Este estudo apresentou uma metodologia para inferir hidrometeoros a partir de medidas polarimétricas de um radar meteorológico de dupla polarização banda X. A metodologia consiste em uma abordagem teórica baseada em simulações numéricas com um modelo de espalhamento Mie (T M atrix e M ueller) e uma experimental pautada na aplicação de um algoritmo de classificação de hidrometeoros (Dolan and Rutledge [2009]). As si- mulações teóricas foram utilizadas para estudar os efeitos da distribuição de tamanho de gotas, temperatura dos hidrometeoros, ângulo de elevação e mistura de hidrometeoros a partir do fator de refletividade do radar (Z), refletividade diferencial (Z DR ), fase diferen- cial especifica (K DP ) e coeficiente de correlação( HV ). Os valores de Z DR são 0.5 dBZ maior para a frequência de banda X do que para um banda S. A partir de Z maior que 45 dBZ K DP começa a ficar maior que 0. Já HV começa a diminuir quando Z é maior que 25 dBZ. Não se observa variações significativas para o graupel, porém para granizo K DP é maior que 0 quando Z é maior que 15 dBZ, entretanto, para água, os valores são consideravelmente menores. Os efeitos de temperatura só são notados quando Z é maior que 60 dBZ. Ao analisar o efeito da elevação, observa-se que Z DR diminui com o aumento da elevação, sendo mais sensível para Z maiores, o mesmo efeito é observado para K DP e HV . Estas variações são mais sensíveis para água e granizo do que para o graupel. Comparando as distribuições exponencial e gama para considerar os efeitos da distribuição do tamanho de gotas para o caso da chuva, nota-se que a distribuição exponenciais é maior que a gama quando as gotas aumentam seu tamanho e diminui sua concentração, devido que na simulação teórica foi utilizado N 0 fixo.Ao analisar os efeitos da co-existência de água e graupel, temos que as gotas de água dominam o sinal de Z quando Z for maior que 30 dBZ, já K DP será positivo(negativo) quando Z for maior(menor) que 35 dBZ da água, desde que Z do graupel seja menor que 10 dBZ, já HV tende a ficar próximo de 1 quanto mais graupel é observado. Para a mistura de granizo e água, Z da água domina o do granizo quando Z é maior que 45 dBZ, K DP é maior(menor) que zero quando Z for maior (menor) que 25 dBZ desde que Z do granizo seja menor que 10 dBZ, já Z DR da água(granizo) domina o do granizo(água) quando Z for maior(menor) que 45 dBZ. Na parte experimental, dois casos observados durante o experimento de campo do Projeto CHUVA no Vale do Paraíba em 8 de Fevereiro e 22 de Março de 2012 foram utilizados. A classificação de hidrometeoros segundo Dolan and Rutledge [2009] indicaram a presença de chuva próximo da superfície proveniente de graupel e granizo. Acima dos 5 km foram identificados a presença de graupel,granizo e cristais de gelo. Ao examinar as regiões classificadas como granizo e graupel dentro da região de 0 e 15 C com os resultados teóricos, é possível explicar a presença concomitante de água e granizo e água e graupel nestas regiões. / This study presents a methodology for inferring hydrometeors from polarimetric mea- surements of a X band double polarization meteorological radar. The methodology consists of a theoretical approach based on numerical simulations with a Mie scattering model (T-Matrix and Mueller) and an experimental approach based on the application of a classification algorithm of hydrometeors (Dolan and Rutledge [2009]). The theoretical simulations were used to study the effects of droplet size distribution, hydrometeor tem- perature, elevation angle and mixture of hydrometeors from radar reflectivity factor (Z), differential reflectivity (Z DR ), specific differential phase (K DP ) and correlation coefficient ( HV ). The values of Z DR are 0.5 dBZ higher for the X band frequency than for the S band frequency. From Z greater than 45 dBZ, K DP starts to get higher than 0. When Z is greater than 25 dBZ, HV starts to decrease. No significant variations are observed for the graupel, however for hail, K DP is greater than 0 when Z is greater than 15 dBZ, but these values were much lower than for water. Temperature effects are only noticed when Z is greater than 60 dBZ. When analyzing the effect of elevation, it is observed that Z DR decreases with increasing elevation, being more sensitive to Z larger; the same effect is observed for K DP and HV . These variations are more sensitive to water and hail than to the graupel. Comparing the exponential and gamma distributions to consider the effects of droplet size distribution in the case of rain, it is noted that the exponential distribution is larger than the gamma when the droplets increase in size and decrease in concentration, due to the fact that in the simulation was used N 0 fixed. When analyzing the effects of co-existence of water and graupel, we have that the water droplets dominate the Z signal when Z is greater than 30 dBZ, K DP will be positive (negative) when Z is greater (lower) than 35 dBZ of water, since Z of the graupel is less than 10 dBZ and HV tends to be close to 1 when more graupel is observed. For the mixture of hail and water, Z of water dominates that of hail when Z is greater than 45 dBZ, K DP is larger (smaller) than zero when Z is larger (smaller) than 25 dBZ since Z of hail is less than 10 dBZ and Z DR of water (hail) dominates hail (water) when Z is greater (lower) than 45 dBZ. In the experimental part, two cases observed during the field experiment of the RAIN Project in Vale do Paraíba on February 8 and March 22, 2012 were used. The classification of hydrometeors according to Dolan and Rutledge [2009] indicated the presence of rain near the surface coming of graupel and hail. Above 5 km were identified the presence of graupel, hail and ice crystals. When examining the regions classified as hail and graupel within the region of 0 and 15 C with the theoretical results, it is possible to explain the concomitant presence of water and hail and water and graupel in these regions.

Este estudo apresentou uma metodologia para inferir hidrometeoros a partir de medidas polarimétricas de um radar meteorológico de dupla polarização banda X. A metodologia consiste em uma abordagem teórica baseada em simulações numéricas com um modelo de espalhamento Mie (T M atrix e M ueller) e uma experimental pautada na aplicação de um algoritmo de classificação de hidrometeoros (Dolan and Rutledge [2009]). As si- mulações teóricas foram utilizadas para estudar os efeitos da distribuição de tamanho de gotas, temperatura dos hidrometeoros, ângulo de elevação e mistura de hidrometeoros a partir do fator de refletividade do radar (Z), refletividade diferencial (Z DR ), fase diferen- cial especifica (K DP ) e coeficiente de correlação( HV ). Os valores de Z DR são 0.5 dBZ maior para a frequência de banda X do que para um banda S. A partir de Z maior que 45 dBZ K DP começa a ficar maior que 0. Já HV começa a diminuir quando Z é maior que 25 dBZ. Não se observa variações significativas para o graupel, porém para granizo K DP é maior que 0 quando Z é maior que 15 dBZ, entretanto, para água, os valores são consideravelmente menores. Os efeitos de temperatura só são notados quando Z é maior que 60 dBZ. Ao analisar o efeito da elevação, observa-se que Z DR diminui com o aumento da elevação, sendo mais sensível para Z maiores, o mesmo efeito é observado para K DP e HV . Estas variações são mais sensíveis para água e granizo do que para o graupel. Comparando as distribuições exponencial e gama para considerar os efeitos da distribuição do tamanho de gotas para o caso da chuva, nota-se que a distribuição exponenciais é maior que a gama quando as gotas aumentam seu tamanho e diminui sua concentração, devido que na simulação teórica foi utilizado N 0 fixo.Ao analisar os efeitos da co-existência de água e graupel, temos que as gotas de água dominam o sinal de Z quando Z for maior que 30 dBZ, já K DP será positivo(negativo) quando Z for maior(menor) que 35 dBZ da água, desde que Z do graupel seja menor que 10 dBZ, já HV tende a ficar próximo de 1 quanto mais graupel é observado. Para a mistura de granizo e água, Z da água domina o do granizo quando Z é maior que 45 dBZ, K DP é maior(menor) que zero quando Z for maior (menor) que 25 dBZ desde que Z do granizo seja menor que 10 dBZ, já Z DR da água(granizo) domina o do granizo(água) quando Z for maior(menor) que 45 dBZ. Na parte experimental, dois casos observados durante o experimento de campo do Projeto CHUVA no Vale do Paraíba em 8 de Fevereiro e 22 de Março de 2012 foram utilizados. A classificação de hidrometeoros segundo Dolan and Rutledge [2009] indicaram a presença de chuva próximo da superfície proveniente de graupel e granizo. Acima dos 5 km foram identificados a presença de graupel,granizo e cristais de gelo. Ao examinar as regiões classificadas como granizo e graupel dentro da região de 0 e 15 C com os resultados teóricos, é possível explicar a presença concomitante de água e granizo e água e graupel nestas regiões. / This study presents a methodology for inferring hydrometeors from polarimetric mea- surements of a X band double polarization meteorological radar. The methodology consists of a theoretical approach based on numerical simulations with a Mie scattering model (T-Matrix and Mueller) and an experimental approach based on the application of a classification algorithm of hydrometeors (Dolan and Rutledge [2009]). The theoretical simulations were used to study the effects of droplet size distribution, hydrometeor tem- perature, elevation angle and mixture of hydrometeors from radar reflectivity factor (Z), differential reflectivity (Z DR ), specific differential phase (K DP ) and correlation coefficient ( HV ). The values of Z DR are 0.5 dBZ higher for the X band frequency than for the S band frequency. From Z greater than 45 dBZ, K DP starts to get higher than 0. When Z is greater than 25 dBZ, HV starts to decrease. No significant variations are observed for the graupel, however for hail, K DP is greater than 0 when Z is greater than 15 dBZ, but these values were much lower than for water. Temperature effects are only noticed when Z is greater than 60 dBZ. When analyzing the effect of elevation, it is observed that Z DR decreases with increasing elevation, being more sensitive to Z larger; the same effect is observed for K DP and HV . These variations are more sensitive to water and hail than to the graupel. Comparing the exponential and gamma distributions to consider the effects of droplet size distribution in the case of rain, it is noted that the exponential distribution is larger than the gamma when the droplets increase in size and decrease in concentration, due to the fact that in the simulation was used N 0 fixed. When analyzing the effects of co-existence of water and graupel, we have that the water droplets dominate the Z signal when Z is greater than 30 dBZ, K DP will be positive (negative) when Z is greater (lower) than 35 dBZ of water, since Z of the graupel is less than 10 dBZ and HV tends to be close to 1 when more graupel is observed. For the mixture of hail and water, Z of water dominates that of hail when Z is greater than 45 dBZ, K DP is larger (smaller) than zero when Z is larger (smaller) than 25 dBZ since Z of hail is less than 10 dBZ and Z DR of water (hail) dominates hail (water) when Z is greater (lower) than 45 dBZ. In the experimental part, two cases observed during the field experiment of the RAIN Project in Vale do Paraíba on February 8 and March 22, 2012 were used. The classification of hydrometeors according to Dolan and Rutledge [2009] indicated the presence of rain near the surface coming of graupel and hail. Above 5 km were identified the presence of graupel, hail and ice crystals. When examining the regions classified as hail and graupel within the region of 0 and 15 C with the theoretical results, it is possible to explain the concomitant presence of water and hail and water and graupel in these regions.

This article, emerged from a bachelor thesis, focuses on the classification of hydrometeors measured by the Laser Precipitation Monitor by the Adolf Thies GmbH & Co. KG. The optical disdrometer can classify measurements of hydrometeor size and fall velocity spectra concerning the precipitation type. The measurement principle of the disdrometer is explained, as well as the classifications. For reasons of calculation time, mostly six main precipitation types are considered (drizzle, rain, snow, ice grains, hail, mixed). It is the goal to understand the process of a reliable classification and to determine how these classifications are implemented. Therefore, the precipitation measurements from the measurement field of the Leipzig Institute for Meteorology from 2021 are used. An analysis of the spectrum consisting of hydrometeor diameter and fall speed is investigated. Afterwards, two machine learning methods are applied to the dataset. The classification of each sample through grouping similar samples using cluster analysis serves as an unsupervised approach and in particular examines the natural clusters present in the dataset. Contrasting that the purely statistical, nonphysical, supervised Random Forest method is applied as well. The comparison of the unsupervised and supervised approach shows that for the classification the supervised method is more promising. / Dieser Artikel konzentriert sich auf die Klassifizierung von Hydrometeoren, welche durch den Laser Niederschlags Monitor der Adolf Thies GmbH & Co. KG gemessen werden. Das optische Disdrometer kann die Messungen von Fallgeschwindigkeits- und Größenspektren der Niederschlagspartikel eigenständig in Gruppen der Niederschlagsart einsortieren. Das Messprinzip, sowie die Klassifizierungsmechanismen werden erklärt. Auf Grund der Rechenzeit werden im Rahmen der folgenden Untersuchungen hauptsächlich die 6 Hauptniederschlagsarten (Niesel, Regen, Schnee, Eiskörner, Hagel, Gemischt) unterschieden. Das Ziel der Analyse ist es, den Prozess einer zuverlässigen Klassifizierung zu verstehen und die Möglichkeiten der Anwendung abzuschätzen. Dafür werden die Niederschlagsdaten der Wetterwiese des Leipziger Instituts für Meteorologie aus dem Jahr 2021 verwendet. Nach erster grundlegender Betrachtung des Datensatzes werden zwei verschiedene Machine Learning Methoden angewendet. Als unüberwachte Methode dient der Ansatz der Clusteranalyse, welcher alle Samples über Ähnlichkeitskriterien gruppiert und dadurch die natürliche Gruppierbarkeit eines Datensatzes aufzeigt. Im Gegensatz dazu steht die rein statistische, unphysikalische Methode des Random Forest mit überwachtem Lernprozess. Im Vergleich beider Ansätze zeigt sich, dass ein überwachter Machine Learning Methode zufriedenstellendere Ergebisse erzeugt als unüberwachte Prozesse.

Cette thèse s’intéresse à la variabilité de la relation mass-diamètre (m(D)) des hydrométéores en phase glace présents dans les systèmes convectif de moyenne échelle (MCS). Elle s’appuie sur une base de données acquise pour 4 types de MCS différents durant 4 campagnes de mesure aéroportée : (i) MCS de la mousson Africaine (Continent ; MT2010), (ii) MCS de l’océan Indien (MT2011), (iii) MCS de la Méditerranée (côtes ; HyMeX), (iv) MCS de la mousson Nord-Australienne (côtes ; HAIC-HIWC). La relation m(D) est calculée à partir de l’analyse combinée des images des hydrométéores enregistrées par les sondes optiques et les facteurs de réflectivité mesurés à l’aide d’un radar Doppler embarqués sur le même avion de recherche. Il est d’usage que la relation m(D) des hydrométéores soit représentée par une loi puissance (avec un pré-facteur et un exposant), qui doit être contrainte par des informations supplémentaires sur les hydrométéores. Une étude théorique sur les formes des hydrométéores à l’aide de simulations en 3 dimensions dans lesquelles les hydrométéores sont orientés aléatoirement et projeté sur un plan, permet de contraindre l’exposant β de la relation m(D) en fonction de l’exposant σ de la relation surface-diamètre (S(D)). La relation S(D) est aussi représentée par une loi puissance, et elle peut-être calculée pour une population d’images d’hydrométéores enregistrés par les sondes optiques. La variabilité de l’exposant est finalement calculée à partir de la variabilité de l’exposant σ déduis des images des hydrométéores. Ensuite le pré-facteur α est calculé à partir de simulations des facteurs de réflectivité, de sorte que les facteurs de réflectivité simulés soient égaux aux facteurs de réflectivité mesurés par le radar nuage le long de la trajectoire de l’avion dans les MCS. Des profils moyens en fonction de la température sont calculés pour les coefficients de la relation m(D), les distributions en tailles des hydrométéores et les contenus massiques de glace dans les MCS (CWC). Les profils moyens pour les quatre types de MCS sont différents les uns des autres. Pour les quatre types de MCS, il est montré que les variations des coefficients de la relation m(D) sont corrélées avec les variations de la température. Four types de paramétrisations de la relation m(D) sont calculées depuis l’analyses des variations des coefficients de la relation m(D). Le bénéfice apporté par l’utilisation de relation m(D) non constante contrairement à l’utilisation de relation m(D) avec α et β constant, est démontré en étudiant l’impact de toutes les paramétrisations de la relation m(D) sur le calcul des relations Z-CWC et Z-CWC-T. / This study focuses on the variability of mass-diameter relationships (m(D)) and shape of ice hydrometeors in Mesoscale Convective Systems (MCS). It bases on data base which were recorded during four airborne measurement campaigns: (i) African monsoon’s MCS (continent; MT2010), (ii) Indian Ocean’s MCS (MT2011), (iii) Mediterranean’s MCS (costs; HyMeX), (iv) North-Australian monsoon’s MCS (costs; HAIC-HIWC). m(D) of ice hydrometeors are derived from a combined analysis of particle images from 2D-array probes and associated reflectivity factors measured with a Doppler cloud radar on the same research aircraft. Usually, m(D) is formulated as a power law (with one pre-factor and one exponent) that need to be constrained from complementary information on hydrometeors. A theoretical study of numerous hydrometeor shapes simulated in 3D and arbitrarily projected on a 2D plan allowed to constrain the exponent β of the m(D) relationship from the exponent σ of the surface-diameter S(D) relationship, which is likewise written as a power law. Since S(D) always can be determined for real data from 2D optical array probes or other particle imagers, the evolution of the m(D) exponent can be calculated. After that, the pre-factor α of m(D) is constrained from theoretical simulations of the radar reflectivity factor matching the measured reflectivity factor along the aircraft trajectory. Mean profiles of m(D) coefficients, particles size distributions and Condensed Water Content (CWC) are calculated in functions of the temperature, and are different for each type of MCS. For the four types of MCS, it is shown that the variability of m(D) coefficients is correlated with the variability of the temperature. Four types of m(D) parametrisations are calculated since the analysis of the variability of the m(D) coefficients. The significant benefit of using variable m(D) relations instead of a single m(D) relationship is demonstrated from the impact of all these m(D) relations on Z-CWC and Z-CWC-T fitted parametrisations.