search

lma6 / GlobalED

0 stars 0 forks source link

Modify photosynthesis module #1

Open lma6 opened 6 years ago

lma6 commented 6 years ago

Modify the whole Farquhar model and stomatal conductance modole, new module is in photosynthesis.c file. To activate this module, the switch COUPLE_FAR and FASTLOAD should be on. Parameters in these module is from Kattge et al 2007 Page 1184

lma6 commented 6 years ago

2

lma6 commented 6 years ago

3

lma6 commented 6 years ago

4

lma6 commented 6 years ago

5

lma6 commented 6 years ago

Test new CRUNCEP climate input Issue: The original climate input (path is listed below) seems wrong in air temperature variables which fluctuate month by month and have no seasonality. This comment is to test the improvement of new processed CRUNCEP on ED with modified and original Farquhar module.

As ED with and without modified Farquhar model have different optimal PFT vcmax. The PFT.cfg file of each case are attached. modifiedFarquhar_ED_pft.txt originalFarquhar_ED_pft.txt

The old and new CRUNCEP input datasets are stored in: old: /gpfs/data1/hurttgp/gel1/leima/AssignTask/gED/Data/Climate/CRUNCEP/v0.9/ /gpfs/data1/hurttgp/gel1/leima/AssignTask/gED/Data/premech/MSTMIP/CRUNCEP/v0.9/

new: /gpfs/data1/hurttgp/gel1/leima/AssignTask/gED/Data/Climate/CRUNCEP/new/ /gpfs/data1/hurttgp/gel1/leima/AssignTask/gED/Data/premech/MSTMIP/CRUNCEP_new/

Vcmax values for modified and original ED are : Modified version: c4g-8, c3g-40, es-15, ms-13, ls-12, cd-13, erg-22 Original version: c4g-15, c3g-15, es-12, ms-12, ls-7, cd-7, erg-7

Farquhar parameterization is based on Kattge et al 2007 including general average (when COUPLE_Accm is off) and with temperature acclimation (When COUPLE_Accm is on). (This may change in the next comments). And LfWidth=0.05

ED with modified Farquhar and forced by new CRUNCEP is fad9bc1837ef4ace9ae16ba87242f095c6c807e2 & b74babf733d4faf360bdf2c6a2b15e25d440100f. The Modified ED results are stored in _GlobalPot1k_FAST_FARCPwoAccm3 & _GlobalPot1k_FAST_FARCP_woAccm3LU.

ED with original Farquhar and forced by new CRUNCEP is 1664a5a21b6424dd2f5bee191edfed358b05a5b4 & e35213cd01e7a3d7476cf2273c320f7c48e0da08. And results are store in _GlobalPot1k_FASTFARCP & _GlobalPot1k_FAST_FARCPLU.

ED with original Farquhar and forced by old CRUNCEP : the results are stored in _GlobalPot1kFAST & _GlobalPot1kLUFAST

All premech variables inc. shortwave radiation, air temperature and specific humidity in the above new CRUNCEP inputs are linear interpolated, which means there is still radiation at night and weaker radiation at noon.

Test results and conclusions: The temperature and air specific humidity variables in MERRA2 are generally lower than CRUNCEP dataset, resulting in reducing half GPP, NPP. These also lead to weak carbon sink during summer period at global. To mitigate this issue, new interpolation method was proposed to produce hourly shortwave radiation, which considers the solar zenith angle changes and location, the new interpolated premech files are named with v2.

For ED with new climate and premech input, the seasonal and latitudinal pattern of carbon fluxes including GPP, RECO are pretty good. Only one issue so far left is low GPP in grassland and cropland, which may due to lower vcmax in c3g and c4g.

lma6 commented 6 years ago

Test cosine-based interpolation of shortwave radiation

Issue: In the last comments, new processed CRUNCEP is used to drive ED and also compared to results from MERRA2. There is a big decrease in vegetation productivity in MERRA2 results, may due to lower temperature and air specific humidity, or simple interpolation of shortwave radiation.

(Feb 23th 2018)Explanation of the above issue. Through comparison, diurnal shortwave radiation based on cosine interpolation move radiation at night to noon, and the Farquhar model based on Kattge et 2007 (General average without acclimation scheme) has low An rate at noon (This is correct, as at noon, VPD increases and results in stomata closed and decreasing Ci). As a result, the new interpolation only has half An rate as before. In addition, a drop in An rate appears at noon for tropical forests because of high temperature. Even the air temperature is not too high, as wind speed is only set as 0.8, resulting much higher temperature in leaf. This may also indicate the general average parameterization scheme from Kattge et 2006 may not accurate for tropical as they are derived from growth temperature at 18 degrees, and the need to consider the parameterization from Bonan et al 2011.

Another factor strongly control midday(ie. noon) depression of An is boundary layer conductance which varies with wind speed. When it increases, the leaf surface humidity decreases as less water vapor is retained around leaf surface, resulting in decreased stomatal conductance, then further decrease net carbon assimilation rate. When it decrease to a very low level (e.g 0.2 mol m-2 s-1), the leaf humidity is no longer a limiting factor, but it reduces the heat exchange between leaf and ambient air, resulting in increased leaf temperature which may exceed optimum of photosynthesis. This mechanism also yield noon depression. (Colatz et al 1991, P122)

Solution: 1)In this comments, the SWD has been reprocessed using solar zenith angle function which varies with latitude, longitude and time. The new process method could be activated if set interp_option as "cosine" in CRUNCEP_premech.py. The updated premech files are named with v2, like CRUNCEP_premech_v2_1901.nc

2)The parameterization in the last comment is changed. When COUPLE_Accm is on, parameters under temperature acclimation from Kattge et al 2007 and Akin et al 2008 are used, and the growth temperature is calculated using current monthly air mean temperature, which is different from Akin et al 2008 (i.e. 10 days running temperature prior days). Through comparison, this difference is small and acceptable. When COUPLE_Accm is off, parameterization from Bonan et al 2011 is used which is static for all PFTs and all latitudes. This parameterization is same as the example from Danica L Lombardozzi et al 2015 Geophysical Research Letters

Test

Using the original Farquhar model from Moorecroft and Vcmax Vcmax: c4g-15, c3g-15, es-12, ms-12, ls-7, cd-7, erg-7 1st try: c4g-15, c3g-15, es-12, ms-12, ls-7, cd-7, erg-7, The results are saved in GlobalPot1k_oldFar_newCRU1. Too lower for tropics. These original values fail to produce Amazon forests due to decreases in new interpolated SWD.

2nd try: c4g-15, c3g-25, es-25, ms-25, ls-20, cd-15, erg-15, The results are saved in GlobalPot1k_oldFar_newCRU2.

3rd try: c4g-25, c3g-25, es-25, ms-25, ls-20, cd-13, erg-12, The results are saved in GlobalPot1k_oldFar_newCRU3 && GlobalPot1k_oldFar_newCRU3_LU. Vcmac for c4g is too high and as well as cd and erg.

4th try:c4g-20, c3g-25, es-25, ms-25, ls-20, cd-9, erg-9. The results are saved in GlobalPot1k_oldFar_newCRU4 && GlobalPot1k_oldFar_newCRU4_LU. Seems not too much difference to 3rd try.

Using Bonan's Parameterization As the original Vcmax value in the last comment experiments is too lower for new interpolation input for trees' survival, the following several rounds of tests are to obtain appropriate settings. Wind speed is 4 and Ca=350, LfWidth=0.05

1st try: c3g-80, c4g-16, es-45, ms-45, ls-40, cd-20, eg-40, outputs are save in GlobalPot1k_FAST_FARCP_woAccm5. ls is near two times higher, cd is 1.5 times higher

2nd try:c3g-80, c4g-16, es-35, ms-35, ls-29, cd-20, eg-45, outputs are save in GlobalPot1k_FAST_FARCP_woAccm6.

3rd try:c3g-80, c4g-16, es-30, ms-30, ls-20, cd-60, eg-45, outputs are save in GlobalPot1k_FAST_FARCP_woAccm7 & GlobalPot1k_woAccm7_MERRA2_LU. The Vcmax of cd is two times higher

4th try:c3g-80, c4g-16, es-30, ms-30, ls-20, cd-30, eg-40, outputs are save in GlobalPot1k_FAST_FARCP_woAccm8 && GlobalPot1k_woAccm8_MERRA2_LU. The Vcmax of cd and eg may be still 5 higher.

5th try: c3g-80, c4g-16, es-28, ms-28, ls-19, cd-20, eg-35. outputs are save in GlobalPot1k_FAST_FARCP_woAccm9. Still too high for tropical region.

In these five attempts, still fail to optimize Vcmax and parameters in 5th try is the best so far. However, it still has problems that like GPP peak in June instead of July in some regions.

Using Kattge 2007 parameterization without acclimation 1st try: c3g-50, c4g-8, es-20, ms-19, ls-13, cd-20, eg-35. outputs are saved in GlobalPot1k_KattgewoAccm1. When Jmax/Vcmax ration of evergreen was changed back to 1.97 as same as others, 35 is too higher for it. Need another run.

2nd try:c3g-60. c4g-10, es-30, ms-28, ls-20, cd-20 ,eg-22, outputs are saved in GlobalPot1k_KattgewoAccm2 & GlobalPot1k_KattgewoAccm2_LU. The followings exp is a comment titled "Divergent NEE seasonal pattern".

lma6 commented 6 years ago

Tuning optimal Vcmax value for c3g and c4g.

results are save in .

lma6 commented 6 years ago

Divergent NEE seasonal pattern

Issue: The results based on Kattge 2007 (GlobalPot1k_KattgewoAccm2_LU) shows a significant difference between CT2016 and ED on seasonal pattern of NEE, GPP and RECO at both global, NH, SH and tropics. The divergence at global is mostly contributed by NH in where abiotic factors govern the seasonality.

Northern hemisphere 1st guess: weak GPP and a little bit stronger RECO from March to September yields lower NEE. Maybe soil respiration is wrong and Vcmax for c3g rather than cold-deciduous and evergreen. In Europe and middle U.S. large area of pasture/rangeland and cropland appear after land use change and only the c3g is allowed to grow. The GPP difference map shows red colors in these regions indicating the c3g should have higher GPP than current results. The next try is to increase Vcmax of c3g from 60 to 80. This try may probably cause problems in other regions where is fine in current version because of the tree-grass competition in one single soil layer. Meanwhile, Vcmax for cold-deciduous and evergreen seems a little higher, may need to be reduced (20-16) and (22-18).**

1st try:c3g-90. c4g-15, es-29, ms-27, ls-20, cd-18 ,eg-20, outputs are saved in GlobalPot1k_KattgewoAccm3 && GlobalPot1k_KattgewoAccm3_LU && GlobalPot1k_KattgewoAccm3_SpTS_LU. GPP in cropland is more closed to reference, but more grass appears resulting in very high GPP in some regions. In this exp, respiration rate of roots is calculated using soil temperature instead of air temperature, but a little improvement. The reason is the most loss of GPP is for growth respiration rather than root respiration (r.g. GPP is 120, 60 is growth respiration, only 6 is root respiration). Due to very small share of root respiration, to separate tf for air and soil may not help a lot. This exp also test the 2nd guess. Besides, the parameters in this exp have relatively worser overall performance than 2nd try in the ['Using Kattge 2007 parameterization without acclimation']. Current ED version is 2c76c0b20e1ada733e63811ae851609237bdd63c.

Change spin-up strategy by cyclically MERRA2 1981-2015 climate rather than using CRUNCEP 1901-1920 avg until 1981 and using MERRA2 from 1981.

1st try: c3g-90. c4g-15, es-50, ms-50, ls-40, cd-45 ,eg-60, outputs are saved in Global_MERRA2_spinup1 & Global_MERRA2_spinup_LU1.

2nd try: c3g-90. c4g-15, es-32, ms-32, ls-25, cd-19 ,eg-22, outputs are saved in Global_MERRA2_spinup2 & Global_MERRA2_spinup_LU2.

3rd try:c3g-90. c4g-15, es-32, ms-32, ls-25, cd-22 ,eg-25, outputs are saved in . In this try, the output timelag is removed.

2nd guess: RECO is strong maybe due to ED currently use air temperature instead of soil temperature to calculate tf which is same for sapwood, root and virtual leaves instead of using soil temperature separately for root. using soil temperature when available to separate tf for root respiration. In 1st try exp (GlobalPot1k_KattgewoAccm3_LU), respiration rate of roots is calculated using soil temperature instead of air temperature, but a little improvement. The reason is the most loss of GPP is for growth respiration rather than root respiration (r.g. GPP is 120, 60 is growth respiration, only 6 is root respiration). Due to very small share of root respiration, to separate tf for air and soil may not help a lot. This exp also test the 2nd guess.

3rd guess, GPP peak should appear in July rather than June. Current Farquhar parameterization based on Kattge 2007 shows maximum GPP in June when temperature and VPD are not too high. After it, temperature and VPD increases and result in more closed stomata and further decrease co2 concentration of leaves. Need to figure out why GPP peak is advanced.

Tropics The results based on Kattge 2007 shows tropics have reverse seasonality which has very carbon sink season.

1st guess (Mar 6th) is temperature dependence of Farquhar model. Maybe during the wet season, the increased air temperature drive leaves out of optimum of photosynthesis.

2nd guess (Mar 6th) is water stress from the dry seasons is inaccurately estimated in ED which has only one layer and roots could not reach the water stored in deep soil layers. A good example solving this issue in SiB3 model could be found in Baker et al.,2008.