INCORPORATION OF PHYSICAL AND REMOTE SENSING DATABASE INTO QUASI-DISTRIBUTED HYDROLOGICAL MODELING
IMERG. CHIRPS. PERSIANN. MapBiomas. HEC-HMS.
Estimating, quantifying or predicting physical quantities related to natural phenomena is a scientific challenge that continually nurtures different areas of knowledge. This thesis contributes with a closer approach to the methodologies that quantify precipitation and flow in hydrographic basins. The accuracy of four precipitation products estimated by remote sensing (PES), from the series IMERG-F, CHIRPS 2.0, PERSIANN-CDR and PERSIANN-PDIR, was compared to the precipitation observed in pluviometers in 15 hydrographic basins of Pernambuco, in annual periods, monthly and daily. The Hydrologic Engineering Center - Hydrologic Modeling System (HEC-HMS) program was applied to quasi-distributed hydrological modeling through the ModClark method, adopting the premise of incorporating open databases, such as MapBiomas, that represent physical characteristics of the studied areas. The main results indicated variations from -53% to 54% over the average annual PES, with differences from -2% to 5% when observing the most accurate annual results, which occurred in basins with semi-arid climate. The best percentage of observed events detection (POD) occurred for the CHIRPS base with 62% of detection in the Brígida river basin, followed by 59%, 56% and 55%, respectively for the PERSIANN-CDR, IMERG-F and PERSIANN-PDIR, both for the Pajeú river basin. The physical data used allowed a quick and satisfactory initialization of the HEC-HMS, when efficiency indicators such as Nash-Sutcliffe (NSE) and the BIAS frequency bias were verified. Sirinhaém, Una and Pajeú rivers basins, representing distinct climatic regions, had only the base flow calibrated, and respectively, resulted in NSE values of up to 0.94, 0.90 and 0.79; and BIAS of less than 5% in at least one fluviometric station in each hydrographic basin. Thus, it is concluded that the PES has distinct error patterns over areas of semi-arid and humid climate, and low potential to feed decision support systems at the scale of the evaluated hydrographic basins. It is also concluded that the physical database used satisfactorily represented the observed flows, and the methodology used is a viable alternative for hydrological modeling in regions without observed data.