| Atmospheric water vapor is the most important component in atmosphere and itstemporal and spatial distribution is very complex. It is important for the generation anddevelopment of weather system and the change of weather. The observation of atmosphericwater vapor distribution and evolution is one of the most difficult and important aspects foratmospheric sounding. The traditional measurement of water vapor limits our knowledgementof water vapor distribution due to the temporal and spatial resolution or the accuracy.Tomographying atmospheric water vapor using Global Positioning System (GPS) is a newtechnology of water vapor sounding, which has developed in recent ten years. The technologyhas potential application value for meso-and micro-scale water vapor distribution andevolution. The thesis studies on the ground-based GPS tomography thenology and itsapplication. It includes the following aspects: A new vertical constraint condition based onstatistical analysis of long time history radiosonde data was formulated;Two vertical layeringmethods were used to analyzed the difference of GPS tomography results; Three differentpriori programs(means the initial tomography model field) were used to compare and analyzetheir impact on the tomography results; The Monte Carlo radom simulation algorithm wasadopted to determine the optimum tomography solution and analyze the simulation times'impact on the Monte Carlo tomography solution; A water vapor tomography experiment wasdid in Fangshan of Beijing and the four-dimensional water vapor variation information wasgot using the primary GPS tomography result;The water vapor distribution chararistic wasstudied under three different weather conditions based on GPS tomography results. The mainconclusions were summarized as follows:⑴The exponential vertical constraint of GPS tomography equation was constructedusing the statistical fitting method based on1999～2009radiosonde data of Nanjiaoobservation station in Beijing. The GPS tomography equations were solved together with thenewly constructed vertical constraint equations. Comparing the water vapor information byGPS tomography method without vertical constraint with the water vapor information byradiosonde materials refutation, the correlation coefficient was about0.95, the root mean square error was0.70g/m3, the average relative bias error was about negative0.29g/m3. Afterthe vertical constraint was added, the correlation coefficient was about0.98, the root meansquare error was0.42g/m3, the average relative bias error (tomography-radionsonde) wasabout negative0.05g/m3. This showed the statistical fitting vertical constraint condition couldimprove the GPS tomography result.⑵In order to study the impact of vertical layering on the tomography result, we did thecontrast experiment using vertical uniform stratification(0.8km vertical height resolution) andvertical non-uniform stratification(0.5km vertical resolution between0～3km,1km verticalresolution between3～6km,2km vertical resolution between6～8km) methods. Thetomography results using non-uniform stratification methods agreed with the radiosonderesults well, the correlation coefficient was about0.98, the root mean square error was0.24g/m3, the average relative bias error was about negative0.1g/m3. The tomography watervapor information using uniform stratification methods compared with the radionsonde watervapor information: the correlation coefficient was about0.95, the root mean square error was0.73g/m3, the average relative bias error was about negative0.32g/m3. Acorrding to thecomparition, the tomography results with vertical non-uniform stratification was moreaccuracy more suitable to study water vapor.⑶The vertical priori information was important for the tomography water vapor result.The thesis proposed three vertical priori programs, those were: the exponential vertical prioriinformation (namely FR program),the monthly mean radiosonde result vertical prioriprogram(namely MR program) and three days mean radiosonde result vertical prioriprogram(namely TR program). Comparison of the tomography result using three verticalpriori programs between the real radionsonde water vapor information was made. The FRprogram results were: the correlation coefficient was about0.98, the root mean square errorwas0.66g/m3; The MR program results were: the correlation coefficient was about0.96, theroot mean square error was0.87g/m3; The TR program results were: the correlationcoefficient was about0.96, the root mean square error was0.99g/m3.The FR program agreedbest with the real radionsonde water vapor information.⑷The monte carlo radom simulation algrithom was used to get the best GPStomography solution and the simulation number of times played a important role on theMonte Carlo result. Comparing the GPS tomography with the radionsonde water vapor resultwhen the number of times was3000: the correlation coefficient was about0.95, the root meansquare error was1.04g/m3; the comparision with the number of times5000: the correlationcoefficient was about0.98, the root mean square error was0.62g/m3; the comparision withthe number of times10000: the correlation coefficient was about0.99, the root mean squareerror was0.49g/m3; In general, the tomography result was more accuracy and the RMS error was less, the correlation coefficient was higher with the simulation times more. It's notnecessary to take infinite large value of the simulation times due to the computationefficiency. What's more, the computation time increased largely with the increase ofsimulation times. We pointed out that the simulation times could be3000or5000times basedon the improved stochastical model and more accuracy simulation priori value consideringboth the accuracy and computation efficiency.⑸The GPS water vapor tomography experiment was do basing on the GPS observationnetwork material of Fangshan in Beijing. The atomosphere above the observation area wasstratified vertically non-uniformly,the exponential veritical constraint was added to thetomography equations,the monte carlo simulation times was chosen5000,then the24hourdaily continuous water vapor three-dimensional distribution information was got under thesepreconditions.The water vapor distribution primary analysis was did combined with theatmospheric dynamical and thermal condition parameters calculated using the Japan20km×20km reanalysis materials.⑹The water vapor distribution characteristic under rainstorm,sunny and light rainyweather conditions was studied based on the best GPS tomography result. The average watervapor value relative bias error between tomography height layers was adopted to describe thedifference of water vapor vertical distribution quantitatively.The average absolute bias errorof the average water vapor value of some tomography height layer with the individual cell'swater vapor value in the same layer was adopted to describe the water vapor differencehorizontal distribution quantitatively. The results were as follows: The heavy rainy weather.In the vertical direction, the water vapor amount decreased with the altitude increase, thewater vapor vary strongly and the water vapor had stratification structure.The water vaporwas mainly focused on below2km layers.Its average was about5.16g/m3and it accounted for77.35%of the total water vapor amount; The vertical distribution difference between the6～8km and5～6km layer was the biggest,the relative bias error was about0.75and it accountedfor28.4%of the total relative bias error；2.5～3km and2～2.5km layers' vertical differencewas the smallest,only0.04and the proportion was about1.51%to the total verticaldifference;The water vapor horizontal distribution was uniform totally,but the individual cellwater vapor different form the average water vapor of the respond layer.The5～6km layer'swater vapor horizontal distribution was the most different,about4.47×10-2g/m3,the layer'saverage water vapor was about0.76g/m3,that was about5.9×10-2every unit water vaporhorizontal difference.The most smallest horizontal difference was in0.5～1km layer,only0.88×10-2g/m3, the average water vapor was about6.65g/m3,that was about1.3×10-3everyunit water vapor horizontal difference. Sunny weather.The water vapor amount still decreased with the height increase and ithad some stratification structure.The water vapor mainly focused on below2km layers.Itsaverage water vapor was about4.56g/m3,the proportion was about76.69%to the total watervapor; The vertical distribution difference between the6～8km and5～6km layer was thebiggest,the relative bias error was about0.82and it accounted for33.7%of the total relativebias error；The vertical distribution difference between the2.5～3km and3～4km layer wasthe smallest,the relative bias error was about0.01and the proportion was about0.41%to thetotal relative bias error;The horizontal distribution was similar with the heavy rain weathertotally. The1.5～2km layer's water vapor horizontal distribution was the most different,about0.1g/m3,the layer's average water vapor was about1.59g/m3,that was about6.6×10-2everyunit water vapor horizontal difference.The most smallest horizontal difference was in0.5～1km layer,only1.65×10-2g/m3, the average water vapor was about6.53g/m3,that was about2.6×10-3every unit water vapor horizontal difference.Light rainy weather.The water vapor amount still decreased with the height increase andit had some stratification structure. Its average water vapor below2km layers was about4.98g/m3,the proportion was about71.94%to the total water vapor; The vertical distributiondifference between the3～4km and4～5km layer was the biggest,the relative bias error wasabout0.42and it accounted for24.53%of the total relative bias error； The verticaldistribution difference between the4～5km and5～6km layer was the smallest,the relativebias error was about0.002and the proportion was about0.11%to the total relative biaserror;The horizontal distribution was similar with the heavy rainy weather and sunny weathertotally. The5～6km layer's water vapor horizontal distribution was the most different,about7.97×10-2g/m3,the layer's average water vapor was about1.16g/m3,that was about6.87×10-2every unit water vapor horizontal difference.The most smallest horizontal difference was in0.5～1km layer,only1.57×10-2g/m3, the average water vapor was about4.85g/m3,that wasabout3.2×10-3every unit water vapor horizontal difference.According to the analysis quantitatively,the water vapor horizontal and verticaldistribution had both similarity and differentce under three different kind of weathercondition.In the horizontal direction,the water vapor horizontal had the same charactericunder various weather condition.That was the water vapor horizontal distribution was moreuniform and the horizontal difference was smaller in the lower layers.The biggest differenceof horizontal distribution was in the5～6km layer both heavy and light rainy weather and thesunny weather was different.In the vertical direction, the water vapor descreased with theheight increase and the vertical water vapor had some stratification structure.The smallestvertical difference was in the mid-and-low layers in various weather.The biggest verticaldifference was in the5～8km layers both in heavy rainy and sunny weather and the light rainy weather was different.
|
PDF Full Text Request