Amazon/Africa precip biases

[islasimpson]

We can gather information that we have here on the Amazon and West Africa precip biases. The too-dry amazon and too-wet West Africa is still a problem in 122. It is there in both the piControl and in the satellite era period of our coupled runs. It isn't there in AMIP mode so it's either associated with SST biases or needs coupled interactions to manifest. Below is a figure showing precipitation in JJA.

Here is the temperature bias in our 121 historical simulation for 1979-2012.

We're too warm in the NH, too cold in the SH. A guess would be that this is pulling precipitation too far North over Africa and making it absent over the Amazon.

CESM2 did have a dry bias over the Amazon but it wasn't as extensive. Neither was the wet bias over West Africa

Here are the JJA temperature biases that we had in CESM2.

The next figure is using the 2000 years from the CESM2 piControl and taking an index of interannual variability in the difference in precipitation between west africa and the amazon, given by the average over the boxes. It shows that in interannual variabilty when you have more precip over west Africa relative to the Amazon, it's warmer in the North tropical Atlantic and cooler in the South.

The next shows the same but for 20 year running means. On these timescales variability in the sub-polar North Atlantic dominates.

So, from the natural variability, we know that when the North Atlantic is warmer, precipitation is higher over West Africa and lower over the Amazon.

We maybe should think about what we can do to try and reduce these biases in the inter-hemispheric temperature gradient.

[islasimpson]

Just summarizing where we were at with the Amazon/Africa precip biases before being derailed by lab sea freezes. This plot is a summary of the precipitation biases relative to GPCP in the boxes on the right. The colored bars that are centered on zero show the range of differences that can occur within segments of the piControl from CESM2 to give a sense of the influence of sample size on internal variability. We are more biased in both these regins in the annual mean than we were in CESM2. JJA is the season of the largest bias but we also had a bias in this season in CESM2 of a comparable magnitude. Where we seem to have got worse compared to CESM2 is in DJF and MAM and maybe SON although to a lesser extent.