AtmosArena Performance Leaderboard ⛅

To submit a model to the AtmosArena Leaderboard, please contact atmosarena@gmail.com 🌞

S2S Forecasting
Weeks 3-4
Weeks 5-6
Model Z500 T850 T2m Q700
RMSE ↓ ACC ↑ Spec ↓ RMSE ↓ ACC ↑ Spec ↓ RMSE ↓ ACC ↑ Spec ↓ RMSE ↓ ACC ↑ Spec ↓
ClimaX frozen 458.53 0.84 0 1.79 0.92 0.3153 1.67 0.96 0.1671 0.69 0.86 0.0789
ClimaX finetuned 453.05 0.84 0 1.77 0.92 0.3224 1.65 0.95 0.2298 0.69 0.86 0.0930
Stormer frozen 461.19 0.78 0 1.77 0.88 0.3307 1.56 0.95 0.4705 0.70 0.81 0.5188
Stormer finetuned 466.82 0.77 0 1.79 0.87 0.3275 1.64 0.94 0.6603 0.71 0.82 0.4337
Unet 498.46 0.84 0 1.90 0.92 0.3863 1.63 0.97 0.2065 0.74 0.85 0.0809
Climatology 475.58 - - 2.00 - - 1.61 - - 0.76 - -
Model Z500 T850 T2m Q700
RMSE ↓ ACC ↑ Spec ↓ RMSE ↓ ACC ↑ Spec ↓ RMSE ↓ ACC ↑ Spec ↓ RMSE ↓ ACC ↑ Spec ↓
ClimaX frozen 471.58 0.81 0 1.84 0.90 0.2894 1.73 0.95 0.1805 0.70 0.84 0.0903
ClimaX finetuned 469.92 0.81 0 1.80 0.90 0.3180 1.70 0.94 0.2093 0.71 0.84 0.0937
Stormer frozen 467.37 0.77 0 1.81 0.87 0.4161 1.69 0.94 0.5971 0.72 0.81 0.7513
Stormer finetuned 475.06 0.77 0 1.84 0.87 0.3024 1.75 0.93 0.6105 0.72 0.82 0.3468
Unet 521.32 0.84 0 2.09 0.91 0.5110 2.29 0.93 0.4647 0.75 0.85 0.8157
Climatology 475.58 - - 2.00 - - 1.61 - - 0.76 - -
Downscaling Performance
Model Z500 T850 T2m Q700 U10m V10m
RMSE ↓ Bias ↓ RMSE ↓ Bias ↓ RMSE ↓ Bias ↓ RMSE ↓ Bias ↓ RMSE ↓ Bias ↓ RMSE ↓ Bias ↓
Stormer finetuned 38.84 0.090 0.57 0.051 0.62 0.031 0.55 0.001 0.64 0.011 0.64 0.017
Unet 47.65 8.790 0.66 0.140 0.73 0.040 0.56 0.005 0.70 0.011 0.70 0.006
ClimaX finetuned 74.62 13.830 0.78 0.153 0.94 0.119 0.61 0.002 0.83 0.007 0.83 0.001
ClimaX frozen 105.49 28.660 0.93 0.167 1.16 0.054 0.70 0.001 1.02 0.032 1.01 0.009
Stormer frozen 104.26 17.540 0.95 0.046 1.12 0.048 0.76 0.001 1.07 0.019 1.05 0.011
Climate Model Emulation
Surface Air Temperature
Diurnal Temperature Range
Precipitation
90th Percentile Precipitation
Model NRMSE_s ↓ NRMSE_g ↓ NRMSE_t ↓
ClimaX frozen 0.085 0.043 0.297
ClimaX finetuned 0.086 0.043 0.300
Stormer frozen 0.117 0.043 0.334
Stormer finetuned 0.126 0.047 0.361
ClimateBench-NN 0.123 0.080 0.524
Model NRMSE_s ↓ NRMSE_g ↓ NRMSE_t ↓
ClimaX frozen 6.688 0.810 10.739
ClimaX finetuned 7.148 0.961 11.952
Stormer frozen 9.123 0.980 14.022
Stormer finetuned 8.598 0.834 12.767
ClimateBench-NN 7.465 1.233 13.632
Model NRMSE_s ↓ NRMSE_g ↓ NRMSE_t ↓
ClimaX frozen 2.193 0.183 3.110
ClimaX finetuned 2.360 0.206 3.390
Stormer frozen 6.159 0.210 7.211
Stormer finetuned 6.180 0.391 8.136
ClimateBench-NN 2.349 0.151 3.104
Model NRMSE_s ↓ NRMSE_g ↓ NRMSE_t ↓
ClimaX frozen 2.681 0.342 4.389
ClimaX finetuned 2.739 0.332 4.397
Stormer frozen 6.773 0.296 8.254
Stormer finetuned 6.797 0.316 8.376
ClimateBench-NN 3.108 0.282 4.517
Event Detection
Model TC Detection (Specificity ↑) AR Detection (Specificity ↑)
ClimaX frozen 0.99 0.96
ClimaX finetuned 0.99 0.96
Stormer frozen 0.98 0.95
Stormer finetuned 0.98 0.95
CGNet 0.99 0.92
Near-Surface Pollutants Downscaling (MAE)
Model NO2 ↓ SO2 ↓ CO ↓ O3 ↓ PM2.5 ↓
ClimaX finetuned 0.069 0.049 0.405 0.0065 0.100
Unet 0.064 0.047 - 0.0071 0.104