Twelve weeks of monitoring is not automatically a baseline

The most common baseline comment we see is about representativeness. A campaign runs through one season, the report presents the averages, and the lender's adviser asks whether those weeks describe the site or only those weeks.

Regulators already define what a representative record looks like for some inputs. India's National Ambient Air Quality Standards define annual values as the arithmetic mean of at least 104 measurements a year at a site, about two a week.¹ For meteorological data used in dispersion modelling, US EPA guidance requires 90 percent data recovery, assessed quarter by quarter, over four consecutive quarters.² A twelve-week campaign meets neither on its own. It does not have to, if the report shows how it relates to a longer record.

What a defensible short campaign shows

State the period and why it was chosen. Put the wind rose and temperature range during the campaign beside the long-term record from the nearest reference station. Report data capture for each parameter at each station, not one headline figure: a 90 percent average can hide an analyser that ran for half the campaign.

Compare results with any regional background data and explain the differences. If the campaign misses a season that matters, say so and describe what the modelling assumes for it.

Where projects actually fall short

The World Bank's Independent Evaluation Group rated environmental assessments and management plans satisfactory in more than 90 percent of the projects it reviewed. In the same evaluation it could substantiate mitigation of negative impacts in only about two-thirds of its sample, and found inadequate environmental and social supervision in more than a quarter of IFC projects.³ The evaluation covers 1999 to 2008 and predates the current frameworks. In our experience the pattern still holds: the weak point is usually whether the numbers behind a chapter can be traced and repeated.