| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: AfricanClimate@Home
Thread: Question about modeling of south african climate |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 19
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello, I have a question about climate modeling of this project.
I understand that the goal of this project is to develop the climate models of the south africa, but is it able to get the accurate climate models, just computing models, whose targeted region is only the south africa? I wonder if there are some influences on the region from the rest of the globe, so that the climate of the south africa would be changed, slightly or not. I'd like to know whether it is enough to get data of the specific regional climate, not of the global climate, in order to develop more accurate climate models of the south africa. thanks, suguruhirahara |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Regional climate models (RCMs) such as this one will be fed climate information from a global climate model (GCM), which is one reason why there is so much data to upload and download. Essentially the global climate model provides the data for the cells immediately surrounding the climate model, for every timestep within the two weeks of the model run.
----------------------------------------I don't know what the resolution of the model is (either time or space), but it'll probably be higher than the CPDN models. Every doubling in resolution requires 16x the amount of CPU time and memory, which is why regional models are necessary (i.e., you only model a small area at high res, and the rest at low res). The other way of running RCMs is to run both a GCM and an RCM at the same time, and produce the edge data that way (i.e., you only download the initial global climate, not the edge data). What I don't entirely understand is how the tiling will work - because this regional model is split up into tiles, and each tile is being fed edge data from a lower-resolution model, the climate within the regional model as a whole will be more constrained by the lower-resolution model than would be normal in a RCM (because there are extra sources of low-resolution data within the overall bounds of the RCM). It will also be a hard job to reassemble the tiles. In addition to that, because the global model is calculated in advance, rather than hand-in-hand with the higher resolution model, the edge outputs of the individual tiles within the regional model can't feed back into the global model or into their adjacent tiles. [Edit 6 times, last edit by Former Member at Sep 4, 2007 11:57:08 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The resolution of the runs in this phase is 30km (yes, higher than CPDN)
About the tiling: each work unit simulates 14 days for the whole domain over Southern Africa. We are searching better ways to create a full Africa domain for the next phase without increasing the size of input data (we already have some ideas). The boundary conditions were given by NCEP global models. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
That's very interesting, thanks.
In that case, I take it that the 84 simulations aren't tiles, but are individual models within an ensemble? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
That's very interesting, thanks. In that case, I take it that the 84 simulations aren't tiles, but are individual models within an ensemble? Exactly. Each of the 82 (not 84) simulations have a different perturbation so we are trying to create a "hyper-ensemble" (that's how our lead researcher Bruce likes to call it) |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks for replying;
The resolution of the runs in this phase is 30km (yes, higher than CPDN) About the tiling: each work unit simulates 14 days for the whole domain over Southern Africa. We are searching better ways to create a full Africa domain for the next phase without increasing the size of input data (we already have some ideas). The boundary conditions were given by NCEP global models. I see, so are you planning to take several phases, including what we are doing now? And one more question: do you enable any kind of optimisation of modelling, for example, SSE? If possible, it would be great, but I fear that would cause errors in the same way as CPDN has experienced... thanks, suguruhirahara |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Regional climate models (RCMs) such as this one will be fed climate information from a global climate model (GCM), which is one reason why there is so much data to upload and download. Essentially the global climate model provides the data for the cells immediately surrounding the climate model, for every timestep within the two weeks of the model run. I don't know what the resolution of the model is (either time or space), but it'll probably be higher than the CPDN models. Every doubling in resolution requires 16x the amount of CPU time and memory, which is why regional models are necessary (i.e., you only model a small area at high res, and the rest at low res). The other way of running RCMs is to run both a GCM and an RCM at the same time, and produce the edge data that way (i.e., you only download the initial global climate, not the edge data). Thanks for great explanation, mike :) Probably many participants like me of the project lacks these information since this is the first time for wcg to introduce a project out of the medicine. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The resolution of the runs in this phase is 30km (yes, higher than CPDN) Very much higher than CPDN: N 48 270km <-- CPDN coupled model (HadCM3, global model, 160 model years) N 96 135km N144 90km <-- CPDN SAP (HadAM3 + slab ocean, global model, 1 model year) N216 60km N312 30km <-- WCG AfricanClimate@Home (WRF, regional model, 2 model weeks) While (say) 90km to 30km may not sound like much, in reality it is a huge jump - 81 times as much CPU time required (since all 4 dimensions have tripled). If the CPDN-SAP model was done at N312 resolution, it'd take 3 years of CPU time to run, instead of 2 weeks of CPU time. [Edit 3 times, last edit by Former Member at Sep 4, 2007 3:52:11 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I see, so are you planning to take several phases, including what we are doing now? And one more question: do you enable any kind of optimisation of modelling, for example, SSE? If possible, it would be great, but I fear that would cause errors in the same way as CPDN has experienced... thanks, suguruhirahara There will definitely be another phase(s), but the results of phase one will drive our approach for the following ones. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello,
Are you planning to take several phases, including what we are doing now? And one more question: do you enable any kind of optimisation of modelling, for example, SSE? If possible, it would be great, but I fear that would cause errors in the same way as CPDN has experienced... There will definitely be another phase(s), but the results of phase one will drive our approach for the following ones. Then, which phase are we now? Is it the first one, in which the aim is to gain the basic information about the south african climate? thanks, suguruhirahara |
||
|
|
|