This thesis presents a new GIS-based scenario volcanic risk assessment model called RiskScape Volcano (RSV) that has been designed for the RiskScape program to advance the field of volcanic risk assessment. RiskScape is a natural hazards risk assessment software tool being developed in New Zealand by GNS Science and NIWA. When integrated into RiskScape, RSV will add proximal volcanic hazard risk assessment capability, and enhanced inventory design; it presently operates outside of RiskScape by combining volcanic hazard models’ output spatial hazard intensity (hazard maps) with inventory databases (asset maps) in GIS software to determine hazard exposure, which is then combined with fragility functions (relationships between hazard intensity and expected damage ratios) to estimate risk. This thesis consists of seven publications, each of which comprises a part of the development and testing of RSV: 1) results of field investigation of impacts to agriculture and infrastructure of the 2006 eruption of Merapi Volcano, Indonesia; 2) agricultural fragility functions for tephra damage in New Zealand based on the observations made at Merapi; 3) examination of wind patterns above the central North Island, New Zealand for better modeling of tephra dispersal with the ASHFALL model; 4) a description of the design, components, background, and an example application of the RSV model; 5) test of RSV via a risk assessment of population, agriculture, and infrastructure in the Rotorua District from a rhyolite eruption at the Okataina Volcanic Centre; 6) test of RSV via a comparison of risk to critical infrastructure in Mammoth Lakes, California from an eruption at Mammoth Mountain volcano versus an eruption from the Inyo craters; and 7) a survey of volcanic hazard awareness in the tourism sector in Mammoth Lakes. Tests of the model have demonstrated that it is capable of providing valid and useful risk assessments that can be used by local government and emergency management to prioritise eruption response planning and risk mitigation efforts. RSV has provided the RiskScape design team with a more complete quantitative volcanic risk assessment model that can be integrated into RiskScape and used in New Zealand and potentially overseas.

This thesis presents a new GIS-based scenario volcanic risk assessment model called RiskScape Volcano (RSV) that has been designed for the RiskScape program to advance the field of volcanic risk assessment. RiskScape is a natural hazards risk assessment software tool being developed in New Zealand by GNS Science and NIWA. When integrated into RiskScape, RSV will add proximal volcanic hazard risk assessment capability, and enhanced inventory design; it presently operates outside of RiskScape by combining volcanic hazard models’ output spatial hazard intensity (hazard maps) with inventory databases (asset maps) in GIS software to determine hazard exposure, which is then combined with fragility functions (relationships between hazard intensity and expected damage ratios) to estimate risk. This thesis consists of seven publications, each of which comprises a part of the development and testing of RSV: 1) results of field investigation of impacts to agriculture and infrastructure of the 2006 eruption of Merapi Volcano, Indonesia; 2) agricultural fragility functions for tephra damage in New Zealand based on the observations made at Merapi; 3) examination of wind patterns above the central North Island, New Zealand for better modeling of tephra dispersal with the ASHFALL model; 4) a description of the design, components, background, and an example application of the RSV model; 5) test of RSV via a risk assessment of population, agriculture, and infrastructure in the Rotorua District from a rhyolite eruption at the Okataina Volcanic Centre; 6) test of RSV via a comparison of risk to critical infrastructure in Mammoth Lakes, California from an eruption at Mammoth Mountain volcano versus an eruption from the Inyo craters; and 7) a survey of volcanic hazard awareness in the tourism sector in Mammoth Lakes. Tests of the model have demonstrated that it is capable of providing valid and useful risk assessments that can be used by local government and emergency management to prioritise eruption response planning and risk mitigation efforts. RSV has provided the RiskScape design team with a more complete quantitative volcanic risk assessment model that can be integrated into RiskScape and used in New Zealand and potentially overseas.

Recurrence rate is often used to describe volcanic activity. There are numerous documented ex- amples of non-constant recurrence rate (e.g. Dohrenwend et al., 1984; Condit and Connor, 1996; Cronin et al., 2001; Bebbington and Cronin, 2011; Bevilacqua, 2015), but current techniques for calculating recurrence rate are unable to fully account for temporal changes in recurrence rate. A local–window recurrence rate model, which allows for non-constant recurrence rate, is used to calculate recurrence rate from an age model consisting of estimated ages of volcanic eruption from a Monte Carlo simulation. The Monte Carlo age assignment algorithm utilizes paleomagnetic and stratigraphic information to mask invalid ages from the radiometric date, represented as a Gaussian probability density function. To verify the age assignment algorithm, data from Heizler et al. (1999) for Lathrop Wells is modeled and compared. Synthetic data were compared with expected results and published data were used for cross comparison and verification of recurrence rate and volume flux calculations. The latest recurrence rate fully constrained by the data is reported, based upon data provided in the referenced paper: Cima Volcanic Field, 33 +55/-14 Events per Ma (Dohren- wend et al., 1984), Cerro Negro Volcano, 0.29 Events per Year (Hill et al., 1998), Southern Nevada Volcanic Field, 4.45 +1.84/-0.87 (Connor and Hill, 1995) and Arsia Mons, Mars, 0.09 +0.14/-0.06 Events per Ma (Richardson et al., 2015). The local–window approach is useful for 1) identifying trends in recurrence rate and 2) providing the User the ability to choose the best median recurrence rate and 90% confidence interval with respect to temporal clustering.