Date of Award
Doctor of Philosophy (PhD)
Civil, Environmental & Architectural Engineering
Indonesia routinely suffers from floods, droughts, landslides and water-borne diseases resultant from significant spatial and temporal variability of precipitation. Mitigation of these hazards and efficient management of resources require tools for understanding and forecasting Indonesian hydroclimate. To date, research efforts aimed at understanding this hydroclimate variability are few and there exists no robust hydrologic modeling tool needed to understand and forecast important variability. Motivated by these gaps and needs, this dissertation makes four unique contributions: (i) A systematic space-time analysis of seasonal precipitation over Indonesia was performed using Principal Component Analysis and Bayesian Dynamic Linear Model. El Nino Southern Oscillation (ENSO) was found to be the driver of leading modes of variability during both the wet (Oct - Mar) and dry (Apr - Sep) seasons. Furthermore, ENSO appeared to drive variability at multi-decadal timescales (8 - 16 year) especially during post 1980 period. The association between ENSO and Indonesian rainfall has strengthened in recent decades, especially during dry season. These findings suggest potential for interannual and multidecadal predictability of Indonesian rainfall. (ii) To understand the processes that drive the hydrologic variability, we built and calibrated a distributed hydrologic model, the Variable Infiltration Capacity (VIC) model for six watersheds over Java, the most populous island of Indonesia. In light of data scarcity and quality issues, model skill scores during calibration period were quite good but comparatively lower skills during validation. The magnitude and variability of baseflow and direct runoff components were found to modulate the model performance. We provided preliminary evidence that performance could be improved by refining the spatial resolution of model and input precipitation and temperature. (iii) Following this, a high resolution gridded daily meteorology - precipitation, maximum and minimum temperature - data set at (~14 km) resolution spanning 30 years from 1985 - 2014 was developed over Java, using an Inverse Distance Weighting method. The data set was stored in a network common data format (NetCDF) and will be publicly available, intended to support basin-scale and island-scale studies of short-term and long-term climate, hydrology and ecology. (iv) In order to provide streamflow forecasts that capture model parametric uncertainty, ensembles of model parameters are necessary. For this, we conducted a multi-objective optimization based calibration of VIC model parameters. The method generated an ensemble of model parameters optimizing six objective functions that capture key aspects of the hydrograph. This improved upon the single objective based calibration. We demonstrated the utility of this in generating skillful seasonal hydrologic forecasts conditioned on seasonal climate forecasts.
The above contributions, especially the watershed modeling tools, are unique and is a first of its kind research efforts in this region. We offered insights into the space-time variability of precipitation and a robust physically-based watershed modeling tool to understand and forecast hydrologic variability over Java in particular and Indonesia in general. Together, this research makes significant strides in providing a framework for understanding, modeling and forecasting Indonesian hydrology and climatology, that will help mitigate natural hazards and enable efficient management of Indonesias water and natural resources.
Yanto, "Development of Data and Modeling Tools for Understanding and Forecasting Indonesian Hydroclimate" (2016). Civil Engineering Graduate Theses & Dissertations. 64.