# ENSOFORECAST archive deposit ## Summary This deposit preserves the 49 regular files in ENSOFORECAST.tgz supplied by the IRI Data Library to Vultax on 21 October 2026. Files are copied without conversion and retain their relative paths inside the tarball. The directory contains 29 netCDF files, 14 Fortran unformatted sequential files, two ASCII text files and four unidentified binary files. Nine files are mapped by the preserved IRI catalog to LDEO1 and LDEO2 forecast leads and observed anomalies. ## Identifiers No DOI yet. Permanent record: . Deposit base URL: . Source catalog identifier: . These URLs identify the record and the deposit as published on 26 October 2026. ## Creators Lamont-Doherty Earth Observatory climate group. In `iri-documentation/entries/ENSOFORECAST/index.tex`, IRI describes the resource as "Equatorial Pacific SSTA forecasts from the Lamont-Doherty Earth Observatory Climate Group". The model is by Stephen E. Zebiak and Mark A. Cane. See their 1987 paper, A Model El Nino-Southern Oscillation, Monthly Weather Review, DOI . The paper identifies the model; it does not establish a separate model version for every file in this deposit. ## Keepers and transfer The prior keeper is the IRI Data Library. Aaron Kaplan of the IRI Data Library help desk, help@iri.columbia.edu, transferred the ENSOFORECAST directory to Vultax for public hosting on 21 October 2026. He supplied and MD5 `1e8b1b29f8e77697fbdbc90d3ca168ac`. IRI's alternate-sources sheet lists ENSOFORECAST with IRI as its only source and no restriction; the row is kept as `iri-documentation/alternate-sources-row.csv`. Vultax holds the preservation copy and its integrity documentation. This is not a claim of IRI endorsement or author review. ## Dates IRI directory last update: "May 2000", as IRI stated in its transfer email of 21 October 2026. The fetched GitHub catalog files do not themselves state that date. Decoded forecast target months span 1972-01 through 2001-02 across the 23 forecast files; each file has its own ending month in `MANIFEST.json`. Observed anomalies in `data/ossta.cdf` span 1972-01 through 2000-02. Absolute ERBE coordinates span 1985-02 through 1988-12 in `erbe.glob.cdf` and 1985-01 through 1989-12 in `erbe.part.cdf`. Three ERBE mean files contain a cyclic January to December axis without an absolute year. In `iri-documentation/descriptions/ensoforecasts.html.txt`, IRI calls the temporal coordinate "verification time". These are target or observation months; issue dates are not recorded in the files. No forecast issue or publication date has been derived from a lead, a target month or a filesystem timestamp. ## Contents All paths below are relative to the deposit folder. Bytes are logical file lengths. `MANIFEST.json` records original netCDF metadata, per-file hashes, decoded time axes and the basis of group assignments. An absent netCDF units or long_name attribute is recorded as null, not inferred. ### Nine IRI catalog files: LDEO1, LDEO2 and observed anomalies 9 files; 15,149,936 bytes. | Path | Bytes | Format | Role | |---|---:|---|---| | `data/fc03mon.cdf` | 1,720,844 | netCDF CDF-1 | LDEO1, lead 3 months | | `data/fc06mon.cdf` | 1,735,976 | netCDF CDF-1 | LDEO1, lead 6 months | | `data/fc09mon.cdf` | 1,751,108 | netCDF CDF-1 | LDEO1, lead 9 months | | `data/fc12mon.cdf` | 1,766,240 | netCDF CDF-1 | LDEO1, lead 12 months | | `data/fc2_03mon.cdf` | 1,594,816 | netCDF CDF-1 | LDEO2, lead 3 months | | `data/fc2_06mon.cdf` | 1,609,948 | netCDF CDF-1 | LDEO2, lead 6 months | | `data/fc2_09mon.cdf` | 1,625,080 | netCDF CDF-1 | LDEO2, lead 9 months | | `data/fc2_12mon.cdf` | 1,640,212 | netCDF CDF-1 | LDEO2, lead 12 months | | `data/ossta.cdf` | 1,705,712 | netCDF CDF-1 | observed anomalies | ### Fifteen additional files in the LDEO2 lead series 15 files; 24,103,824 bytes. | Path | Bytes | Format | Role | |---|---:|---|---| | `data/fc2_00mon.cdf` | 1,559,508 | netCDF CDF-1 | Lead 0 months, tentative from filename | | `data/fc2_01mon.cdf` | 1,564,552 | netCDF CDF-1 | Lead 1 months, tentative from filename | | `data/fc2_02mon.cdf` | 1,569,596 | netCDF CDF-1 | Lead 2 months, tentative from filename | | `data/fc2_04mon.cdf` | 1,579,684 | netCDF CDF-1 | Lead 4 months, tentative from filename | | `data/fc2_05mon.cdf` | 1,584,728 | netCDF CDF-1 | Lead 5 months, tentative from filename | | `data/fc2_07mon.cdf` | 1,594,816 | netCDF CDF-1 | Lead 7 months, tentative from filename | | `data/fc2_08mon.cdf` | 1,599,860 | netCDF CDF-1 | Lead 8 months, tentative from filename | | `data/fc2_10mon.cdf` | 1,609,948 | netCDF CDF-1 | Lead 10 months, tentative from filename | | `data/fc2_11mon.cdf` | 1,614,992 | netCDF CDF-1 | Lead 11 months, tentative from filename | | `data/fc2_13mon.cdf` | 1,625,080 | netCDF CDF-1 | Lead 13 months, tentative from filename | | `data/fc2_14mon.cdf` | 1,630,124 | netCDF CDF-1 | Lead 14 months, tentative from filename | | `data/fc2_15mon.cdf` | 1,635,168 | netCDF CDF-1 | Lead 15 months, tentative from filename | | `data/fc2_16mon.cdf` | 1,640,212 | netCDF CDF-1 | Lead 16 months, tentative from filename | | `data/fc2_17mon.cdf` | 1,645,256 | netCDF CDF-1 | Lead 17 months, tentative from filename | | `data/fc2_18mon.cdf` | 1,650,300 | netCDF CDF-1 | Lead 18 months, tentative from filename | ### Five ERBE netCDF files 5 files; 5,191,508 bytes. | Path | Bytes | Format | Role | |---|---:|---|---| | `data/erbe.sarb/erbe.glob.cdf` | 1,977,968 | netCDF CDF-1 | srb field | | `data/erbe.sarb/erbe.glob.cld.mean.cdf` | 506,144 | netCDF CDF-1 | srb field | | `data/erbe.sarb/erbe.glob.clr.mean.cdf` | 506,144 | netCDF CDF-1 | srb field | | `data/erbe.sarb/erbe.glob.for.mean.cdf` | 506,144 | netCDF CDF-1 | srb field | | `data/erbe.sarb/erbe.part.cdf` | 1,695,108 | netCDF CDF-1 | srb field | ### Remaining inputs and auxiliary files 20 files; 73,754,224 bytes. | Path | Bytes | Format | Role | |---|---:|---|---| | `data/CHRIS4yrH` | 1,598,400 | binary unidentified | Input or auxiliary file; see interpretation limits below | | `data/CHRIS4yrrm` | 1,598,400 | binary unidentified | Input or auxiliary file; see interpretation limits below | | `data/CHRISclimo` | 88,800 | binary unidentified | Input or auxiliary file; see interpretation limits below | | `data/cac.ssta` | 4,025,112 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/cac.sstf` | 4,025,112 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/cac.sstm` | 121,056 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/cacsm.ssta` | 4,025,112 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/cacsm.sstf` | 4,025,112 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/fsu.anom` | 9,962,992 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/fsu.full` | 9,962,992 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/fsu.mean` | 241,928 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/fsu4yrclim.ano` | 9,236,944 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/fsusm.full` | 9,236,944 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/fsusm4yrclim.a` | 9,236,944 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/rcmstress` | 273,696 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/servain.tau.RA` | 2,006,400 | binary unidentified | Input or auxiliary file; see interpretation limits below | | `data/slp.7989` | 777,616 | ASCII text | Input or auxiliary file; see interpretation limits below | | `data/tau7088.anom` | 767,904 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/tau8189am` | 2,542,752 | Fortran unformatted sequential big-endian | Input or auxiliary file; see interpretation limits below | | `data/test1` | 8 | ASCII text | Input or auxiliary file; see interpretation limits below | The four named LDEO1 leads, four named LDEO2 leads and shared CACSM observation file are directly mapped in IRI `LDEO1` and `LDEO2`. The remaining fc2 files are grouped with that filename series; their individual lead labels are tentative and do not imply an undocumented model version. ## How to read the files The 29 netCDF files are classic CDF-1 and can be opened with a netCDF reader that supports the classic format. `NETCDF-HEADERS.txt` is a CDL-style header dump of each file in path order, including attribute types. `MANIFEST.json` retains all dimensions, variable attributes and global attributes exactly as read. IRI catalog definitions sometimes add units; those catalog additions have not been inserted into the original file metadata. Time variable T with units monthtime uses months since January 1960. The preserved IRI manual `iri-documentation/references/ingrid-units.source.txt` describes the epoch as "months since January 1960". Integer part 0 is January 1960; integer part 144 is January 1972. Coordinates ending in .5 identify the containing month. Decode by taking the floor and adding that many calendar months to January 1960, not by assuming a fixed number of days. The ERBE files with units month and gridtype 1 retain only a cyclic month of year. Fourteen files have Fortran unformatted sequential records with big-endian 32-bit markers. Every leading marker equals its trailing marker and the records exhaust the file. A marker is the payload length in bytes, not a value count. Record payload schemas, grids, units and epochs are not supplied, so record validation does not decode their scientific contents. The CAC, FSU, CHRIS, stress, tau and Servain names support only tentative associations. IRI `descriptions/ensoforecasts.html` mentions FSU pseudo-stress forcing but gives no mapping to these particular input files. CHRIS4yrH, CHRIS4yrrm, CHRISclimo and servain.tau.RA remain binary unidentified. `slp.7989` is ASCII text with repeated SLP headers and two-digit year tokens; no century, grid or units have been assumed. `test1` is eight bytes of ASCII text with unknown purpose. Per-file marker counts and interpretation evidence are in `MANIFEST.json`. No archive member was executed or sourced. ## Integrity `SHA256SUMS` and `MD5SUMS` each cover all 49 data files, in sorted path order. Both use the standard two-space delimiter and LF line endings. Their paths begin with data/ and are relative to this folder. Source documentation is tracked separately in `iri-documentation/_fetch/receipts.json`. Original tgz MD5: `1e8b1b29f8e77697fbdbc90d3ca168ac`. Original tgz SHA-256: `ee0e93452bafaa08e64ade880c5b548d389d0653099be6e8b19de79e112037c7`. Original tgz compressed bytes: 78,463,022. Extracted regular-file bytes: 118,199,492. From the deposit folder, verify all data files in one POSIX shell line: ```sh sha256sum -c SHA256SUMS && md5sum -c MD5SUMS ``` ## Provenance and chain of custody - 21 October 2026: IRI supplied the tarball through Aaron Kaplan, by email. The saved HTTP headers report Last-Modified 2026-10-21T15:26:33Z. This is a server modification time; the actual tarball creation time is not recorded. Its gzip timestamp is zero. - 22 October 2026 at 13:06:29.695 UTC: the original download receipt records 78,463,022 bytes and the MD5 and SHA-256 above, with MD5 verified against Aaron's value. The receipt is preserved in `iri-documentation/transfer-records/archive-receipt.json`. - 22 October 2026: the archive was extracted and inventoried as 49 regular files. The first inventory was generated at 13:17:38.100 UTC. Original tar member timestamps are retained as provenance metadata and do not establish forecast or observation dates. - 26 October 2026: the tgz hashes were rechecked, tar header checksums were validated, extracted files were compared byte for byte against tar members, and local deposit copies were compared byte for byte and hashed. NetCDF metadata and time coordinates were checked against the 22 October inventory. IRI catalog documentation was fetched again from GitHub at commit `7b8bcdd8be1c2beec01e441bcc09e31d5854958a`. - Publication: 26 October 2026, at https://vultax.com/archive/ensoforecast with the files under https://vultax.com/datasets/ensoforecast-archive/. The record page lists every file with its checksum and marks any file that is not yet online. ## Rights and reuse IRI's alternate-sources sheet lists ENSOFORECAST as unrestricted and IRI transferred the directory to Vultax for public hosting on 21 October 2026. No written licence accompanies the files. Please cite the archive and the model as below, and name IRI and the Lamont-Doherty Earth Observatory as the source. If IRI states reuse terms, this file will record them. ## Citation Archive: Lamont-Doherty Earth Observatory climate group. 2026. ENSOFORECAST archive deposit. Published 26 October 2026. Preservation custodian: Vultax. . Files: . No DOI yet. `CITATION.cff` and `citation.bib` describe the archive, rather than assigning the model paper DOI to the data. Model: Stephen E. Zebiak and Mark A. Cane. 1987. A Model El Nino-Southern Oscillation. Monthly Weather Review. DOI . Related Vultax edition: . The edition is a separate interpretive work; it is not the unconverted deposit. ## Corrections research@vultax.com Report a path, checksum, metadata or provenance problem with the file name. Corrections should be dated and described without silently changing data bytes. ## Published 26 October 2026. Initial deposit. Data files are unchanged from the verified source tarball.