Determination of sampling methods. Sampling methods are determined on a site-specific basis, based on habitat conditions of the sampling site, and the season sampled; (i) The preferred method for sampling hard-bottomed substrates is the rock basket/cone/bag method as described in "Methods for Biological Sampling and Analysis of Maine's Rivers and Streams" (DE P LW 0387 -B2002).(ii) Soft-bottomed substrates will, whenever ecologically appropriate and practical, be sampled by core or dredge of known dimension.(c) Other methods. Other methods may be used where ecologically appropriate and practical.(d) Classification attainment decisions. Classification attainment decisions may be based on a determination of the degree to which the sampled site conforms to the narrative aquatic life classification criteria provided in statutory standards for water quality classification. The decision is based on established principles of water pollution biology and must be fully documented.(e) Site specific impact decisions. Site specific impact decisions may rely on established methods of analysis of comparative data between a test community and an approved reference community.(f) Determination of detrimental impact. A determination of detrimental impact to aquatic life of a test community without an approved reference community may be made if it can be documented, based on established methods of the interpretation of macroinvertebrate data, and based on established principles of water pollution biology, that the community fails to demonstrate the ecological attributes of its designated class as defined by the narrative aquatic life standards in the water quality classification law.06-096 C.M.R. ch. 579, § 3