NASA's global temperature record (GISTEMP) in a spreadsheet
NASA's GISTEMP global temperature record in a spreadsheet: reading anomalies against the 1951-1980 base, decade averages, RANK formulas and hemisphere files.
NASA's GISTEMP record gives the global mean surface temperature as an anomaly, the difference from a 1951-1980 average, for every year from 1880. In the file retrieved on 2026-10-09, 2024 is the warmest year at 1.29 °C above that base period, and 2025 is second at 1.19 °C. The 2025 value is 1.41 °C above the average for 1880 to 1900. These figures come from the file as it stood on that date. NASA can revise the record, so state the retrieval date whenever you quote a value.
This article explains what GISTEMP is and why it uses anomalies, shows how to read the annual and seasonal columns, and gives the spreadsheet formulas for decade averages, changes since the late 1800s, rankings and moving averages. It also covers the hemisphere files, a warming-stripes chart, and the mistakes that most often distort the comparison.
What is GISTEMP and what data does it combine?
GISTEMP is the GISS Surface Temperature Analysis, produced by NASA's Goddard Institute for Space Studies. It combines two inputs. For land, it uses station records from NOAA's Global Historical Climatology Network (GHCN). For the ocean, it uses sea surface temperatures from NOAA's Extended Reconstructed Sea Surface Temperature dataset (ERSST). The version used here is version 4, published in the tabledata_v4 folder.
The analysis has three files that share one layout:
GLB.Ts+dSST.csv, the global land and ocean meanNH.Ts+dSST.csv, the Northern Hemisphere meanSH.Ts+dSST.csv, the Southern Hemisphere mean
Each file has a header line that names the series, then one row per year.
Why does GISTEMP use anomalies, not absolute temperatures?
Absolute temperatures differ widely from place to place. A desert station, a mountain station and a coastal station can differ by many degrees in the same month, and the station network has changed greatly since 1880. An anomaly compares each station with its own normal for the same month, so the stations can be averaged without the results depending on where they happen to sit.
GISTEMP expresses anomalies in degrees Celsius relative to the 1951-1980 average. A value of 0 means the same as that 30-year average. A value of −0.22 means 0.22 °C colder than it. Because the base is fixed, every year in the file can be compared with every other year.
How do you read the columns in a GISTEMP file?
Each row has the following columns:
| Column | Meaning |
|---|---|
| Year | Calendar year |
| Jan to Dec | Monthly anomalies |
| J-D | January to December annual mean |
| D-N | December of the previous year to November of this year |
| DJF | December of the previous year, January and February |
| MAM, JJA, SON | March to May, June to August, September to November |
A *** marks a missing value. Three cases come up often:
- The 1880 row shows
***for D-N and DJF, because December 1879 is not in the record. - The J-D value stays
***until all twelve months of a year are in. The 2026 row has January to August, so its J-D is***. - A missing month in the middle of a year leaves that year's J-D and seasons blank.
Some values from the global file, retrieved on 2026-10-09:
| Measure | Value (°C, relative to 1951-1980) |
|---|---|
| Last complete year, 2025 J-D | 1.19 |
| Warmest year in the file, 2024 J-D | 1.29 |
| Average of 1880 to 1900, J-D | −0.22 |
| 2025 minus the 1880 to 1900 average | 1.41 |
| Average of the 2010s | 0.81 |
| Average of 2020 to 2025 (six years) | 1.07 |
The 2025 row is complete for all twelve months, so its J-D is a full-year value. Its December anomaly is 1.06 °C.
How do you calculate GISTEMP decade averages?
The table gives the mean of the annual J-D values in each decade and the difference from the 1880 to 1900 average of −0.224 °C. The 2020s have six years of data, so their average is for 2020 to 2025 only.
| Decade | Average J-D (°C) | Difference from 1880-1900 (°C) |
|---|---|---|
| 1880s | −0.218 | 0.006 |
| 1890s | −0.243 | −0.019 |
| 1900s | −0.321 | −0.097 |
| 1910s | −0.335 | −0.111 |
| 1920s | −0.241 | −0.017 |
| 1930s | −0.124 | 0.100 |
| 1940s | 0.042 | 0.266 |
| 1950s | −0.048 | 0.176 |
| 1960s | −0.030 | 0.194 |
| 1970s | 0.035 | 0.259 |
| 1980s | 0.247 | 0.471 |
| 1990s | 0.387 | 0.611 |
| 2000s | 0.588 | 0.812 |
| 2010s | 0.807 | 1.031 |
| 2020s (2020 to 2025) | 1.067 | 1.290 |
The ten warmest years in the file, by J-D value, are:
| Rank | Year | J-D (°C) |
|---|---|---|
| 1 | 2024 | 1.29 |
| 2 | 2025 | 1.19 |
| 3 | 2023 | 1.17 |
| 4 | 2016 | 1.01 |
| 5 | 2020 | 1.01 |
| 6 | 2019 | 0.98 |
| 7 | 2017 | 0.92 |
| 8 | 2015 | 0.90 |
| 9 | 2022 | 0.89 |
| 10 | 2018 | 0.85 |
2016 and 2020 are tied at 1.01 in the two-decimal file, so their order is a tie, not a ranking.
Which formulas calculate GISTEMP averages in a sheet?
Import the CSV into a sheet named GISTEMP, with the header in row 1. Put the years in column A, rows 2 to 148, so that 1880 is in row 2, 2025 is in row 147 and 2026 is in row 148. The J-D column is N. Keep the values numeric, so that the averages below work.
- Average of the 1880s, in any cell:
=AVERAGE(N2:N11), which returns −0.218 - Average of the 2010s:
=AVERAGE(N132:N141), which returns 0.807 - Average of 1880 to 1900:
=AVERAGEIFS(N2:N148,A2:A148,">=1880",A2:A148,"<=1900"), which returns −0.224 - Change of 2025 since the 1880 to 1900 average:
=N147-AVERAGEIFS(N2:N148,A2:A148,">=1880",A2:A148,"<=1900"), which returns 1.414 °C - Rank of 2025 among the years:
=RANK(N147,N2:N148), which returns 2 - Warmest year:
=INDEX(A2:A148,MATCH(MAX(N2:N148),N2:N148,0)), which returns 2024 - Second warmest value:
=LARGE(N2:N148,2), which returns 1.19 - Number of years with a J-D value:
=COUNT(N2:N148), which returns 146
For a 10-year moving average, put =AVERAGE(N2:N11) in a new column, in row 11, and fill it down. Each row then averages itself and the nine years before it. Row 147 returns 1.016 °C for 2016 to 2025, which rounds to 1.02.
A least-squares line through the 1980 to 2025 values has a slope of 0.21 °C per decade. The formula is =SLOPE(N102:N147,A102:A147)*10, with 1980 in row 102. Read a slope from one dataset as a summary of that dataset, not as a forecast.
AVERAGE and AVERAGEIFS skip text, so the *** cells are left out. Do not replace *** with 0. A zero for a missing year would pull every average below its true value.
How do you build GISTEMP warming stripes in a spreadsheet?
Ed Hawkins of the University of Reading designed the warming stripes. Each year is one vertical stripe, coloured by its annual anomaly, with no axes, labels or numbers. The design shows the long-term change without the reader needing a scale.
To build a version in a spreadsheet, select N2:N147, the J-D values for 1880 to 2025, and add a three-colour scale from blue at the lowest value, through white, to red at the highest. Set the midpoint type to Number and enter 0, so that an anomaly of zero is white. Hide the grid and labels and narrow the columns, so that each year is one stripe. The stripes show the sequence of years, not the size of the change, so use the tables above for values.
How do the hemispheres compare in GISTEMP?
The Northern and Southern Hemisphere files have the same layout as the global file, so the same formulas apply. The values for 2024 and 2025 are:
| Series | 2024 J-D (°C) | 2025 J-D (°C) | Average 1880 to 1900 (°C) |
|---|---|---|---|
| Global (land and ocean) | 1.29 | 1.19 | −0.22 |
| Northern Hemisphere | 1.66 | 1.49 | −0.30 |
| Southern Hemisphere | 0.91 | 0.89 | −0.15 |
The global value is close to the average of the two hemispheres. In every year of the file, the global J-D value is within 0.02 °C of the mean of the NH and SH values, as expected when the two hemispheres have equal area. To check 2025, copy the NH and SH J-D values for 2025 into cells U147 and V147 of the global sheet, and use =(U147+V147)/2, which returns 1.19.
The hemispheres have not always been in the same order. The Northern Hemisphere J-D value is below the Southern Hemisphere value in 64 years of the file, including every year of the 1880s and nine of the 1900s. Read the hemisphere files as separate series, and do not assume that the Northern Hemisphere is always warmer.
For station-level monthly temperatures, the monthly weather for 60 world cities workbook covers 2025. It answers a different question: what individual places experienced, not the global mean.
What mistakes make GISTEMP comparisons misleading?
Mixing base periods. GISTEMP's anomalies are relative to 1951-1980. Other global datasets use different base periods. HadCRUT5 is relative to 1961-1990, and NOAA's GlobalTemp is relative to the 20th-century average. The anomalies from different bases differ by a near-constant offset, so comparing them directly gives the wrong level, even though the shape of the trend is similar. Check each dataset's documentation, and rebase one series before combining them.
Treating a partial year as complete. The 2026 row has eight months. Its January to August mean is 1.22 °C, which looks higher than the 2025 full-year value of 1.19 °C, but the periods differ. The 2025 January to August mean is 1.20 °C, so compare the same months of each year, and label partial-year values as year to date.
Replacing missing values with zero. A zero for *** is a value of 0 °C relative to the base period, which the record does not say. Leave the cell as text, and use functions that skip it.
Reading one year as a trend. 2024 is the warmest year in the file, but a single year is noisy. The 1980 to 2025 trend line is the better summary.
How do you cite GISTEMP data?
Credit NASA Goddard Institute for Space Studies, GISS Surface Temperature Analysis (GISTEMP), version 4. Name the file and the date you retrieved it, because NASA revises the analysis and a later download may differ. The reliable public datasets for spreadsheets article covers other sources that keep their own base periods and revision histories.