Two reputable sources can describe the same event with different numbers without contradicting each other. They may use different reference periods, rounding conventions, coverage or methods. A careful comparison begins by identifying the quantity each source reports, then asks which conclusions survive after those differences are made explicit.
This guide works through NASA and NOAA reports published in January 2024 about global surface temperature in the calendar year 2023. It is a historical source-comparison exercise, not the latest annual climate assessment. Statements about a record are read within the period covered by those reports; this page does not claim that 2023 remains the warmest year today.
You will build a compact source record, compare definitions, work through original arithmetic examples and write a reasoned judgement. You need basic subtraction and percentages only for a few extension questions. The important skill is keeping a number attached to its meaning while moving between sources.
This workshop belongs to Guided Reading and Inquiry. For the underlying research method, use How Research Methods and Source Evaluation Work. For the climate system itself, use How Climate Works.
Frame a question that the sources can answer
Our question is: “How should a reader interpret the two agencies’ reported annual temperature anomalies and record statements for 2023?” It is narrow enough to answer from the selected reports, while still requiring attention to definitions and evidence. It does not ask us to reproduce an entire global temperature analysis.
A less useful starting question would be “Which agency is right?” That framing assumes a disagreement before we have established that the numbers refer to comparable quantities. Another weak question is “Are both sources reliable?” Reliability depends on the claim, method and context; a general reputation judgement does not perform the comparison for us.
Write down the event period separately from the publication period. The year being measured is not the year in which the summary appears. This distinction is routine in annual statistics, school results, financial reporting and historical research. Losing it can turn a report about a completed period into a mistaken claim about the present.
The intended output is a short comparison supported by an evidence table and a calculation notebook. The table preserves source details; the notebook separates reported values from your own transformations. Together they make the judgement easier for another reader to check.
The two source records
Source A is NASA’s NASA Analysis Confirms 2023 as Warmest Year on Record, a January 2024 agency release. Its opening temperature paragraph reports about 1.2°C above the 1951–1980 average. Its record discussion refers to modern record-keeping beginning around 1880. Use the opening paragraph, record discussion and “Open Science in Action” section as the relevant locators.
Source B is NOAA NCEI’s Assessing the Global Climate in 2023, published on 12 January 2024. Its “Surface Temperature” section reports 1.18°C above the twentieth-century average and places the annual result within the 1850–2023 record. That section is the locator for the comparison here; other sections cover different indicators and periods.
| Field | Source A: NASA release | Source B: NOAA NCEI summary |
|---|---|---|
| Measured period | Calendar year 2023 | Calendar year 2023 |
| Quantity selected here | Global surface temperature anomaly | Global surface temperature anomaly |
| Celsius anomaly as displayed | About 1.2°C | 1.18°C |
| Reference average | 1951–1980 | Twentieth century |
| Record context | Beginning around 1880 | 1850–2023 |
| Useful locator | Opening paragraph; record discussion | Surface Temperature section |
This is a deliberately small source record. It contains what the comparison needs without copying whole reports. Preserve the page title and URL with your notes, along with the date on which you consulted it. If a later task requires the underlying numerical series, record that dataset’s version and variables separately rather than treating the release as a substitute for the data.
Observation before interpretation
Begin with observations that someone else can find in the selected sources: a reported number, its unit, its baseline and its time period. Then identify your interpretation: the numbers should not be subtracted as if they used one common reference. The interpretation is a reasoned conclusion from the metadata, not another sentence simply copied from a report.
This separation makes disagreements easier to resolve. If two readers disagree about the baseline, they can inspect the relevant paragraph. If they agree about the baseline but disagree about what comparison is legitimate, the problem is conceptual or mathematical. Mixing these levels makes every disagreement look like a dispute over facts.
Use three labels in a working notebook: reported, calculated and inferred. “Reported” preserves what a source states. “Calculated” records a transformation with its inputs and operation. “Inferred” explains a judgement and its assumptions. The labels need not appear in every final paragraph, but they help prevent an invented calculation from acquiring the authority of a cited source.
For example, a hypothetical rebasing exercise below is calculated from invented teaching values. It is not a newly discovered agency estimate. Keeping that distinction visible allows us to explain the method without pretending to have performed the agencies’ analyses.
An anomaly is a difference from a reference
A temperature anomaly is a difference between a temperature measure and a specified reference average. The reference is part of the quantity’s definition. Saying “the anomaly is one degree” without naming the reference leaves an important piece of meaning unspecified.
Consider an original fictional example unrelated to the agencies’ actual climatologies. A teaching series has an annual value of 15.4°C. Reference B has a mean of 14.2°C, so the anomaly relative to B is 15.4 − 14.2 = 1.2°C. Reference C has a mean of 14.0°C, so the anomaly relative to C is 15.4 − 14.0 = 1.4°C.
The annual value has not changed. The reference has changed. Both anomalies correctly describe the same teaching value relative to their stated baselines. Treating their 0.2°C difference as a disagreement about the annual value would be a category mistake caused by dropping the reference information.
An anomaly is also different from the absolute temperature itself. It is not automatically the increase from the previous year, and it is not a statement that every place on Earth experienced that exact change. Those are different quantities requiring different comparisons or spatial information.
Rebase one series step by step
For one consistently defined series, write the anomaly relative to B as annual value minus B. To express it relative to C, add B minus C. In words: anomaly relative to C = anomaly relative to B + the difference between baseline B and baseline C.
Using the fictional values above, start with 1.2°C relative to B. The baseline difference is 14.2 − 14.0 = 0.2°C. Add it to obtain 1.4°C relative to C. You can check the result directly against 15.4 − 14.0. Both routes produce the same answer because they describe the same series and period.
The phrase “one consistently defined series” matters. Rebasing does not automatically remove differences in data coverage, processing, averaging or later revisions between distinct series. A common reference makes quantities more comparable, but it does not make every other methodological difference disappear.
Do not use the two agencies’ rounded headline values to invent the exact difference between their baseline means. That difference would need appropriate information from the relevant series. The arithmetic rule tells us what information is needed; it does not supply missing data by itself.
Separate annual change from baseline difference
Extend the fictional series with a previous annual value of 15.1°C. Its anomaly relative to B is 0.9°C, and relative to C it is 1.1°C. The later year’s anomalies are 1.2°C and 1.4°C. In either reference system, the year-to-year difference is 0.3°C.
| Fictional quantity | Relative to B, mean 14.2°C | Relative to C, mean 14.0°C |
|---|---|---|
| Earlier annual value, 15.1°C | 0.9°C | 1.1°C |
| Later annual value, 15.4°C | 1.2°C | 1.4°C |
| Later minus earlier | 0.3°C | 0.3°C |
Subtracting two anomalies using the same fixed baseline cancels that baseline. Subtracting anomalies from different baselines does not have the same meaning. This is why a chart assembled from several sources needs careful harmonisation before its points can be treated as one time series.
The exercise also shows why an anomaly headline should not be read as an annual jump. The later anomaly relative to C is 1.4°C, but the annual increase in this invented example is 0.3°C. The two numbers answer different questions, even though both are measured in degrees Celsius.
Rounding is information about presentation
A value shown to one decimal place should not silently acquire an extra digit of certainty. Writing 1.2 as 1.20 may be algebraically harmless in a calculation, but it can suggest a reporting precision the original display did not supply. Keep a note of how the source actually presented the value.
Suppose a fictional measurement rounds to 1.2 at one decimal place. Under a stated round-to-nearest convention, many underlying values close to 1.2 can produce that display. The rounding interval describes display behaviour; it is not a scientific confidence interval or a complete statement of measurement uncertainty.
Likewise, two displayed numbers can differ because of rounding while their underlying values are closer than the display suggests. They can also agree after rounding while their underlying estimates differ. Neither display agreement nor display disagreement substitutes for understanding what was measured and how uncertainty was assessed.
In our historical comparison, baseline definitions must be considered before attaching meaning to the subtraction 1.20 − 1.18. The calculator can return 0.02 correctly while the proposed interpretation remains invalid. Arithmetic correctness is only one requirement for a meaningful comparison.
Convert temperature differences correctly
For a temperature difference, multiply a Celsius difference by 9/5, or 1.8, to express it in Fahrenheit degrees. Do not add 32. The offset used when converting an absolute Celsius temperature cancels when one temperature is subtracted from another.
An original example makes this visible. Absolute temperatures of 10°C and 12°C convert to 50°F and 53.6°F. Their difference is 2°C or 3.6°F. Applying 2 × 1.8 + 32 would incorrectly produce 35.6°F because it would treat a difference as though it were an absolute reading.
When converting a rounded published value, carry its approximate status into your wording. Extra digits produced by multiplication do not restore digits that were never reported. If exact cross-unit consistency is important, consult appropriately precise values from the source instead of repeatedly converting rounded headlines back and forth.
This distinction transfers beyond climate reporting. Changes in length, time and currency values may have their own transformation rules and reference conditions. Before applying a familiar formula, identify whether you are converting an absolute quantity, a difference, a rate or an index.
Compare the record statements within their windows
A record statement describes a maximum or minimum within a defined collection of observations. Its time window belongs in the claim. “Highest in this series through the reported year” does not mean “highest for all time,” and it does not predict what a later observation will show.
Two series can agree on the highest year within their respective windows even if one window begins earlier. That agreement is meaningful within the stated records. It does not establish that the records have identical coverage or that their numerical estimates must match digit for digit.
For a historical assignment, write in the time frame of the selected releases. If you want to answer a current question, retrieve current assessments and define a new comparison. Do not quietly combine an old headline with a later data point and present the result as if both belonged to the same publication moment.
This habit is useful in sport, economics and education as well as science. A record based on a particular series can change when new observations arrive or when the series is revised. The appropriate response is to preserve the earlier claim’s context and identify the basis of the newer claim.
Two sources are not automatically two independent datasets
NASA’s Open Science in Action discussion explains that the agencies use much of the same temperature data while applying different methods. Agreement between analyses is therefore valuable, but it should not be described as two wholly separate sets of raw observations.
Distinguish organisational independence, analytical independence and independence of underlying observations. Two organisations may make separate methodological decisions while drawing partly on the same measurement networks. A newspaper can add editorial explanation while simply repeating an agency’s number. These relationships affect what each additional source contributes.
Consider a fictional source chain. An observatory publishes a count. A newspaper summarises it. A school newsletter copies the newspaper’s figure. Finding the same count in all three places is evidence of transmission, not three independent counts. The newsletter may still help explain local relevance, but it does not multiply the original measurement.
A useful source comparison therefore asks both “who produced this page?” and “where did this claim originate?” Follow the specific number or observation to its source when possible. Avoid making independence a yes-or-no badge attached to an entire publication when different claims may have different lineages.
What the selected evidence supports
The chosen passages support a comparison of reported annual anomalies, reference periods and record contexts. They also allow us to explain why a direct subtraction of the displayed anomalies is not a fair measure of disagreement. These are bounded conclusions that can be traced to the selected material and the arithmetic principles developed here.
The two selected ranking statements alone do not reproduce the full evidence for climate mechanisms or attribution. A source may discuss those subjects elsewhere, but a comparison limited to its headline numbers should not pretend to have assessed every underlying analysis. Use the wider climate reading when that becomes the research question.
Similarly, a global annual summary cannot establish the experience of every neighbourhood or every day. A reader asking about a particular Singapore location needs appropriately local observations and time definitions. The broad result and a local pattern answer different questions; the existence of spatial variation does not by itself refute a global average.
This is a disciplined use of scope, not a reason to disregard the reports. Strong evidence becomes more useful when we identify the claim it supports precisely. Overextending it creates avoidable weaknesses; refusing to recognise its legitimate conclusion is a different mistake.
Build a claim-to-evidence table
| Proposed claim | Evidence needed | Judgement for this exercise |
|---|---|---|
| The reports use the same baseline | Each report’s reference-period definition | Not supported; the definitions differ |
| The displayed values can be subtracted as a direct disagreement measure | Comparable quantity, baseline, precision and method context | Not justified by the two headline values |
| Both reports concern the same calendar year | Stated measured period | Supported by the selected passages |
| Three news pages repeating one agency figure are three independent measurements | Evidence of separate observation processes | Repetition alone does not establish this |
| The historical record statement remains current indefinitely | Later observations and current assessment | Outside the selected historical comparison |
The table helps separate different strengths of conclusion. “Not supported by these passages” does not always mean “false.” Sometimes the claim requires additional evidence. In the first row, we can identify a specific incompatibility with the claim. In the final row, the problem is a move beyond the available period.
Use this distinction in your own writing. “The source does not establish X” is often more accurate than “the source proves X is wrong.” A good evaluation names whether the obstacle is contradiction, missing evidence, incomparable definitions or an unsupported extension of scope.
A short model comparison
“The two historical reports describe the same annual period, but their displayed anomalies use different reference averages. Their numerical difference should therefore not be treated as a direct measure of disagreement. A fair comparison preserves the baseline, unit, rounding and record window attached to each value. Agreement in the record conclusion is informative within the reports’ stated periods, while shared observational inputs mean that two agency analyses should not be counted as wholly independent measurement collections.”
This paragraph does several jobs without repeating the entire evidence table. It answers the research question, identifies the principal comparability problem and recognises the value and limit of agreement. The source records above supply the exact historical details; the paragraph explains how to interpret them.
Notice its restraint. It does not announce that the numbers are identical, that one organisation has made an error or that all methodological differences have been resolved. It also avoids a vague conclusion that “both sources are useful.” The reader learns what kind of comparison is legitimate and why.
For an assessed piece of writing, attach the relevant links or citation style to the claims and include the table as supporting work if permitted. A polished paragraph should remain traceable to evidence even when the working notes are more detailed than the final prose.
Improve a weak comparison
Here is an original flawed draft: “The figures differ by 0.02 degrees, so one source must be inaccurate. The newer-looking page is probably better. Three websites repeat its number, proving that it is independently confirmed. The record headline means the same statement must still be true today.”
The first sentence subtracts values before aligning their definitions. The second substitutes visual appearance for evidence about the relevant publication and dataset. The third confuses repeated reporting with independent measurement. The fourth removes the time boundary from a historical claim. These are distinct errors and need distinct repairs.
A stronger draft is: “Before comparing the numbers, I need to preserve their units, reference periods and presentation precision. The page’s appearance does not establish which analysis is appropriate for this question. I should trace repeated figures to their origin and identify what each source contributes. Because the selected reports describe a completed historical period, a current record claim would require a current assessment.”
The revision is a method paragraph rather than a complete factual answer. To finish the assignment, add the actual source details from the table and explain the supported conclusion. Good source evaluation combines an accurate record with reasoning; a list of sensible cautions alone does not answer the question.
When a later version changes a number
Imagine a separate fictional agency report that originally gives an annual index of 104.2. A later dataset gives 104.3 after a documented revision. The difference could be real and legitimate; it does not automatically mean that the earlier reporter fabricated a number. The first task is to establish the version history and the reason for the change.
For a question about what was reported at the time, the original release is relevant evidence. For a question requiring the best current estimate, the revised series may be appropriate. For a question about revision itself, both versions and the change documentation matter. The research purpose determines which record is needed.
Do not silently replace a quoted historical value with a current dataset value while leaving the old citation. That creates a source mismatch. Record the publication, version or retrieval information that lets another reader locate the value you actually used, and explain any transformation or update.
The library’s How Statistical Revisions, Data Vintages and Benchmark Updates Work develops this issue further. The present exercise needs only the practical habit: a number belongs to a defined source state, not merely to a website’s general reputation.
Practice questions before the answers
Use the source table for the historical comparison and the explicitly fictional values for calculations. Keep reported and invented numbers separate in your notes.
- Why is publication in January 2024 compatible with a report about the calendar year 2023?
- What must be checked before interpreting the subtraction of two reported anomalies?
- In the fictional series, a value is 16.0°C, baseline B is 14.5°C and baseline C is 14.1°C. Find both anomalies and the rebasing adjustment.
- Convert a fictional temperature difference of 0.5°C into Fahrenheit degrees. Should you add 32?
- Three articles repeat one institute’s count. What would establish an additional independent measurement?
- Does a rounding interval around a displayed value automatically describe its scientific uncertainty?
- A historical report gives a record through its final observation year. What evidence is needed to turn it into a current record statement?
- A global average rises while one local area cools. Is that pattern logically impossible?
For Questions 2, 5 and 7, name a concrete piece of information you would request. An answer such as “check reliability” is too broad to guide the next step. A useful request points to a definition, source lineage, version or observation that could resolve the uncertainty.
Worked answers and commentary
Question 1: An annual summary can be published after the measured year ends. The observation period and publication date serve different roles. A careful citation records both so that a later reader does not mistake the report date for the period measured.
Question 2: Check that the quantities refer to comparable time periods, spatial scope, units and reference averages, then consider rounding and methods. In this case, the baseline definitions already prevent treating the bare subtraction as a direct disagreement measure. Ask for suitable series information before attempting harmonisation.
Question 3: The anomaly relative to B is 16.0 − 14.5 = 1.5°C. Relative to C it is 16.0 − 14.1 = 1.9°C. The adjustment from B to C is 14.5 − 14.1 = 0.4°C, which is added to 1.5°C. These are teaching values, not agency observations.
Question 4: Multiply 0.5 by 1.8 to obtain 0.9 Fahrenheit degrees. Do not add 32 because the quantity is a difference. Adding the offset would convert the wrong kind of quantity and produce a result with the wrong meaning.
Question 5: You need evidence that another observation or measurement process produced a result, rather than merely repeating the original institute’s count. Separate analysis of shared observations may add value, but its contribution should be described accurately instead of being labelled wholly independent raw data.
Question 6: No. Rounding describes how a value is displayed. Scientific uncertainty may involve measurement, sampling, coverage, models or other features of the analysis. A display rule does not supply a complete uncertainty assessment.
Question 7: Retrieve a current assessment with the later observations and relevant series definition. Preserve the earlier statement as a historical claim. A page remaining online is not evidence that no later observation has changed the record.
Question 8: No. An average can rise while some components fall, depending on their values and weights. A local question needs local evidence; the local exception does not by itself contradict a properly defined global aggregate.
Change the question and choose evidence again
Suppose your new question is: “Was my neighbourhood warmer on a particular afternoon than on the same date last year?” The selected annual global reports are no longer sufficient. You need appropriately located observations, comparable times, measurement definitions and a reason to choose that particular comparison.
Or suppose the question becomes: “How did the agencies’ methods differ?” The releases may point you toward the answer, but you would need the relevant methodological documentation and dataset context. A source that is sufficient for a brief annual summary may be insufficient for reproducing a technical comparison.
This is a useful test of source selection. Keep the sources fixed for a moment and change the question. If you continue using them to answer everything, you may be treating institutional reputation as unlimited authority. If you identify precisely what new evidence is needed, you are matching the source to the task.
Write a two-sentence handover for either new question: what the present sources establish, and what must be obtained next. That handover allows another reader to continue the inquiry without inheriting a claim stronger than the evidence you actually examined.
A delayed task with a different kind of number
Tomorrow, compare two fictional school-library reports. Report A says that borrowing rose 20% from 500 loans. Report B says that borrowing rose by 100 loans. Ask whether the reports might describe the same change. The calculation 20% of 500 = 100 shows that they can, provided the period, unit and population of loans align.
Now change Report B to 100 additional borrowers. The numerical match no longer settles the comparison because loans and people are different units. One borrower can account for several loans. You need definitions and counts suited to the revised claim rather than another repetition of the percentage calculation.
This task transfers the central habit beyond temperature: align the quantity before comparing the number. Return to the Geography and Maps Library for spatial representations, or to Guided Reading and Inquiry for another worked path. Carry the source record, the transformation and the scope of the conclusion together.