Stanford Earth Undergraduate Research Symposium
Summer research projects from the 2020 SESUR cohort
Terachet Rojrachsombat, Salvatore Pascale, Morgan E. O’Neill
Past comprehensive climate models exhibit large biases in their simulation of the North American Monsoon (NAM), a major contributor to annual rainfall in a vast area of the North American Southwest (Geil et al., 2013; Liang et al., 2008). As such, future projections of the NAM remain uncertain, limiting future adaptation planning needed to manage water resources. This study analyzes the model depictions of NAM precipitation in the core monsoon domain (25-29ºN 105-109ºW) and the sea surface temperature (SST) in adjacent oceans, using the recently released Coupled Model Intercomparison Project Phase 6 (CMIP6). Compared to satellite and reanalysis data, coupled CMIP6 models tend to overestimate both the NAM precipitation and the SST off the coast of Mexico. However, uncoupled CMIP6 models, which use observed SST instead of model-run SST, depict NAM precipitation more realistically. Therefore, SST biases reduce the accuracy of NAM precipitation depictions. However, in-depth investigations into the relationship between local SST biases and NAM precipitation yields no statistically significant results yet. Spatial resolution also influences model accuracy, but the effect is small compared to SST biases. When global warming is replicated using five different model experiments, in four of them, most climate models predict a decrease in NAM precipitation. The results are consistent with Cook & Seager, 2013, Pascale et al., 2017, and Pascale et al., 2019. This future decrease in precipitation also correlates with model precipitation biases. The more biased-dry the model, the greater the precipitation decrease due to global warming, which is consistent with Bukovsky et al., 2015. Thus, when studying the NAM, especially when dealing with global warming, one must use models that are low in precipitation bias, low in SST bias, and high resolution.
Great job, Drive! I enjoyed hearing about your summer project. That was an interesting result about the connection between SST and the North American Monsoon strength, and a very clear presentation!
Thank you! I really appreciate that
Drive
Hi Drive,
Great presentation! You did a really nice job explaining your concepts and walking us through your research.
Since your research shows that uncoupled models might be better at predicting than coupled models, I was wondering how you can address these types of SST biases in reality? It makes sense that using actual SST, over model predicted SST, would yield more accurate results, yet I can also see many situations where actual SST may not be available in certain places/time periods. Are there other ways that we can help correct for these biases in the data without specific space-time SST data?
I also liked how you touched upon some of the ramifications of NAM with climate change, and how it may be weakening. What types of impacts might less rainfall have on people in this area (for example, might areas no longer be suitable for food crops, increase in major storm events, etc.)?
Hi Bianca,
Great question! SST data only starts from 1979 when satellite data begins, so we are limited to the 40 or so years of information to inform out climate models.
Still, I haven’t gone into climate modeling as a field yet. I just use the results of the model to study stuff. So, I have absolutely no idea how to really fix these SST biases. To answer that question, I might have to study the parameters fed into the climate models or the underlying code, which I am very unqualified to do.
As for the weakening NAM, I can only say for sure that, from the literature and from my research, the precipitation would be decreasing due to global warming. My coupled models says that rainfall in western Mexico will decrease by 10%; my uncoupled models says 25%. This means less water supply for crops and other usage, which would have bad ramifications for political and economic stability for the nation. But, these are results for year 2070-2100. So, engineers and policymakers still have time to respond.
Cheers,
Drive
Great presentation! Do you think a hybrid of the coupled and uncoupled models would work, as well?
Hi Shridhar,
The oceanic realm and the atmospheric realm can either be connected or not connected. So, there are no hybrids by definition. BUT! CMIP6 has hundreds upon hundreds of experiments I can choose from with different boundary conditions and settings. I played with like 10 experiments or so. For example, “AMIP future 4K” is an uncoupled model experiment with more realistic ocean warming patterns due to global warming. Many insights can definitely be gained from studying the experiment closer.
But, since I want to make SST biases the independent variable, I compared the most basic of the coupled and uncoupled models.
Cheers,
Drive
Well done Drive, a productive summer and really interesting and important project! Excellent presentation too, very easy to follow. It’s too bad that the coupled models aren’t as accurate, because we expect global warming to be a highly coupled ocean/atmosphere phenomenon. In your research and reading for this project, what did you find as possible explanations for why the coupled models aren’t as good? Is it just a matter of resolution? Or is the spin-up time for these simulations just too long for more accurate results? Really nice work!
Thank you Morgan!
I indeed investigate model resolution effect on accuracy. Accuracy definitely play a role since most models do not have a high-enough resolution to resolve the Gulf of California, which is an important source of moisture for the monsoon. Since the models will treat the Gulf as land, the precipitation response is sure to be inaccurate.
I categorize the models into two groups, those with grids around 100 km in length and 250 km in length. When I use the uncoupled models (to make SST a controlled variable) the higher resolution models (100 km) indeed are more accurate than the lower resolution models (250 km). The effect is small but measurable.
However, when I use coupled models, the effect of resolution is completely drowned out by the SST biases. The error bars are just too big for every comparison. This means SST biases play a key role in determining the accuracy of the precipitation field. And, CMIP6 models’ resolution is still not high enough to counteract SST biases.
From my investigation, I really suspect the overestimation of SST in the Gulf of Mexico as the culprit of this inaccuracy. All models thinks the Gulf of Mexico is hotter than it really is; some models have biases of over 7 degrees celsius (Most sea life would be dead if those models are correct). Sadly, almost all my scatterplot relating SST biases in the Gulf of Mexico and measures of model accuracy are inconclusive so far. So, I did not conclude anything regarding that for this video.
As for global warming, I agree that it is too bad that coupled models might not be accurate. To make this worse, the many effects of climate change compete with each other as well. Higher CO2 concentrations seems to enhance the hydrologic cycle which means a stronger monsoon. Higher SST due to ocean warming, however, seems to weaken the monsoon due to less ocean-land temperature gradient. My research suggest that the effects of SST is more important that of CO2. Still, since the two might cancel each other out, it unfortunately adds to the uncertainty of global warming predictions.
Despite all the inaccuracies, there are some agreements among the models. For example, in the appropriate experiment, most coupled models still agree that the monsoon precipitation will decrease due to global warming. So, the climate models are still trustworthy to some extent.
I’m not sure what you mean spin-up time for these simulations. But yes, AMIP experiments only starts from 1979. I wish there are more years so that any trend lines would be more prominent.
As for this project, I wish I took the initiative to collaborate with you more. Maybe the next project, if I would like to study storms or things of that matter, I would correspond with you often.
Thank you,
Drive
Hi Drive!
I really enjoyed your presentation. I thought your visuals were really great and the way you explained everything was straightforward. I was going to ask where the SST biases come from/why coupled models aren’t as good, but I see that you already answered others. I loved the last graph you showed about potential effects of global warming on NAM. I can imagine how devastating that might be for the many countries in the monsoon’s path reliant upon agriculture.
Great job!