A Parallel PCG Solver For Groundwater Model In The Area Of Energy Base Of Mongolia

Energy Base of Mongolia lies in an arid and semiarid climatic region, and the annual evaporation is far larger than the annual rainfall. Now, It has a certain influence to the local economy and ecosystem, because of the increasing shortage of groundwater made by the local exploitation and utilization of the resource of Energy Base. Thus knowing the occurrence of ground water has the guiding significance to mark out the local exploitation and utilization of ground water properly.Based on lots of data on the hydrogeology, simulate the 3D groundwater flow region from the 2000 to 2006 years through GMS software. The result of simulation shows the total recharge of the study area is 465866.66*104m3/a, the total discharge is 535356.34*104m3/a, and the total released storage of aquifer is 69491.55*104m3/a. the primary recharge comes from precipitation infiltration, the recharge of which is 456903.71*104m3/a,accounting for 98.08% of the total recharge. In the meantime, the main discharge comes from evaporation of groundwater, the discharge of which is 428492.12*104m3/a, accounting for 80.04% of the total discharge. And the secondary discharge comes from artificial extraction and boundary discharge, accounting for 12.21%.However, due to the large study area, the outcome is rough to subdivide study area by using 1000m*1000m resolution. It could not describe groundwater flow regime finely, so the water level data of observing wells are useless, which are used to verify the simulative water level. Therefore, it needs us to describe investigated zone more finely, which will increase the amount of computation greatly and need processor's capability higher. Nevertheless, current single processor barges up against an avoidless capability bottleneck. So this study uses multiprocessor to parallelly compute the 3D flow expression.Based on the coefficient matrix of 3D flow expression anto created by the serial programme of MODFLOW-2000, this study use the OpenMP programming paradigm to parallelize the preconditioned conjugate-gradient(PCG) solver to solve 3D flow expression. From the outcome, we can know the execution time in the serial environment, which costs 20.705s, is typically about 1.05-1.11 times slower than that in the parallel environment. Apparently, it can easily solve the capability bottleneck of single processor, moreover, it can promote calculating efficiency greately.