Check interception loss

[luyssaert]

In r7274 the interception loss is about 3 times lower than in ORC2 and ORC3 despite having more or less the same LAI. r7274 uses Pgap which might be responsible for simulating a smaller leaf area for rain interception. A smaller intercepting leaf area is in line with the larger bare soil evaporation of r7274 compared to ORC2 and ORC3. Before accepting this difference as a consequence of Pgap, the parameter values and the calculations should be checked to confirm that no accidental changes were made.

[luyssaert]

Prescribing veget and lai
Even when prescribing veget_max, ORC3 and ORC4 use slightly different values. Neither model uses exactly the same values. ORC3 corrects the veget_max of all PFTs. ORC4 puts the residual in PFT1.

When forcing lai, veget and qsintcst, ORC3 and ORC4 generate almost identical results. The difference comes from the small differences in veget_max and veget (was set equal to veget_max in this test). Given the calculation is rather straightforward, this result is not a huge surprise, but it shows at least that the code was not changed by accident (Kim had already checked the relevant parts of the code and the parameter values).

Prescribing veget, calculating lai
ORC2 and ORC3 still have a prescribed maximum lai which is used to initialize qsintmax. ORC4 uses the actual lai (which is only calculated in stomate after qcintmax is initialized). This results in a large difference in qsintmax for the first day. Note the first day ORC3 has a qsintmax even for PFTs that don’t have leaves yet.

The second day ORC3 use the actual lai and all three models are better in line with each other than the first day. For the first month (length of the test), qsintmax in ORC4 is often higher then in ORC3.

When using the calculated lai instead of hacking the lai by fixing it, the interception of ORC4 was 0.015 kg m-2 and 0.009 kg m-2 for ORC3. This is exactly the opposite as what is shown by the mapper (for much longer time period and the global domain) and what triggered this investigation.

Not surprisingly, interception scales with lai. Sometimes ORC3 is higher, sometimes ORC4 is higher.

Calculating lai and veget
Veget is to a large extend determined by lai. A lower lai will also result in a lower veget.
In ORC4 veget can be calculated in two ways: Pgap and Lambert Beer (same as in ORC3). The method of how veget is calculated has an impact (interception is 10-15% higher with Lambert Beer) but interception is still too low: ORC3 0.19 kg m-2, ORC4-Pgap: 0.09 kg m-2 and ORC4-LB: 0.10 kg m-2. This suggests that the underlying cause is differences in LAI.

The global mean LAI between ORC3 and ORC4 is very similar, but the global maps shows ORC4 has more LAI in the tropics and less LAI outside of the tropics compared to ORC3. How it is possible to have the same mean LAI but only half the interception? The mapper does not reveal anything suspicious.

Initially the the mapper did not show the relationship between LAI and precipitation (too low LAIs in regions/seasons with a lot of precipitation?). An updated version of the mapper confirms this issue. This ticket was therefore considered a direct consequence of #698. Although no code changes were made, the ticket is no longer considered essential to finalize milestone 4.1.

As interception is included in the mapper, progress on this issue will be evaluated for future model versions following tuning and benchmarking.