Table 4 Mean absolute error across global grid for HASDM-ML as a function of space weather conditions.

From: Uncertainty quantification techniques for data-driven space weather modeling: thermospheric density application

 

\({F_{10} \le 75}\)

\({75 < F_{10} \le 150}\)

\({150 < F_{10} \le 190}\)

\({F_{10} > 190}\)

All \({ {F}_{ {10}}}\)

MC dropout

\({a_p \le 10}\)

8.96%

9.78%

9.97%

9.14%

9.50%

\({10 <a_p \le 50}\)

9.76%

10.05%

10.87%

9.90%

10.09%

\({a_p > 50}\)

15.35%

12.86%

13.23%

12.55%

13.01%

All \({a_p}\)

9.12%

9.92%

10.36%

9.55%

9.71%

Direct probability

\({a_p \le 10}\)

8.64%

9.33%

9.35%

9.11%

9.10%

\({10 <a_p \le 50}\)

9.18%

9.51%

9.69%

9.64%

9.48%

\({a_p > 50}\)

11.14%

11.23%

11.34%

10.30%

11.11%

All \({a_p}\)

8.74%

9.42%

9.52%

9.34%

9.23%

Back to article page