The ISO 19906 arctic structures standard specifies that ice loads be calculated at the extreme level (100 year return period) for verifying ultimate limit states and at the abnormal level (10,000 years) for accidental/abnormal limit states. Since ice load measurements on structures have only been made over much shorter time periods, concerns are often expressed about the accuracy to which 10,000 year values can be estimated.
In this paper, the uncertainties in 100 and 10,000 year loads are considered through examples based on experience with calculations of loads on structures in different iceberg and sea ice environments. For icebergs, it is necessary to consider the size distribution of icebergs (including the potential presence of extremely large icebergs and ice islands) as well as drift velocities and shapes that can govern high return- period loads for fixed structures. With sea ice, abnormal-level loads can be governed either by the presence and geometrical properties of large discrete features (e.g. first-year ridges and stamukhi, or in the arctic, multi-year floes with thick ridges), or by very thick ice as a result of thermal growth. It is demonstrated how errors in key contributing ice parameters can influence extreme-level loads, and the relationship between level/rafted ice loads at the abnormal level and the factored (1.35) extreme-level values, and how these uncertainties might be considered in the design process.