Discharge monitoring report data, worked up from your results for the permittee's signatory.

Today a project manager or reporting specialist pulls every result for the outfall and period, adds the client's flows, works out each statistic the permit asks for, applies the state's rule for non-detects, rounds to the permit's decimals and sends the worksheet to the client, who enters it in NetDMR or the state's own system.

In the proof of concept, AI reads copies of the same results, flows and permit pages and drafts the value for every parameter and statistic, with each input result and each step of the calculation beside it. The excursion count and every compliance call are left to people. Nothing is entered, signed or submitted.

The permit sets the statistics. Under 40 CFR 122.45, a continuous discharger other than a POTW gets maximum daily and average monthly limits, and a POTW gets average weekly and average monthly. Section 122.41 makes averages arithmetic unless the permit says otherwise, so a geometric mean for bacteria has to be written into the permit, and it puts any extra monitoring done with a Part 136 method into the calculation, so a result from an added sample can't be dropped. Pennsylvania DEP's DMR instructions turn a concentration into a load by multiplying milligrams per liter by flow in million gallons a day and by 8.34, run weekly averages Sunday to Saturday, and round each value to as many decimals as the permit limit carries. Florida DEP counts a partial week at the end of a month with the following month.

Every state treats a non-detect its own way. Connecticut DEEP asks for a zero in the value field with the detection level in the comments, uses zero for averaging, and asks permittees not to use NODI code B. Pennsylvania uses the reporting limit, drops the less-than sign for the arithmetic and puts it back on the answer if any value in the set was a non-detect, so less than 1.0, 2.0, less than 1.0 and 1.0 average to less than 1.3. Florida DEP uses half the method detection limit or half the effluent limit, whichever is lower, and doesn't accept a lab's I qualifier on Part B of its form. The draft applies the rule your reviewer names for each client's state and permit, and says on every line which rule it used.

Bacteria and missing data break the usual math. Pennsylvania's instructions substitute 1 for a zero in a geometric mean, report a too-numerous-to-count plate as a value with a greater-than sign rather than as TNTC, let the greater-than sign win when a set has both signs, and ask for the largest volume filtered whenever a result is a non-detect. A month with no discharge, a conditional parameter that wasn't triggered and a sample never taken each has its own No Data Indicator on EPA's list, codes C, 9 and E among them, and Connecticut reminds permittees that a NODI code can never excuse a limit exceedance. The draft suggests a code only where the client's own records show why a value is missing, and marks it for the signatory.

Codes, comments and the excursion column. One quantity can wear several parameter codes: Connecticut lists 74076 for flow on the day of sampling, 00056 for average daily flow, 50047 for the maximum during 24 hours and 00058 for an instantaneous flow rate, and wants the sample date of monthly and quarterly monitoring in the comments. The No. Ex. column counts results outside the limits; in Pennsylvania's example, a pH permit of 6.0 to 9.0 with readings of 5.9 and 9.1 gets a 2. Filling it means judging each value against its limit, so the draft leaves it blank for the permittee's people.

Signing stays with the permittee. Since December 21, 2016, 40 CFR 127 has had DMRs submitted electronically. For NetDMR, the Texas Commission on Environmental Quality says consultants, labs and contractors register as data providers who can enter and view DMR data but cannot sign or submit, and a new signatory signs a subscriber agreement for each permit first. Under 40 CFR 122.22 the person who signs certifies under penalty of law that the information is true, accurate and complete. Your reviewer will check each input result, each formula, every rounding, every non-detect and every suggested NODI against the worksheet your lab sent.

If your reviewer says yes, new periods for your permit clients get the same drafting for about 30 days.

Each worksheet is checked by your reviewer before it goes to the client, and the permittee's signatory still reviews, signs and submits the DMR.

Scoring carries on the same way all month, and nothing else is added until the run is judged.