Changes between Version 16 and Version 17 of DevelopmentActivities/AccelerationSpinup

Timestamp:

2012-01-13T10:42:39+01:00 (12 years ago)

Author:

dsolyga

Comment:

--

DevelopmentActivities/AccelerationSpinup

@@ -278 +278 @@
 '''Results :'''
 
-We launched ORCHIDEE for 250Y on the same point described above then we restart and solve the system for 5 years. We obtained the same
+I launched ORCHIDEE for 250Y on the same point described above then I restart and solve the system for 5 years. We obtained the same
 results as ORCHIDEE in 7000Y (we tested the PFT 6 also)! [[BR]]
 
@@ -285 +285 @@
  * when the starting condition is OK, we solve the system until the stopping condition is reached (that will stop the job)
 
+== WORK DECEMBER 2011/JANUARY 2012 ==
+
+After the meeting of September 2011, we agree that is unuseful to start the algebraic method without the equilibrium of the biomass (litter inputs).[[BR]]
+
+For testing this point, I make a 250Y run without the method. Then I solve for 5Y after restarting this job. [[BR]]
+[[BR]]
+For the PFT 10 and the PFT 6, I obtain good results (0.05% relative error to the referenced values - 7000Y ORCHIDEE run). [[BR]]
+We divide the simulation lenght by 6 for the grasses 
+and by 12 to have the same relative error (see above)! [[BR]]
+
+It proves well that the analytic spinup is very sensitive to the biomass and its equilibrium.
+
+
+=== 1) First implementation : ===
+
+In order to optimize the time for the equilibrium of the biomass, we look at 3 criterions (2 fluxes and one stock):
+ * turnover_daily_accu : the cumulated sum of the turnover daily over the forcing period 
+ * bm_to_litter_accu : the cumulated sum of bm_to_litter over the forcing period
+ * biomass : the stock of biomass
+
+If the relatives errors for the three previous criterions between two forcing period is less than a given threshold (in %) simultaneously, we start the
+algebraic method.[[BR]]
+ The error is calculated by taking the maximum relative error over all the biomass compartment. To simlify the survey, we set the same
+threshold for the three criterions.[[BR]]
+
+If this condition cannot be reached, we force after the resolution after a prescribed length defined by the user (typically
+the estimated time for the biomass to be at equilibrium). [[BR]]
+
+
+Similarly, the carbon equilibrium is not reached until a threshold defined by the user is reached. We consider the maximum relative error over the carbon pools.[[BR]]
+
+The simulation finishes when the equilibrium is reached or normally.
+
+'''Conclusion :'''
+
+I test this method for the PFT 6 and PFT 10.
+
+I conclude that we cannot set a strict criterion over the carbon (0.01%) and over the biomass (0.1% didn't work) due to the reactif pools (metabolic litter :
+a little variation implies a "big" relative error). [[BR]]
+So we decide to focus only to the passive pool for computing the error.
+  
+
+=== 2) Second implementation : ===
+
+The same as above except the relative error for the carbon is based on the passive pool only. [[BR]]
+We choose to set the threshold for the carbon to 0.01%.[[BR]]
+
+I make a run whose the simulation lenght was 500Y max. [[BR]]
+If the biomass equilibrium is not reached, the resolution start after 495Y for the last five years.[[BR]]
+
+Here are the results for the PFT 6 :
+
+
+|| BIOMASS THRESHOLD || TIME EQUILIBRIUM BIOMASS || PASSIVE THRESHOLD || SPINUP LENGTH || SOLUTION ||
+|| 0.5% ||  85Y || 0.01% || 246Y || ||
+|| 0.3% || 103Y || 0.01% || 222Y || ||
+|| 0.2% || 118Y || 0.01% || 206Y || ||
+|| 0.1% || 495Y (forced resolution) || 0.01% || 498Y || ||
+
+For the PFT 10, the maximum length simulation was 200Y. The forced resolution start after 195Y.
+|| BIOMASS THRESHOLD || TIME EQUILIBRIUM BIOMASS || PASSIVE THRESHOLD || SPINUP LENGTH ||
+||  ||  Y || 0.01% || Y ||
+||  ||  Y || 0.01% || Y ||
+||  ||  Y || 0.01% || Y ||
+||  || 495T (forced resolution) || 0.01% || 498Y ||
+
+
+'''Conclusion :'''
+
+The threshold 0.1 % for the biomass stock cannot be reached so I have to wait 200Y for the equilibrium of the biomass for the grasses![[BR]]
+
+Because of the grasses and of the intrisic instabilities of the model, we decide to make a third change. [[BR]]
+For the biomass criterions, we focus now only on the fluxes : turnover_daily_accu and bm_to_litter_accu. [[BR]]
+We look at the relative errors every 10 years (decennal mean of the fluxes).
+
+
+
+=== 3) Third implementation : ===
+
+In order to reduce intrisic instabilities of the model which affect the method, we look at the relative error on the decennal mean fo the biomass fluxes.[[BR]]
+
+For the same reason, we solve the spinup system every 10 years after the starting of the method. I used the same protocol for the PFT 6 and 10 described 
+above :
+
+Results for PFT6 :
+
+|| BIOMASS THRESHOLD || TIME EQUILIBRIUM BIOMASS || PASSIVE THRESHOLD || SPINUP LENGTH || NUMBER OF SYSTEM RESOLUTION || SOLUTION ||
+|| 0.5% ||  90Y || 0.01% || 330Y || 15 || ||
+|| 0.25% || 210Y || 0.01% || 311Y || 11 || ||
+|| 0.1% || 250Y || 0.01% || 260Y || 2 || ||
+|| 0.01% || 340Y || 0.01% || 361Y || 3 || ||
+
+Results for PFT10 (only one test made):
+
+|| BIOMASS THRESHOLD || TIME EQUILIBRIUM BIOMASS || PASSIVE THRESHOLD || SPINUP LENGTH || NUMBER OF SYSTEM RESOLUTION || SOLUTION ||
+|| 0.1% || 40Y || 0.01% || 60Y || 3 || ||
+
+