problem in calculation of gdd_m5_dormance

[nicolasviovy]

I found a problem that could be a critical bug for phenology. The gdd_m5_dormance used to calculate begin leaves is calculated in a way which seems wrong:
in stomate_season:

WHERE ( ( t2m_daily(:) .GT. ( ZeroCelsius? - gdd_threshold ) ) .AND. &

( gdd_m5_dormance(:,j) .NE. undef ) )

gdd_m5_dormance(:,j) = gdd_m5_dormance(:,j) + &

dt * ( t2m_daily(:) - ( ZeroCelsius? - gdd_threshold ) )

ENDWHERE

gdd_threshold been set to 5°C
So the minus sign is weird here as gdd is normally calculated with a threshold above ZeroCelsius? ? the consequence is accumulation of gdd even for temperature lower than 0 which can explain the too early begin leave we observe....
The problem seems to be related to a confusion with the NGD model used for larix.. In this model for very cold condition the model use an number of days above -5°C which lead to calculation of ngd_minus5(:,j)

WHERE ( t2m_daily(:) .GT. (ZeroCelsius? - gdd_threshold) )

ngd_minus5(:,j) = ngd_minus5(:,j) + dt

ENDWHERE

which use the same gdd_threshold ... but in this case used with a minus sign which give the good result (but in fact another variable should be used as there is no reason to fix the same threshold for both model and for this case too we should fix a value of -5 and have a plus sign)....
So in principle the code should be modified with:

WHERE ( ( t2m_daily(:) .GT. ( ZeroCelsius? + gdd_threshold ) ) .AND. &

( gdd_m5_dormance(:,j) .NE. undef ) )

gdd_m5_dormance(:,j) = gdd_m5_dormance(:,j) + &

dt * ( t2m_daily(:) - ( ZeroCelsius? + gdd_threshold ) )

ENDWHERE

and have:

WHERE ( t2m_daily(:) .GT. (ZeroCelsius? + ngd_threshold) )

ngd_minus5(:,j) = ngd_minus5(:,j) + dt

ENDWHERE

with a new constant ngd_threshold fixes to -5

[luyssaert]

The proposed changes have been implemented in r7928. In the same revision, GDD_threshold was externalized and made PFT-specific.