RAINFALL-RUNOFF MODELING: INTEGRATION OF DATA-BASED AND PROCESS-INFORMED MODELS TO STRENGTHEN CONCEPTUAL HYDROLOGICAL PREDICTION AND INTERPRETATION

RAINFALL-RUNOFF MODELING IN ANDEAN SYSTEMS REMAINS CHALLENGING DUE TO THE COMPLEXITY OF THE PROCESSES AND THE LIMITED OBSERVABILITY OF VARIABLES. HERE, DATA-BASED AND PROCESS-INFORMED MODELS ARE ESPECIALLY VALUABLE, SINCE THEY CAPTURE THE ESSENTIAL DYNAMICS WITH HIGH PREDICTIVE CAPACITY FROM SPACE INFORMATION. THIS PROJECT FOCUSES ON TWO APPROACHES OF THIS NATURE. THE FIRST, A FIRST ORDER NONLINEAR DYNAMIC SYSTEM (SDNPO), SUITABLE FOR ESTIMATING UNOBSERVED PRECIPITATION AND EVAPOTRANSPIRATION, DERIVING UNIT HYDROGRAMS AND INFERING INTRINSIC PROPERTIES THROUGH BACK-ENGINEERING. THE SECOND, A DATA-BASED MECHANICIST MODEL (MBD), RECOGNIZED FOR ITS HIGH PREDICTIVE ACCURACY THANKS TO ROBUST ASSIMILATION TECHNIQUES, INCLUDING STATE-DEPENDENT PARAMETERS (PDE). HOWEVER, THE SDNPO IS HIGHLY SENSITIVE TO THE IDENTIFICATION OF THE STORAGE-DISCHARGE CURVE. IN CONTRAST, THE MBD CONCEPTUALIZES RAINFALL-RUNOFF THROUGH RESERVOIRS, BUT USUALLY STICKS TO CONVENTIONAL HYDROLOGICAL INTERPRETATIONS, WHICH LIMITS ITS SCOPE. THIS PROJECT EXPLAINS AND INTEGRATES BOTH FRAMEWORKS WITH TWO COMPLEMENTARY PURPOSES: (I) APPLY PDE TECHNIQUES TO OBJECTIVELY IDENTIFY THE STORAGE-DOWNLOAD FUNCTIONS OF THE SDNPO AND, THEREFORE, INCREASE THE CONFIDENCE OF ITS PERFORMANCES; AND (II) FORMULATE A MBD EQUIVALENT TO THE SDNPO TO EXPAND ITS INTERPRETABILITY AND INFER PROCESSES BEYOND CONVENTIONAL APPROACHES -FOR EXAMPLE, STATE-DEPENDENT EVAPOTRANSPIRATION CURVES-. THIS RESEARCH HAS NO PRECEDENTS AND IS USEFUL, SINCE ITS SYNERGIES MAXIMIZE CONFIDENCE IN THE MODELS AND SUPPORT THEIR APPLICATION TO OTHER SYSTEMS WITH LITTLE MONITORING, BEYOND THE ANDEANS.

XXII UNIVERSITY COMPETITION FOR RESEARCH PROJECTS