Clouds and OLR

Estimating solar SW radiation from OLR or cloudiness

This set of figures compares two ways of estimating the net solar radiation at 0°, 170°W.
First, Reed (1977) found the empirical formula Q=Q0(1-0.62C+0.0019alpha), where C is total cloud cover and alpha is noon solar angle. (For C less than 0.3, Reed finds no reduction in insolation (Q = Q0)).
Second is a regression directly between OLR and SW: Q=0.9285(OLR)-0.6377. (This one was used in Fig 5 of the bottom set).

Some preliminaries:
1. Time series of OLR and ISCCP cloud fraction at 0°, 170°W.
   You want monthly-average OLR? We got monthly-average OLR
2. Time series of OLR and net insolation (Reed formula) at 0°, 170°W
3. Time series of SW at 0°, 170°W: Clear-sky and Reed formula (ISCCP)

   Direct comparisons between Reed formula and OLR regression, for monthly clouds and OLR during 1983-91 when ISCCP clouds were available. Overall, the agreement is good at the lower values (high cloudiness), but the Reed formula suggests that at low cloudiness the insolation is much greater than the OLR regression would indicate. The Reed formula gives a range of SW values about twice as large as the OLR regression.

4. Time series of net SW radiation at 0°, 170°W: Reed (ISCCP) and OLR regression
5. Net SW radiation at the equator: Reed (ISCCP) and OLR regression. Min/max during Jul 83-Jun 91

   Try a regression between OLR and ISCCP clouds.

   Figure 1 suggests that a regression could be found between OLR and cloud fraction that might improve the estimate of solar radiation. The regression from OLR to cloudiness could then be used in the Reed formula. This intuitively makes sense, since OLR is really closest to a measure of cloudiness (at least in the west where much of the cloud cover is associated with deep convection that is well-represented by OLR). Another advantage of this method (rather than regressing directly from OLR to radiation), is that Q0 in the Reed formula contains much of the annual cycle variability. Finally, variations of OLR above about 250 W/m2 or so are not meaningful in the context of radiation received at the sea surface. It would be desirable to be able to ignore those variations. Regressing directly OLR -> radiation means that these meaningless signals have as much significance as the low values. However, since the Reed formula ignores cloud variations below 0.3 sky cover, high values of OLR (that regress to low cloud cover values) are less of a factor.

   Clouds are available between July 1983 and June 1991. Use monthly-average OLR.
   Examples at longitudes across the Pacific (scatter diagrams and regression lines):

6. 130°E
7. 160°E
8. 170°W
9. 140°W
10. 110°W

    Some statistics. The correlation is good (>0.8) west of about 140°W, and excellent (>0.9) west of 170°W. East of 140°W the relation is unusable. In the east a lot of cloudiness is stratus, which does not produce an OLR signature. Note the increase in cloud variance, particularly east of 120°W, that is not seen in OLR.

11. Standard deviation of monthly clouds and OLR
    
12. Regression slope and correlation
13. Basic stats of OLR

    Cloud time series estimated by regression:

14. 130°E
15. 160°E
16. 170°W
17. 140°W
18. 110°W

    Regression over the entire tropical Pacific

19. Correlation, slope and intercept based on total ISCCP clouds
20. Correlation, slope and intercept based on ISCCP high clouds only
21. Average real and regressed clouds
22. Standard deviation of real and regressed clouds