In hydrological researches, the fine-scale precipitation data is always a matter of importance. The obstacle of seldom availability of such data, however hugely affects the research processes. Although there are many models proposed to cover the lack of data, usually they fail to simulate the intermittency of rainfall due to its highly stochastic variation. Intermittency refers to alternate behavior of rainfall which starts with a storm then followed by a dry period, then again after a while another storm starts. In the present thesis, it is tried to propose a model to produce synthetic data series while it simulates intermittency of rainfall at a satisfactory level. Then simulated data series can be used to cover the lack of real data. For this purpose, the Neyman-Scott Rectangular Pulse model (NSRP model) is applied to produce the synthetic data series based on the statistical information gathered from the observed data series. The observed data, which is used as input for the model, is rain depth in the period of 6 years from 1st of January 2012 till 31st of December 2016 for seven specific stations in Milan city. The fine-timescale available with the observed data is with a time step of 1 minute. One hundred simulations have been obtained as the synthetic data series. The average, variance, skewness and autocorrelation of data are taken as sample moments for both synthetic and observed data series. Although the synthetic data cannot coincide with the observed data based on NSE and R^2 statistic evaluations, the sample moments are in coincidence at a satisfactory level. Hence, synthetic data can show the behavior of observed data in an acceptable standard. Also, intermittency of rainfall, the proportion of dry, wet days and extreme values of rainfall are compared between synthetic and observed rainfall series. The NSRP model parameters are optimized with the main time scales of 1 minute and 8 hours’ sample moments. Besides, sample moments in some other time scales are considered in optimization function which leads to improving optimization. The comparisons of sample statistics of observed data and synthetic data are shown in the diagrams. Furthermore, the sample statistics of each synthetic data are compared with means of box plots. The result of comparisons proves the ability of synthetic data in preserving the main statistic behavior of observed ones. The main purpose in the present thesis is the production of synthetic data series which preserve not only sample statistics of observed data series but also its intermittent behavior. The intermittency curves are plotted with the use of the expected maximum value of precipitation in different moving time durations. Promisingly, the intermittency curve of synthetic data shows similar behavior change in a critical time scale of about 1 hour, it simulates the general trend of the observed intermittency curve. The model seems successful in terms of simulating intermittency and dry periods.
In Idrologia, i dati di precipitazione su larga scala sono sempre una questione importante. La scarsa disponibilità di tali dati, tuttavia, influisce enormemente sulle analisi e sugli studi presenti in letteratura. Sebbene siano disponibili molti modelli di simulazione della variabilità temporale della precipitazione, di solito questi non riescono a simulare il carattere intermittente della precipitazione. L'intermittenza si riferisce al comportamento ON-OFF delle precipitazioni che vede l’alternanza di periodi di calma a periodi di intensa attività, elevate intensità in brevi periodi di tempo. Nella presente tesi, si tenta di proporre un modello per generare serie sintetiche di dati di precipitazione ad elevata risoluzione temporale. A tal fine, il modello di Neyman-Scott Rectangular Pulses (modello NSRP) è stato considerato per generare serie sintetiche sulla base di dati osservati ad una risoluzione temporale di un minuto su un periodo di 6 anni dal 1 ° gennaio 2012 al 31 dicembre 2016 per sette stazioni nella città di Milano. E’ bene notare il modello NSRP non è mai stato utilizzato per lo studio dell’intermittenza della precipitazione, né calibrato su serie temporali ad elevata risoluzione come invece fatto nel presente lavoro di tesi. Pertanto, sono state generate 100 serie di dati sintetici. I seguenti momenti statistici: media, varianza, asimmetria e autocorrelazione sono stati considerati come metriche di paragone tra dati osservati e dati simulati. Sebbene i dati sintetici possano non coincidere con i dati osservati in base a indici di performance come NSE e R^2, i momenti statistici menzionati tra dati simulati e quelli osservato sono in buon accordo. Fase critica in questo studio è stata la calibrazione dei parametri del modello. Essi sono stati ottenuti ottimizzando le performance del modello su alcune scale temporali, nella fattispecie quelle tra 1 minuto e 8 ore. Lo scopo principale della presente tesi è la generazione di serie sintetiche di precipitazione che conservano non solo le statistiche campionarie delle serie di dati osservate, ma anche il comportamento intermittente della precipitazione. Effettivamente nella tesi si è dimostrato che il comportamento intermittente della precipitazione è stato ben riprodotto anche nelle serie sintetiche.
Neyman-Scott rectangular pulses model's calibration for fine-resolution rainfall data
KIAN, SINA
2018/2019
Abstract
In hydrological researches, the fine-scale precipitation data is always a matter of importance. The obstacle of seldom availability of such data, however hugely affects the research processes. Although there are many models proposed to cover the lack of data, usually they fail to simulate the intermittency of rainfall due to its highly stochastic variation. Intermittency refers to alternate behavior of rainfall which starts with a storm then followed by a dry period, then again after a while another storm starts. In the present thesis, it is tried to propose a model to produce synthetic data series while it simulates intermittency of rainfall at a satisfactory level. Then simulated data series can be used to cover the lack of real data. For this purpose, the Neyman-Scott Rectangular Pulse model (NSRP model) is applied to produce the synthetic data series based on the statistical information gathered from the observed data series. The observed data, which is used as input for the model, is rain depth in the period of 6 years from 1st of January 2012 till 31st of December 2016 for seven specific stations in Milan city. The fine-timescale available with the observed data is with a time step of 1 minute. One hundred simulations have been obtained as the synthetic data series. The average, variance, skewness and autocorrelation of data are taken as sample moments for both synthetic and observed data series. Although the synthetic data cannot coincide with the observed data based on NSE and R^2 statistic evaluations, the sample moments are in coincidence at a satisfactory level. Hence, synthetic data can show the behavior of observed data in an acceptable standard. Also, intermittency of rainfall, the proportion of dry, wet days and extreme values of rainfall are compared between synthetic and observed rainfall series. The NSRP model parameters are optimized with the main time scales of 1 minute and 8 hours’ sample moments. Besides, sample moments in some other time scales are considered in optimization function which leads to improving optimization. The comparisons of sample statistics of observed data and synthetic data are shown in the diagrams. Furthermore, the sample statistics of each synthetic data are compared with means of box plots. The result of comparisons proves the ability of synthetic data in preserving the main statistic behavior of observed ones. The main purpose in the present thesis is the production of synthetic data series which preserve not only sample statistics of observed data series but also its intermittent behavior. The intermittency curves are plotted with the use of the expected maximum value of precipitation in different moving time durations. Promisingly, the intermittency curve of synthetic data shows similar behavior change in a critical time scale of about 1 hour, it simulates the general trend of the observed intermittency curve. The model seems successful in terms of simulating intermittency and dry periods.| File | Dimensione | Formato | |
|---|---|---|---|
|
10.pdf
non accessibile
Descrizione: Thesis text
Dimensione
16.89 MB
Formato
Adobe PDF
|
16.89 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/152781