obtaining, by one or more processors, real-time operating data indicating operating conditions of the autonomous vehicle during a current trip;

determining, by the one or more processors and based on a geographic location of the autonomous vehicle, first environmental data indicating one or more first weather conditions encountered by the vehicle as it traverses a current route during the current trip;

obtaining, by the one or more processors, contextual data comprising historical weather data indicating weather conditions encountered by the autonomous vehicle during a previous trip taken along a first route, wherein the first route is different from the current route associated with the current trip;

obtaining, by the one or more processors, other data associated with other vehicles operating in an environment surrounding the autonomous vehicle, the other data comprising (i) other real-time operating data indicating one or more behaviors of the other vehicles during the current trip, and (ii) historical operating data indicating past behaviors of the other vehicles during at least one trip taken along a second route different from the current route;

selecting, by the one or more processors, from a plurality of models, and based at least in part on the first environmental data, a first model

characterized by a first plurality of weights associated with the real-time operating data, the contextual data, and the other data;

determining, by the one or more processors, using the first model, and based at least in part on the first plurality of weights, the real-time operating data, the contextual data, and the other data, a first real-time safety threshold corresponding to the one or more first weather conditions;

determining, by the one or more processors and based at least in part on the real-time operating data, that the operating conditions of the vehicle satisfy the first real-time safety threshold;

determining, by the one or more processors and during the current trip, second environmental data, different from the first environmental data, indicating a change in the one or more first weather conditions to one or more second weather conditions along the current route;

selecting, by the one or more processors, from the plurality of models, and based at least in part on the second environmental data, a second model characterized by a second plurality of weights, different from the first plurality of weights;

determining, by the one or more processors and using the second model, a second real-time safety threshold corresponding to the one or more second weather conditions;

determining, by the one or more processors and based at least in part on the real-time operating data, that a particular vehicle operating condition exceeds the second real-time safety threshold;

determining, by the one or more processors, an instruction to modify a particular vehicle operation as the vehicle continues to traverse the current route during the current trip, based at least in part on the determining that the particular vehicle operating condition exceeds the second real-time safety threshold; and

providing, by the one or more processors, the instruction to a particular processor that is on-board the autonomous vehicle and that controls the particular vehicle operation, the instruction, when executed by the particular processor, causing the particular processor to modify the particular vehicle operation.