April 2023
I’ll start off this blog with a simple and very obvious message: Our planet is warming due to humans. This a result of an increase in the emission of greenhouse gases from the burning of fossil fuels. Yet, the exact future evolution of global climate change and its consequences still remains uncertain, especially on regional scales. Decision makers and other community partners therefore are tasked with considering a wide range of possible future climate change pathways to address adaptation and mitigation planning (see Deser, 2020). One approach for evaluating future projections is by using data from global climate models. Unlike forecasting day-to-day changes in weather, climate models project the average of a variable over decadal to centennial and longer time-scales.
There are three key sources of uncertainty for assessing and communicating future climate change projections…
Of course, there are other uncertainties, such as climate sensitivity, the regional effective radiative forcing of anthropogenic aerosols, land cover/land use change, etc., but a discussion of those are better left for another day. 🙂
One source of uncertainty is due to the climate model itself; this is otherwise known as “model structural uncertainty.” In this particular visualization, we are using the SPEAR large ensemble (https://www.gfdl.noaa.gov/spear_large_ensembles/), which is based on GFDL’s CM4.0 physical climate model (https://www.gfdl.noaa.gov/coupled-physical-model-cm4/). Another form of uncertainty is associated with the level of future warming. This type of uncertainty is called “scenario uncertainty,” and it arises from the various socioeconomic and technological decisions that could evolve and consequently affect the amount of emitted greenhouse gases. In this visualization, we are evaluating a projection called SSP2-4.5 (i.e., moderate amount of future climate change emissions), which is a pathway closely aligned with our current trajectory.
However, for this month’s climate viz of the month, we are focused on the third big source of uncertainty, which is referred to as “internal climate variability.” You can think of internal climate variability as analogous to the butterfly effect, in which a tiny change in one place can lead to an unexpected change in another place. In other words, internal climate variability describes the chaotic (generally unpredictable) behavior of the atmosphere, which is an intrinsic property of the atmosphere in the absence of human-caused climate change. Recent studies have shown a better appreciation for considering the influence of internal climate variability, especially at regional spatial scales. Though it’s important to note that internal climate variability can even temporarily mask or accelerate short-term trends in the global mean surface temperature (see Medhaug et al. 2017). If you are interested in this effect, check out our new study on this effect for predicting slowdowns in the rate of decadal warming using machine learning: https://doi.org/10.1029/2022GL098173.
To better consider how internal variability impacts future climate change projections, international modeling centers often run experiments using the same climate model and future emissions scenario but vary the initial conditions of each simulation by a very very small round-off error. This type of experiment is called a “large ensemble,” where each ensemble member is run using the same climate model protocol but differ by a tiny change in those initial conditions. In large ensemble experiments, the spread in ensemble member outcomes represents the uncertainty due to internal climate variability, and the effect of human-caused climate change can then be revealed by averaging over all the ensemble members (i.e., getting rid of the noise). Check out a user guide on evaluating large ensembles in http://dx.doi.org/10.5065/h7c7-f961.
Internal climate variability is quite large in the Arctic, and for example, this is one reason that we have not observed a new record low in Arctic sea-ice extent since 2012. I’ve discussed this point in other monthly blogs – so now let’s turn to April’s visualization. We can see the differences between ensemble member #30 (i.e., one possible realization of our future) versus the average across ensemble members in GFDL SPEAR (i.e., the ensemble mean).
For those interested in the technical details, this large ensemble (NOAA GFDL SPEAR_MED) is fully-coupled (i.e., land, ocean, atmosphere, ice components), leverages CMIP6 forcing, and has an unusually high spatial resolution at approximately 50 km.
You can see the effect of internal climate variability on the left side, where there are much larger differences from year-to-year in the magnitude and location of the temperature anomalies. But on the right side, the long-term trend from only climate change results in a steady warming and little year-to-year variability (remember, by construction, the noise is averaged out in the ensemble mean).
Importantly, the left map is what we experience – a world with climate change and internal climate variability. If I showed an animation using ensemble member #29, you would see the same long-term trend, but different interannual variations. And if I showed an animation using ensemble member #28, you would again see the same long-term trend, but yet another realization of year-to-year variability. And so on…
I can’t stress enough how important it is to consider the influence of internal climate variability when thinking about climate change, especially for an area like the Arctic. In fact, I don’t think any future impact studies should be conducted without including this source of uncertainty. Though, maybe I am a tiny bit biased, because I think large ensemble are the coolest thing ever.
That’s all for now! Thanks for stopping by my blog! Feel free to reach out any time if you are interested in large ensembles and/or internal climate variability.
Otherwise, for the month of April 2023, it was fairly quiet in the Arctic. The largest absolute magnitude of temperature anomalies was found around Alaska, where unusually persistent cold weather occurred nearly all month long. Sea-ice extent and thickness levels are also pretty unremarkable, well compared to recent years anyways. But we are only at the very beginning of the melt season, so it’s far too early to make any conclusions or outlooks about this year’s September minimum. We will just have to wait and see. Stay tuned here!
March 2023
Hope you all are having a great start to spring – at least in the Northern Hemisphere anyways. My March ‘climate viz of the month’ is not the prettiest graph that I have shared, but I think it displays some cool information.
Each panel of this graphic shows the decadal temperature trends for bands of latitudes from the South Pole (90°S; y-axis minimum) to the North Pole (90°N; y-axis maximum) for four different seasonal periods. The temperature trends are calculated from 1990 to 2022. I chose this period, as it coincides with a greater acceleration of long-term warming compared to earlier in the 20th century. Note that this graphic uses data at a horizontal resolution of 1.9° of latitude by 2.5° of longitude (i.e., 96 by 144 gridpoints), and there is no scaling by latitudinal distance in the visualization of the y-axis (hint: Earth is a sphere). Using data from ERA5 reanalysis, I then calculate the zonal mean (averaging across all longitude points), which leaves the temperature data in dimensions of time by latitude (of which there are 96 latitude bands).
I am such a huge fan of zonal mean climate plots, as they are basically a short course on understanding climate dynamics. One area of my research has involved understanding linkages between Arctic sea ice and mid-latitude weather, and a favorite paper of mine that I first read in grad school is titled “The Role of Ocean–Atmosphere Coupling in the Zonal-Mean Atmospheric Response to Arctic Sea Ice Loss.” This paper is a great example of understanding various climate responses in a zonal mean sense. If you also like this style of visualization, check out some of my other zonal mean animations at https://zacklabe.com/arctic-climate-seasonality-and-variability/.
In summary, this visualization helps to show the evolution of seasonal temperature trends and reveals where the fastest rate of warming is occurring. Arctic amplification is extremely striking, particularly during the early winter period. This is due to that fact that sea ice begins refreezing by October and subsequently all the heat that was absorbed into the Arctic Ocean during summer is transferred back into the overlying atmosphere.
Nearly all latitude bands are warming during the four seasons, except for portions of the Southern Ocean and the South Pole during their austral winter. Remember that the Southern Ocean plays an important role in being a significant absorber of heat associated with anthropogenic climate change, and raw data observations are quite sparse in this region. Understanding temperature trends in the Antarctic also remains a very active area of research due to high uncertainties in different reanalysis and station-based observational datasets. In any case, in the annual mean, all latitude bands are warming during this time period.
Otherwise, March 2023 was fairly uneventful in the Arctic aside from some anomalous warmth in northern Greenland and stretching toward the Beaufort Sea. Both sea-ice extent and volume rankings for March were insignificant compared to other recent years. At this point in the year, though, we still don’t know how summer will unfold in the Arctic. There is very low sea-ice predictability right now, particularly related to forecasting the September sea ice minimum. As we get closer to June, we’ll have a much better idea for how much melt can be expected this year.
In other news, we have a new study out this month on using machine learning to understand regional climate change in a climate intervention experiment (called ARISE-SAI-1.5). Check it out at https://doi.org/10.1088/1748-9326/acc81a. I also wrote a blog which summarizes the study and a few of our other explainable artificial intelligence work for climate science at https://zacklabe.com/climate-signals-and-explainable-ai/. As usual, thanks for stopping by my blog, and feel free to reach out anytime if you have any questions on it.
February 2023
Happy start to the Arctic summer! Well, the official melt season anyways. According to the National Snow & Ice Data Center (NSIDC), the annual maximum Arctic sea-ice extent was set on 6 March 2023 at 14.62 million square kilometers (5.64 million square miles). This occurred about 6 days earlier than the 1981-2010 average. Since the maximum (read more here), Arctic sea ice has quickly declined due to recent weather conditions, and current levels are actually close to the lowest on record for the date. Yikes! While there is very low predictability (correlation) between the March sea ice maximum and the September sea ice minimum, it’s certainly a busy start to the melt season.
Anyways, more on that later. My next ‘climate viz of the month’ is an animation showing daily Arctic sea-ice concentration in February across the Atlantic side of the Arctic Ocean. This animation uses high-resolution data (approximately 3 kilometers) derived from an algorithm associated with the ASMR2 satellite instrument.
Long-term trends in winter are smaller than summer for sea ice in the Arctic. The entire Arctic Ocean is actually ice-covered in February, and therefore, the variability and long-term declines are found at the sea-ice edge. The Barents Sea has observed the largest declines in February, which are due to both atmospheric and oceanic processes. For this reason, I keep a close eye on weather conditions in the Barents Sea throughout much of the year.
The animation speed is intentionally sped up to highlight the large amount of daily variability in this part of the Arctic. A lot of this sea-ice variability is due to local weather conditions, especially as storms from the North Atlantic begin to move poleward. Although these storms usually start to weaken and fizzle out, they still are a source of warmth and moisture from the lower latitudes into the high Arctic. In addition to the warmth and precipitation, there is also plenty of wind and wave activity that can drive the movement of sea ice drift along the ice edge. This region is often referred to as the marginal ice zone. If that wasn’t enough, there’s also strong ocean currents to deal with in this region, which are an important source of heat transport/exchange at the boundary of the Arctic and Atlantic Oceans. Quantifying changes in the influx of deep ocean heat and salinity from the North Atlantic Current (e.g., Atlantification – warm and salty waters from the Atlantic to the Arctic) is a key research topic for understanding Arctic climate change. In other words, sea ice can melt from both above and below! All these processes and interactions result in a very turbulent region across the Greenland and Barents Seas, especially during winter.
This data visualization is particularly striking for highlighting the effects of a strong storm near Svalbard that coincided with southerly winds pushing sea ice well northward in the middle of winter. Watch closely as the ice pack moves back and forth over the course of a week, especially the opening near Franz Josef Land.
Notably, sea ice in the Barents Sea set a record for the lowest daily measurement in the entire month of February (since satellite records have been kept (from 1978/1979)) during this period. As a result of this southerly flow and the wide-open ocean waters (typically sea-ice covered), surface air temperatures were also more than 5°C above the 1981-2010 average around Svalbard for the monthly mean. This is concerning, but unsurprising, since long-term temperature trends near Svalbard are warming faster than anywhere else on Earth. It’s a striking regional example of Arctic amplification.
My advice to those monitoring climate change in the Arctic – keep a close eye on the Fram Strait and Barents Sea and always appreciate the importance of synoptic meteorology (i.e., the chaotic noise of our atmosphere).
Elsewhere, a strong dipole in temperature anomalies was evident in February. This included regions of anomalous warmth over western Siberia compared with brutally cold temperatures stretching from the Canadian Arctic to the Kamchatka Peninsula. Relative to recent months (and since 2021 to be honest), sea-ice extent was particularly low for its monthly ranking. With El Niño just around the corner (check out sea surface temperatures in the equatorial Pacific lately!!) after a rare triple La Niña, it will be ‘interesting’ to see how (quickly) the large-scale circulation responds in the Arctic and Northern Hemisphere. Stay tuned…
January 2023
Hi! My next ‘climate viz of the month’ evaluates changes in daily Arctic sea-ice thickness during 2022. While I normally share lots of sea-ice thickness visualizations on weekly and monthly timescales, it is also informative to compare its daily evolution. It may take a few seconds for this GIF to load (on the left). You can also click on it to see a higher resolution animation.
Specifically, this visualization uses daily sea-ice thickness data from an ice-ocean model called the Pan-Arctic Ice Ocean Modeling and Assimilation System (PIOMAS; Version 2.1). This reanalysis-like dataset extends from 1979 to present day, which is generated through the assimilation of daily surface fields including: 2-m temperature, 10-m winds, specific humidity, precipitation, evaporation, radiative fluxes, sea-ice concentration, and sea surface temperature. In a previous study (Labe et al. 2018), we evaluated PIOMAS with observations from satellites (CryoSat-2 and ICESat) and submarine track data to understand patterns of Arctic sea-ice thickness. In summary, we found that PIOMAS compares remarkably well in reproducing the spatial distribution, variability, and long-term trends of Arctic sea-ice!
As you watch this animation, note the prominent seasonality of sea-ice thickness. The maximum mean pan-Arctic thickness is set in late May, and the minimum is set in early November. The thickest ice is almost always found north of Greenland and the Canadian Arctic Archipelago. This area is often known as the Last Ice Area (e.g., the Wandel Sea), and the spatial distribution is broadly due to the transpolar drift stream. In other words, surface winds, ocean currents, and cold temperatures act to move and compress the thickest and oldest (multi-year) sea ice along this region before slowly drifting toward the Fram Strait and out into the Atlantic Ocean. Understanding variability and long-term trends of sea ice in the Last Ice Area is particularly important for assessing Arctic climate change. Meanwhile, sea ice is much thinner at the outer edges of the Arctic, like in the Barents-Kara Seas or Chukchi Sea.
Another important consideration for evaluating daily sea-ice thickness is the large spead in regional mean thickness. Ice thickness ranges from 1-2 meters (3-7 feet) in most regions throughout the year with seasonal ice cover, but the thickest ice can be more than 4 meters (13 feet) thick north of Greenland and the Canadian Arctic Archipelago. Though, it’s important to note that this model’s horizontal resolution (nominal 22 km) cannot resolve finer details, such as the actual thickness of ridging on ice floes. And of course, some areas have observed a decrease in ice thickness by more than 50% since 1979. As climate scientists improve model simulations of Arctic sea ice, it will become increasingly important to capture this spatial distribution of ice thickness to make projections of future Arctic climate change.
That’s all for now! Thank you for taking the time to visit my website! You can find the associated monthly climate data rankings for 2023 at https://zacklabe.com/archive-2023/. Although January 2023 observed anomalously low Arctic sea-ice extent due to wide open ocean waters in the Barents Sea region, temperatures were unusually cold near Siberia, and the total ice volume was unremarkable compared to some recent years.
My visualizations:
The views presented here only reflect my own. These figures may be freely distributed (with credit). Information about the data can be found on my references page and methods page.