{"slug":"noaa-isd-data-aermod-2025","title":"NOAA ISD Stopped in 2025: Meteorological Data for AERMOD Now","excerpt":"NOAA’s legacy Integrated Surface Database files stopped updating in late August 2025, so airport data for later study periods is no longer available in the format AERMET reads. What changed, who is affected, and the six practical alternatives for AERMOD studies in India.","date":"2026-11-03","author":"DS Technoverse","image":"/blog-images/aermod-software/aermod-report-met-stations-table.webp","category":"Air Dispersion Modelling","tags":["NOAA ISD","AERMOD meteorological data","AERMET surface data","NASA POWER AERMOD","ERA5 AERMOD","GHCNh","site meteorology EIA"],"content":"For years, the first step of almost every AERMOD study outside the United States was the same: find the nearest airport in NOAA's Integrated Surface Database (ISD), download its hourly records for the study period, and feed them to AERMET. In India, where few projects have long-term site weather data, the ISD record of the nearest airport was the default surface data set for screening studies, CTE applications and the meteorological cross-check in EIA reports.\n\nThat default has gone. NOAA's legacy ISD station files stopped updating in late August 2025. If your study period is after that — a post-monsoon 2025 baseline, winter 2025–26, summer 2026 — there is no ISD airport record to download, and AERMET cannot be fed the way it used to be.\n\nThis article sets out what we found, who is affected, and the six practical alternatives, with the strengths and the review risks of each. It finishes with how our [AERMOD Studio](/services/aermod-software/) software handles the change automatically and documents it in the report.\n\n![Surface stations table from an AERMOD Studio report: every airport within 150 km rejected because its record ends before the study period](/blog-images/aermod-software/aermod-report-met-stations-table.webp)\n\n## What Changed\n\nNOAA's National Centers for Environmental Information (NCEI) publish the ISD as one compressed file per station per year, with a station history file that lists, for every station, the dates its record begins and ends. When we checked in November 2026:\n\n- The station history file had last been modified on **30 August 2025**.\n- For the great majority of stations — about nine thousand of them — the recorded end date was **24 August 2025**, with most of the rest ending a day later or years earlier.\n- There was **no 2026 folder** of station files at all.\n\nIn other words, the legacy ISD stopped at the end of August 2025 for every station, not just for a few closed airports. Data up to that date is still available and still perfectly usable; data after it is not, in that product.\n\nNCEI's newer hourly product is the Global Historical Climatology Network – hourly (GHCNh). It is distributed in a different format from ISD, which AERMET does not read directly.\n\n### What it looks like for a real site\n\nHere are the airport stations within 150 km of a dummy test site near Pithampur, Madhya Pradesh, as listed in the ISD station history file, with the reason each could not be used for an October–December 2025 study:\n\n| Station | Distance from site | ISD record ends | Usable for Oct–Dec 2025? |\n|---|---|---|---|\n| Devi Ahilyabai Holkar (Indore airport) | 17.6 km | 24 August 2025 | No |\n| Dhar | 39.0 km | 27 June 2025 | No |\n| Ujjain | 64.6 km | 24 August 2025 | No |\n| Khargone | 88.4 km | 24 August 2025 | No |\n| Ratlam | 101.7 km | 24 August 2025 | No |\n| Khandwa | 111.6 km | 24 August 2025 | No |\n| Shajapur | 112.4 km | 25 May 2022 | No |\n| Alirajpur | 136.5 km | 17 September 2021 | No |\n| Dohad | 147.0 km | 27 June 2024 | No |\n\nEvery station is ruled out, including a major airport 18 km away. The same pattern holds across India and, indeed, worldwide.\n\n### What has not changed\n\nUpper-air data is still flowing. The Integrated Global Radiosonde Archive (IGRA), which provides the twice-daily soundings AERMET needs for the morning temperature profile, continues to be updated: in November 2026, the IGRA file for Bhopal contained soundings up to the first days of that month. Upper-air data is not the problem; surface data is.\n\n## Who Is Affected\n\n- **EIA baselines from autumn 2025 onwards.** A post-monsoon (October–December 2025), winter (December 2025–February 2026) or summer (March–May 2026) baseline season cannot be paired with ISD airport data for the same period.\n- **CTE, CTO and expansion studies** that relied on the nearest airport for a recent year.\n- **Annual studies** covering any part of the period after August 2025.\n- **Screening studies and stack-height checks** that used \"last year at the nearest airport\" as a convenient default.\n- **Consultants and software** whose workflows download ISD automatically. A tool that looks only for stations \"covering the period\" may now find nothing — or, worse, may quietly fall back to an older year or a distant station without saying so.\n\nStudies whose period falls entirely before late August 2025 are not affected: the historical ISD files are still there.\n\n## Why AERMET Needs Surface Data at All\n\nAERMOD does not use raw weather observations. AERMET turns them into the boundary-layer parameters AERMOD needs for every hour: the friction velocity, the Monin–Obukhov length (a measure of stability), the convective and mechanical mixing heights, the surface heat flux, and the wind and temperature at the reference height. To do that, it needs, for every hour:\n\n- **Wind speed and direction** — the transport of the plume and the mechanical turbulence.\n- **Temperature** — for buoyancy and plume rise.\n- **Cloud cover or solar radiation** — to estimate how much the ground heats in the day and cools at night, which sets stability. Without either, AERMET cannot tell a sunny afternoon from an overcast one.\n- **Pressure and humidity**, used in several calculations, with defaults when missing.\n\nIt also needs the morning upper-air sounding to calculate how the convective boundary layer grows through the day, and the surface characteristics of the area (albedo, Bowen ratio, roughness length).\n\nAirport ISD records supplied all of the surface variables in one hourly record, including the cloud cover observed by the airport's meteorological staff or instruments. That convenience is what has been lost.\n\n## The Six Alternatives\n\n### 1. Site-specific meteorological data\n\nFor EIA studies this was always the expected primary source, and it is now the strongest one. The baseline monitoring programme normally includes an automatic weather station at the site, recording hourly wind speed and direction, temperature, relative humidity and rainfall, and often solar radiation and pressure.\n\nAERMET reads such data through its ONSITE pathway. What to check:\n\n- **Measurement height and exposure.** Wind should be measured at about 10 m in an open location, away from buildings and trees; the height is entered in AERMET.\n- **Averaging.** Hourly averages, with wind direction averaged correctly (as a vector or by the unit-vector method).\n- **Time convention.** AERMET numbers hours 1 to 24 by the end of the hour, in local standard time; a record stamped 00:00 belongs to hour 24 of the previous day.\n- **Completeness.** At least 90% valid hours over the study period, as US EPA guidance on meteorological monitoring for regulatory modelling expects.\n- **Stability inputs.** If the station records solar radiation, AERMET can use it for daytime conditions. If it records neither solar radiation nor cloud cover, the cloud cover has to come from another source — and after August 2025 that can no longer be the nearest airport's ISD record.\n\nSite data has one weakness: the season is short. A three-month record is fine for a three-month study, but an annual study needs a full year.\n\n### 2. NOAA GHCNh\n\nNCEI's newer hourly data set covers many of the same stations as ISD, including Indian airports. Its advantage is that it is observed data from familiar stations, so representativeness arguments built around an airport still apply.\n\nIts disadvantage, for now, is practical: AERMET does not read it. Each station-year has to be converted into a format AERMET accepts — for example its ONSITE format — with careful handling of units, time zones, missing-value codes and cloud cover. Conversion is a one-off piece of software, but it must be validated, and the conversion method documented in the report. Check, too, that the hourly record for your station is complete for your period: many Indian stations report less frequently at night.\n\n### 3. NASA POWER\n\nNASA's POWER project (Prediction Of Worldwide Energy Resources) offers free hourly data for any latitude and longitude through a public web service, with no key. Its meteorology comes from the MERRA-2 reanalysis on a grid of about 0.5° by 0.625°, and its solar radiation from CERES satellite products. For AERMOD, the useful hourly parameters are wind speed and direction at 10 m, temperature and humidity at 2 m, surface pressure, all-sky surface solar radiation and cloud amount.\n\nStrengths:\n\n- **Complete records.** No missing hours, so data capture is not an issue.\n- **Any location, any recent period**, including after August 2025.\n- **All the variables AERMET needs**, including cloud cover and solar radiation, in one place.\n\nLimitations, which a report must state:\n\n- **It is modelled, not measured.** Reanalysis blends observations with a weather model; it is a best estimate of conditions, not an observation at your site.\n- **It is smooth.** A grid cell tens of kilometres wide averages out local effects. Light winds and calms tend to be under-represented, and valleys, coasts and urban heat islands are not resolved.\n- **Heights differ.** Wind is at 10 m but temperature at 2 m; when the data is fed to AERMET as a single-level site record, that is an approximation.\n\nNASA POWER is a reasonable choice for screening, stack-height comparisons and studies where no representative observations exist, and it is far better than using a distant or incomplete station without saying so. For an Environmental Clearance study, it should support, not replace, the site data from the baseline season.\n\n### 4. ERA5 reanalysis\n\nThe European Centre for Medium-Range Weather Forecasts' ERA5 reanalysis is available free, after registration, from the Copernicus Climate Data Store. It is hourly, global, on a grid of about 0.25° (roughly 30 km), and extends back to 1940.\n\nERA5 is finer than MERRA-2 and is widely used to drive mesoscale models. Used directly as a surface record for AERMET, it shares the limitations of any reanalysis — smoothing, under-represented calms, single-cell representativeness — and needs the same conversion work as GHCNh. Its most valuable role is as the starting point for option 6.\n\n### 5. India Meteorological Department data\n\nThe India Meteorological Department (IMD) maintains the national observing network and supplies station data on request, for a charge. IMD records are observed data from official stations, which strengthens representativeness arguments.\n\nCheck the reporting frequency before you buy: many stations report synoptic observations every three hours rather than every hour, and AERMET needs hourly data. Three-hourly records require interpolation, which must be justified, and the result is weaker than a true hourly record. As with GHCNh, the data has to be converted into a format AERMET accepts.\n\n### 6. Prognostic meteorology (WRF with MMIF)\n\nWhere no representative observations exist, a mesoscale weather model such as WRF can simulate the study period at fine resolution — a few kilometres or less — driven by a global reanalysis such as ERA5. The US EPA's Mesoscale Model Interface Program (MMIF) converts WRF output into files AERMOD can use. The 2017 revision of the US EPA Guideline on Air Quality Models allows prognostic meteorological data where there is no representative observational data, subject to an evaluation of the model's performance.\n\nThis is the most capable option for complex terrain and coastal sites, and the most demanding: it needs modelling expertise, computing time and a performance evaluation against whatever observations are available. It is usually reserved for large or contested studies.\n\n## How Much Does the Choice of Data Change the Answer?\n\nIt is fair to ask whether any of this matters for the result. We ran the same dummy study twice with AERMOD Studio to find out: a single 30 m stack near Kolar, Karnataka, with a 300 m plant boundary, for March to May 2025 — a period still covered by ISD, so both options were available. The first run used the nearest qualifying airport (Bengaluru HAL airport, 62 km away, 96% valid hours) with Bengaluru upper air; the second used NASA POWER hourly data for the site, with everything else unchanged.\n\n| Result (incremental, outside the boundary) | NOAA airport data | NASA POWER | Difference |\n|---|---|---|---|\n| Data capture (AERMOD) | 92.7% | 99.9% | Complete record with POWER |\n| PM10 24-hour maximum | 4.86 µg/m³, 0.45 km ENE | 5.98 µg/m³, 0.40 km E | +23% |\n| SO₂ 24-hour maximum | 12.97 µg/m³, 0.45 km ENE | 15.96 µg/m³, 0.40 km E | +23% |\n| SO₂ 1-hour maximum | 32.86 µg/m³, 0.30 km N | 40.48 µg/m³, 0.81 km WSW | +23%, different location |\n| SO₂ period average | 2.72 µg/m³ | 2.11 µg/m³ | −22% |\n| NOx 24-hour maximum | 8.11 µg/m³, 0.45 km ENE | 9.97 µg/m³, 0.40 km E | +23% |\n\nThree lessons from one site and one season — not a general rule, but typical of what we see:\n\n- **The short-term maxima moved by about a quarter**, and the 1-hour maximum moved to a different side of the plant. For a project close to a standard, that is the difference between compliance and exceedance.\n- **The period average moved the other way.** Reanalysis winds are smoother and steadier, which changes how often the plume is carried over the same receptors. Neither data set is \"conservative\" for every averaging period.\n- **Both runs were valid.** The airport run passed the 90% completeness test; the POWER run had no gaps. Validity is necessary, not sufficient: representativeness is a separate question, and the report must address it.\n\nThe practical conclusion is to treat the meteorological source as a key assumption, state it prominently, and — where a result is close to a standard — test it with a second data set.\n\n## A Checklist for Your Next Baseline Season\n\nIf you are planning an EIA baseline from 2026 onwards, the meteorological record you collect is now likely to be the backbone of the modelling, not just a cross-check. Before the season starts:\n\n- **Site the station well.** An open location, wind at about 10 m, away from buildings, trees and the plant's own structures.\n- **Record the right variables.** Wind speed and direction, temperature, relative humidity and pressure at minimum; add solar radiation, which AERMET can use for daytime stability, and record cloud cover if an observer is on site.\n- **Log hourly averages in local standard time**, and document the time convention (hour beginning or hour ending).\n- **Check completeness every week.** A failed logger discovered at the end of a three-month season cannot be repaired.\n- **Keep the raw data and the calibration records.** Reviewers increasingly ask for them.\n- **Plan the upper-air station.** Identify the nearest IGRA station and check that it has soundings in your season; some Indian stations have long gaps.\n- **Decide the fallback in advance.** Agree with the client which source fills gaps — and say so in the methodology — rather than choosing after the fact.\n\n## Comparing the Options\n\n| Option | Observed or modelled | Spatial detail | Cost | Effort | Data capture | Review risk |\n|---|---|---|---|---|---|---|\n| Site station | Observed | The site itself | Part of the baseline | Low to medium | Depends on the station | Lowest, if complete and well sited |\n| NOAA GHCNh | Observed | The station | Free | Medium (conversion) | Depends on the station | Low, if representative and complete |\n| NASA POWER | Modelled | ~50 km grid | Free | Low | Complete | Medium: must be disclosed and justified |\n| ERA5 | Modelled | ~30 km grid | Free | Medium (conversion) | Complete | Medium |\n| IMD stations | Observed | The station | Charged | Medium (conversion, often 3-hourly) | Depends on the station | Low to medium |\n| WRF + MMIF | Modelled | 1–4 km | Expert time and computing | High | Complete | Low to medium, with a performance evaluation |\n\n## Completeness: the 90% Rule and How to Check It\n\nWhatever the source, completeness decides whether the result is defensible. US EPA guidance on meteorological monitoring for regulatory modelling expects at least 90% valid data. Two places to check:\n\n- **Before AERMET**: count the valid hours of each variable in your period. A station that reports every three hours at night can look complete in a daily summary and still have a third of its hours missing.\n- **After AERMOD**: AERMOD reports the number and percentage of missing hours in its output. This is the figure that matters, because it counts the hours AERMOD could not use for any reason, including missing upper-air data or failed AERMET calculations.\n\nLow completeness does more than add uncertainty. AERMOD applies rules for averaging periods with missing hours, and a sparse record can produce results that are internally inconsistent — for example, a period average higher than the highest 24-hour average, which is impossible with complete data. Treat any such result as a sign that the meteorology is not fit for purpose.\n\n## What to Write in the Report\n\nReviewers will ask about the meteorology first. A report that answers these questions in advance avoids most of the queries:\n\n1. **Which data was used, and from where?** Name the source, the station identifiers and distances, the period and the variables.\n2. **Why this source?** List the stations considered and why each was used or rejected — including, now, that the legacy ISD record ends before the study period.\n3. **How complete is it?** Give the percentage of valid hours, before and after AERMET.\n4. **Is it representative?** Compare the wind rose with site or regional data; note terrain, coast and land-use differences.\n5. **If modelled data was used, say so plainly** — that it is reanalysis, its resolution and its known limitations — and how it relates to the site data from the baseline season.\n6. **How was it processed?** The AERMET version and options, the surface characteristics and how they were chosen, the time zone, and any substitution of missing data.\n7. **Upper air.** The IGRA station, its distance, the number of soundings and how missing soundings were handled.\n\nThe upper-air part of an AERMOD Studio report, for example, lists every station considered, the number of soundings in the period, the share of days with a morning sounding, and the decision for each.\n\n## Studies Already in Progress or Submitted\n\nThe change affects studies differently depending on where they are:\n\n- **Submitted studies with periods before late August 2025** are unaffected: the ISD data they used remains valid and available for anyone who wants to check it.\n- **Studies run recently with automated tools** deserve a check. If the study period is after August 2025 and the report names an airport station, find out exactly which period of that station's data was used. Some workflows silently fall back to an earlier year or a more distant station when the requested period is not available. That may be defensible, but only if the report says so and explains why.\n- **Studies being prepared now** should settle the meteorological source before the modelling starts, with the client and, where appropriate, in line with the Terms of Reference, rather than discovering the gap when the run fails.\n- **Studies under review** that receive a query about the meteorology can usually be answered by documenting the stations considered and the reason for the source used. Where the source is reanalysis, a comparison with the site station data from the baseline season strengthens the answer considerably.\n\n## How AERMOD Studio Handles It\n\nWhen we hit this change in our own runs, we rebuilt the meteorology step of [AERMOD Studio](/services/aermod-software/) around three principles: choose by documented rules, never hide a fallback, and put the reasons in the report.\n\n- **Surface data.** Every ISD station within 150 km is listed. For stations whose record covers the period, the valid hours are counted and the nearest station with at least 90% is used. Stations whose record ends before the study period — which after August 2025 means all of them — are listed with that reason.\n- **Fallback.** When no airport qualifies, the software uses NASA POWER hourly data for the exact site coordinates, says so in the methodology, the conclusions and the limitations, and gives the reason each airport was rejected.\n- **Site data.** You can upload the hourly record from your baseline weather station instead; it is processed through AERMET's ONSITE pathway, with gaps filled from the nearest airport where that airport's record covers the period.\n- **Upper air.** The nearest IGRA station with a morning sounding on at least 75% of the days is used, with the next one filling missing soundings, and every candidate is listed with its decision.\n- **Quality gate.** If AERMOD reports fewer than 90% valid hours, the run produces no report and no NAAQS verdict, unless you choose a watermarked draft; inconsistent averages are flagged.\n\n![Meteorological stations map from an AERMOD Studio report (dummy test study)](/blog-images/aermod-software/aermod-met-stations-map.webp)\n\nThe demo is free: send us your site and season, and we will show you which data the software selects for it and why.\n\n## Practical Recommendations\n\n- **For EIA studies, plan the site weather station as the primary source** and make sure it records solar radiation or cloud cover, not just wind and temperature. Check its completeness weekly during the season, not at the end.\n- **For screening and design studies**, NASA POWER is an acceptable, disclosed choice where no representative observations exist — and it removes the completeness problem.\n- **For periods before late August 2025**, ISD airport data is still available and still a good choice where the airport is representative.\n- **For annual or contested studies**, consider prognostic meteorology or a converted observed record, and budget for the extra work.\n- **Whatever you choose, document it.** The reason a station was not used is now as important as the station you did use.\n\nOur broader guide to [meteorological data for air dispersion modelling in India](/blog/meteorological-data-for-air-modelling-india/) covers representativeness and site station design in more depth, and [AERMET explained](/blog/aermet-meteorological-data-preprocessing/) covers the processing itself.\n\n## Frequently Asked Questions\n\n### Is NOAA ISD data discontinued?\n\nThe legacy ISD station files that most AERMOD workflows download stopped updating in late August 2025: the station history file was last modified on 30 August 2025, most station records end on 24 August 2025, and there is no 2026 folder. The historical data up to that date remains available. NCEI's newer hourly product is GHCNh.\n\n### Can I still use ISD data for my AERMOD study?\n\nYes, if your study period ends before late August 2025 and the station is representative and complete for that period. For later periods, there is no ISD record to use.\n\n### Does AERMET read GHCNh data?\n\nNot directly. GHCNh is distributed in a different format from ISD, so the data has to be converted into a format AERMET accepts, such as its ONSITE format, and the conversion has to be validated and documented.\n\n### Is NASA POWER data acceptable for AERMOD?\n\nIt is used for screening and design studies and where no representative observations exist, provided the report states clearly that it is modelled reanalysis data with known limitations. For Environmental Clearance studies, site-specific data from the baseline season is normally expected; NASA POWER can support it but should not silently replace it.\n\n### Will the legacy ISD files start updating again?\n\nWe cannot say, and it would be unwise to plan on it. Check the dates in NCEI's station history file before you rely on ISD for any period after August 2025, and design new baseline programmes so that the site station can carry the modelling on its own.\n\n### Is IGRA upper-air data still available?\n\nYes. IGRA continues to be updated; Indian stations such as Bhopal had soundings into November 2026. Check the number of soundings in your period, because some stations have long gaps.\n\n### What data capture does AERMOD need?\n\nUS EPA guidance on meteorological monitoring for regulatory modelling expects at least 90% valid data. Check it both before AERMET and in AERMOD's count of missing hours.\n\n### My site station does not record cloud cover. What can I do?\n\nIf it records solar radiation, AERMET can use that for daytime stability. If it records neither, cloud cover has to come from another source for the same period — no longer the airport's ISD record after August 2025, but, for example, NASA POWER cloud amount or a converted observed record — and the combination must be documented.\n\n### How does AERMOD Studio choose the meteorology?\n\nIt checks every NOAA ISD station within 150 km, uses the nearest one with at least 90% valid hours, and otherwise uses NASA POWER hourly data for the site; you can also upload your own site data. Upper air comes from the nearest IGRA station with enough morning soundings. Every station considered and the reason for its decision is printed in the report.\n\n---\n\n**See AERMOD Studio run your own site — the demo is free.** Send us the site coordinates, the stacks and the season you want to model; we run the study and take you through the meteorology, the isopleths and the report in a short online session. See [AERMOD Software (AERMOD Studio)](/services/aermod-software/), or if your study needs terrain, building downwash or area sources, our [AERMOD modelling service](/services/aermod-air-dispersion-modelling/). DS Technoverse is based in **Indore, Madhya Pradesh** and works with consultants and industries across India.","related":[{"slug":"common-aermod-mistakes","title":"10 Common AERMOD Modelling Mistakes That Fail Regulatory Review"},{"slug":"aermet-meteorological-data-preprocessing","title":"AERMET Explained: How Meteorological Data Is Prepared for AERMOD"},{"slug":"aermod-building-downwash-bpip","title":"AERMOD Building Downwash (BPIP-PRIME): A Practical Overview"},{"slug":"aermod-modelling-consultant-india","title":"AERMOD Modelling Consultant in India: What They Do and How to Hire One"}]}