Weather computer resolution
The global computer models take weather observations and transfer them onto a grid of dots, thereby averaging the weather out. Features that are occurring between the reports of the observation network are missed altogether, but some data such as temperature is observed in a continuous fashion by remote sensing from satellites and is covered well. The model has to “balance” this matrix of dots to ensure that certain rules are observed, for example all the computed upward motions need to be balanced by down ward motions so that the total atmosphere remains intact. Once a captured pattern is obtained, it is then pushed into the future using dynamic equations using small time-steps. Then another balancing computation is done and so forth. SO the GRIB data of a weather forecast is just a mathematical idea based on the extrapolation of a captured patter. In the real world chaos is continuously jiggling the weather pattern in to different direction, so that the weather forecast deviates from the real world.
You may already be aware of the limitations of GRIB data. Sometimes someone will offer you BRIB data with a better resolution, and the question arises if it will have better accuracy. The answer is somewhat mixed, as I hope to show you today:
The graphic above shows the normally accessible GRIB data (GFS) over Fiji on a typical trade wind day showing wind flow to be from E to SE 10 to 15 knots.
Here is a simultaneous output from the 8km resolution Predictwind version of the GFS model (PWG) as available (under registration) from www.predictwind.com
This model has captured some terrain effects—lowering the air pressure in the interior of the main islands of Fiji in the heat of the day, and there by forming a westerly sea-breeze around Nadi/Lautoka and causing an acceleration of the wind on the SW end of Fiji near Navula Passage. All of these terrain effects truly happen, even if they are not resolved by the normal GRIB data.
These models with higher resolution and closer dots (and smaller time-steps) require heaps more computing than the normal global models and give a pattern that is somewhat closer to the real world. However there is no increase in the basic weather observation resolution that goes into these models, so they still have limitations. Their output is still just an averaged idea.
From Bob McDavitt http://metbob.wordpress.com/