What's Happening with Global Precipitation?

Land vs. Ocean, Observations vs. Models and the Official Narrative

This is another post with two main topics: climate data and the climate narrative. The data focuses on global precipitation—how it has changed over the past forty years, especially during the anomalously warm year of 2023. The narrative aspect is about how these changes align with expectations from global climate models, and how reporting differs when models agree or disagree with observed data.

In a previous post, “Why it is important to read scientific papers beyond their abstracts, especially when it is about the role of CO₂ in climate”, I discussed that our information space dominated by the carbon-focused narrative makes for

a caricature world where some important pieces of information and thoughts that we do not know about have been erased while others are arbitrarily highlighted. And this caricature reinforces itself, because those who do not fit in, have no voice.

How a rigid narrative steals information from us

A striking example: after the devastating Los Angeles fires in January, a non–peer-reviewed study that lacked statistically significant results—but attributed the fires to fossil fuel–driven global warming—received widespread media coverage. In contrast, more informed discussions framing the fires as a complex land use issue, emphasizing landscape drying from mismanagement and vegetation degradation, remain severely underreported. This lack of meaningful debate enables harmful policies like the “Fix Our Forests” Act in the U.S.

Image taken from Lucie Wuetrich’s post on LinkedIn.

Or consider this abstract of a Science paper:

Carbon isotopes in fossil emu (Dromaius novaehollandiae) eggshell from Lake Eyre, South Australia, demonstrate that the relative abundance of C₄ grasses varied substantially during the past 65,000 years. Currently, C₄ grasses are more abundant in regions that are increasingly affected by warm-season precipitation. Thus, an expansion of C₄ grasses likely reflects an increase in the relative effectiveness of the Australian summer monsoon, which controls summer precipitation over Lake Eyre. The data imply that the Australian monsoon was most effective between 45,000 and 65,000 years ago, least effective during the Last Glacial Maximum, and moderately effective during the Holocene.

It is interesting, but doesn’t sound particularly thrilling or existential. Yet the paper discusses one of the most dramatic and large-scale human-made ecological, hydrological, and climatic catastrophes—one that has nothing to do with CO₂. As if not to cause a second thought in the reader, who must remember at all times that all climate change is CO₂-related, the abstract does not even hint at the paper’s most important conclusions. I will discuss this paper at the end of today’s post, as it also relates to precipitation.

Precipitation in a warming world

In global climate models, global precipitation increases with rising planetary temperatures. This growth reflects a corresponding increase in evaporation, as evaporation and precipitation must balance globally over short timescales due to the limited amount of water vapor in the atmosphere.

Evaporation is ultimately constrained by net radiation, which comprises absorbed solar radiation and downwelling thermal radiation from the atmosphere, minus upwelling thermal radiation from the surface. As planetary albedo decreases—such as through ice and snow melt—the Earth absorbs more sunlight, leading to increased evaporation, assuming other factors remain constant.​

Furthermore, in warmer air, a larger fraction of net energy is allocated to latent heat (evaporation and transpiration) rather than sensible heating. Sensible heat flux is proportional to the temperature difference (ΔT) between the surface and the atmosphere, while evaporation is proportional to the difference in water vapor concentration (Δq) just above the surface versus a reference height (typically around 10 meters).​

The ratio between latent and sensible heat fluxes is therefore proportional to Δq/ΔT, indicating how rapidly water vapor concentration increases with temperature. Since this rate itself grows as the air warms—governed by the Clausius-Clapeyron relation—a warmer climate, for the same net radiation, will direct a larger share of energy into evaporation rather than into sensible heating.

Considering these two effects—an increase in absorbed solar radiation and a growing proportion of energy allocated to evaporation—global climate models project a 1–3% increase in precipitation for each degree Kelvin of global temperature rise.

What do the data say?

There are several global precipitation datasets, and today, we will focus on GPCP — the Global Precipitation Climatology Project. This dataset “provides a consistent analysis of global precipitation from an integration of various satellite data sets over land and ocean and a gauge analysis over land.” The data are freely accessible to all Earth citizens, and as a theorist, I am deeply grateful to humanity for this treat. (I repeat that it will be a shame if we don’t understand anything from the wealth of data at our disposal and proceed to the approaching dark ages as ignorant about the role of natural ecosystems in climate stabilization as before.)

Below is Figure 1 from the study of Gu and Adler 2023 “Observed variability and trends in global precipitation during 1979–2020”, which compares observations (solid black curve, GPCP) with models (red, green and blue curves).

The blue curve (CMIP6-hist) shows the ensemble mean of approximately thirty global climate models run from 1979 to 2014 using historical data on greenhouse gas concentrations and other radiative forcing inputs, such as albedo. The green curve (CMIP6-histGHG) represents the same model ensemble extended to 2020, only using historical greenhouse gas data. The red curve (AMIP6) shows simulations in which surface temperature was constrained to follow observed historical values. The numbers indicate the corresponding trends in precipitation, expressed in mm/day per decade.

The researchers found that for 1979–2014, model trends were larger and statistically significant at the 95% confidence level over land, ocean, and globally. In contrast, observed trends over the same period were not significant, reaching statistical significance only at the 90% level when extended to the longer period of 1979–2020.

However, the broad quantitative similarity between the modeled and observed trends, with the latter nearing statistical significance, led the researchers to anticipate that models and observations may converge in the near future as more data become available.

Reflecting this expectation, while the abstract of the paper acknowledges that:

Long-term change/trend in global mean precipitation is generally weak in GPCP. Although the GPCP trend is statistically significant at the 90% confidence level over global land + ocean during 1979–2020, it is not significant over either global land or ocean separately. For the shorter, overlap period with the CMIP6 historical experiments (1979–2014), GPCP positive trends can’t reach the 90% confidence level, while significant and more intense precipitation trends appear in CMIP6 ensemble-means.

it concludes with the implication that:

the GHG effect would become more readily detectable in observed precipitation in the near future with regards to both global mean and regional precipitation changes.

The Extra Warm Year 2023

The above analysis covered the period from 1979 to 2020. Then came the “near future” in the form of 2023, marked by an abrupt positive anomaly in global mean surface temperature. As discussed in “Seeing Forests Through Clouds - A 300-Words Science Haiku,” this anomaly was likely driven by reduced cloud cover and lower planetary albedo—meaning more energy available for evaporation and precipitation. It thus seems that 2023 offers a particular opportunity to see whether the models are getting the global water cycle right.

Probably motivated by these considerations, the same authors, Adler and Gu (2024), published a study “Global Precipitation for the Year 2023 and How It Relates to Longer Term Variations and Trends”. This is what they found:

Note that between 2000 and 2023, the GPCP dataset underwent a major update, transitioning from version 2.3—used by Gu and Adler (2023), with a resolution of 2.5° × 2.5°—to version 3.2—used by Adler and Gu (2024), which offers a higher resolution of 0.5° × 0.5°. In GPCP v. 3.2, data are available from the year 1983.

Comparing Fig. 7 of Adler and Gu (2024) with Fig. 1a of Gu and Adler (2023), we can see that the anticipation of increasing precipitation in the near future was based on several recent annual peaks that appeared to be more frequent. Had any of the years 2021, 2022, or 2023 shown a similar peak, the upward trend might have reached statistical significance. However, this did not happen. Both 2020 and 2021 saw an unprecedented negative anomaly, and even the exceptionally warm 2023 barely returned to the long-term mean.

It is hard to imagine that, had the exceptionally warm year of 2023 shown a global precipitation peak, it wouldn’t have been highlighted in the abstract as evidence supporting climate model predictions. In contrast, the actual abstract of Adler and Gu (2024) makes no mention of any mismatch between models and observations. Instead, it states:

In this paper, the global distribution of precipitation for 2023, in terms of global totals and regional anomaly patterns, is analyzed using information from the new Global Precipitation Climatology Project (GPCP) V3.2 Monthly product, including how the precipitation amounts and patterns from 2023 fit into the longer record from 1983–2023. … Comparison of the observed regional trend maps with climate model results indicates similarity between the observations and the model results forced by observed SSTs, while the “free-running” model ensemble shows only a broad general agreement over large regions.

One might conclude from this that observations from 1983 to 2023 generally agree with model projections. However, the paper actually compares models and observations only for 1983–2014—an even shorter period than in the authors’ previous study. Despite 2023 being the warmest year on record, its global precipitation was unremarkable. Yet the abstract still ends with a nod to global warming:

The ITCZ (Inter-Tropical Convergence Zone) latitude band, 0–10◦ N, sets a record high mean rain rate in 2023 after a steady upward trend over the decades, probably a response related to global warming.

In other words, while some regions saw below-average precipitation—including a significant negative anomaly over land—the abstract highlights isolated findings that can be read as supporting the carbon-focused narrative. Land use is never mentioned.

Negative precipitation anomaly over land in 2023

In their earlier study, Gu and Adler (2023) found that while the global (land + ocean) precipitation trend was significant at the 90% confidence level, the trends over land and ocean individually were not. In contrast, Adler and Gu (2024) do not present separate graphs for land and ocean, focusing solely on global precipitation—without reporting a trend, unlike in their previous work. They do note, however, that 2023 featured a negative precipitation anomaly over land and a positive one over the ocean.

My colleagues and I decided to look into this matter. Here’s what we found, having analyzed GPCP v. 3.2 for land and ocean separately:

Fig. 3 from Makarieva A.M., Nefiodov A.V., Cuartas L.A., Nobre A.D., Andrade D., Pasini F., Nobre P. “Assessing changes in atmospheric circulation due to ecohydrological restoration: How can global climate models help?” (in revision).

For the period 1983–2023, there are no significant trends (indicated by dashed lines) over land, ocean, or globally, with the warmest year 2023 showing an unprecedented negative anomaly over land. In the global mean (panel c), the trend—though statistically insignificant—is several times lower than what global climate models would predict, assuming a 1 K temperature rise over the period and a 1.5% increase in precipitation per 1 K of warming.

Communication nuances that matter

While Adler and Gu (2024) did not report any trends for their global data (Fig. 7 above), in the text they stated as follows:

Global total precipitation has remained nearly steady over the GPCP era, with only a small 1.5%/K rise (not statistically significant) compared to global surface temperature, and similar to climate models [Gu and Adler 2023]. …

Figure 7 is a plot of annual global total precipitation anomalies from 1983 to 2023. The variations have about a 3% range around the mean of 2.81 mm day⁻¹ and there is a slight upward trend (not statistically significant) as previously mentioned, which may be more of an effect of a shift in the Pacific Decadal Oscillation (PDO) around 1998 [Gu and Adler 2023].

In other words, when the GPCP v. 2.3 data for 1979–2020 showed a trend significant at the 90% level and broadly consistent with model projections, it was explicitly displayed on the graph—along with other statistically insignificant trends that were nonetheless numerically close to the modeled values. In contrast, when the updated GPCP v. 3.2 data for 1983–2023 (including the warmest year on record) showed a negligible trend, it was not explicitly mentioned. Nevertheless, the earlier study is cited in a way that (quite misleadingly) suggests the present findings are in agreement with it.

Another nuance is that when we compared our own analysis with the data manually read from Fig. 7 of Adler and Gu (2024), we obtained the following:

This figure shows the global precipitation anomaly in mm/day, based on Fig. 7 of Adler and Gu (2024) (black curve) and our own analysis of the same data (dashed curve). The two curves match perfectly for all years except 2023, where our analysis indicates a slight negative anomaly, while Adler and Gu (2024) report a slight positive one. The cause of this discrepancy remains unclear.

Finally, it's also relevant to the communication problem that the results of Gu and Adler (2023), which more explicitly conformed to model predictions, were published in Climate Dynamics—a Q1 journal in the field of climate science. Meanwhile, the follow-up study by the same authors using the same dataset, but without explicitly endorsing model projections, was published in Atmosphere, a Q2 journal. When researchers apply for grants, it matters which journals they publish in—Q1 ranking higher than Q2, Q3, or Q4. This gives some insight into how researchers can be nudged toward the “right” narrative and potentially penalized for straying from it.

Implications for science

A long-term, ubiquitous but nearly invisible manipulation of which data should be highlighted and which should be hidden or downplayed results in the degradation of the scientific process.

Stating clearly that the observed trends in global precipitation do not conform to climate models is important. As we have seen in “We are losing soil moisture, why?”, a decline in precipitation over land is consistent with independent assessments of soil moisture stores—and would be incompatible with the rising precipitation projected by the models. Also, a relative reduction in precipitation by 1% due to reduced transpiration (as compared to the pure CO₂-driven warming) can result in significant additional global warming, as we have discussed in “Global cooling from plant transpiraton”.

I’ll conclude with a quote from the Science paper by Johnson et al. (1999), “65,000 years of vegetation change in central Australia and the Australian summer monsoon.” I quoted its abstract at the beginning of this post—which, notably, said little more than that the paper was about extinct Australian fauna and monsoons. You may want to scroll back and reread it to appreciate just how remarkably non-ideological it was. But the paper itself stated something quite revolutionary:

Over the past 65,000 years, environmental factors other than climate have substantially influenced Australian ecology (25). Vegetation change in northeastern and southeastern Australia, brought about by an increase in fire frequency (26), has been attributed to the arrival of the first human immigrants at ∼60 ka (5) and has been suggested as the cause of extinction of G. newtoni at ∼50 ka (6). Our isotopic data are consistent with a human overprint on natural climate change. The effectiveness of the summer monsoon at Lake Eyre decreased substantially at approximately the same time as megafauna extinction (6) and never fully recovered, despite an invigorated planetary monsoon during the early Holocene (27). The transfer of moisture from the biosphere to the atmosphere is an important feedback mechanism that enhances the penetration of monsoon moisture into the interior of other continents (28). A change in vegetation type across northern Australia brought about by the burning practices of the first human colonizers may have reduced this wet-season feedback and, consequently, diminished the effectiveness of the summer monsoon at Lake Eyre during the early Holocene (29).

Preserving natural ecosystems is crucially important for the water cycle on land. Let us stop burning them out.

Implications for our ability to think

A recent Nature study showed that when AI systems are left to recycle their own outputs as inputs, they rapidly degenerate into something completely nonsensical. The researchers concluded that continued human input to the web will be crucial.

But if we limit ourselves to rigid narratives that are forced to propagate from one scientific abstract to the next, we won’t do much better than those poor AI systems feeding on their own garbage.

As Noam Chomsky stated, human thinking is based on “the innate, genetically installed “operating system” that endows humans with the capacity to generate complex sentences and long trains of thought”. This internally ingrained sanity-check system is linked to our genome, which has evolved to encode information about proper interactions with our natural ecosystem. As long as this link remains intact, we retain our sanity and can endure reality checks. Contact with nature—in all its multidimensional richness—is crucial for human cognition, especially for our long-term thinking to make sense. Recognizing the central role of natural ecosystems in climate is just one way to begin rethinking and repairing our painful detachment from the natural world.