A discard ban for fish was introduced in Norway in 1987, which requires that all commercial catches must be landed and reported. In theory, this regulation creates a full record of total removals from all fisheries. However, exemptions and varying compliance rates create a risk that unreported catches still occur. Estimating unreported catches of all species in multiple fisheries is a large task that is complexified by the many influential factors related to unique fishery regulations, market demands, fishing gear, and species biology. There is therefore a need to standardise the estimation procedure, but this requires compromises that affect the bias and precision variably across individual species which must be understood if results are used as scientific advice. In Norwegian fisheries, the largest source of detailed data on unreported catches comes from the Norwegian Reference Fleet, a group of active fishing vessels that are paid to sample their catches at sea. However, participation in the programme is voluntary, meaning there are uncertainties about how representative the Norwegian Reference Fleet are of the wider fisheries. In such a complex system, it is important to address uncertainties in the entire estimation process, including from sampling data and the estimators used. The aim of this thesis is to develop standardised estimators for unreported catches in Norwegian fisheries. To identify the current knowledge gaps in Norwegian fisheries, global best practices for estimating unreported catches were collated and applied to Norwegian fisheries. Following from this, two research paths were identified. Firstly, there is a demand to understand the quality of data collected by the Norwegian Reference Fleet. Based on the available data, this was confined to quantifying the representativeness of samples. Secondly, previous studies estimating unreported catches have used bespoke model-based approaches to improve predictive performance, but simple design-based approaches have been applied based on assumptions that have not yet been tested. There is therefore a demand to evaluate the assumptions behind the current design-based estimators. To evaluate representativeness, the sampling design of the Norwegian Reference Fleet was simulated using reported catches, for which fleet-level information is available. The simulation study identified that nonprobability sampling of vessels in the Norwegian Reference Fleet results in a tendency to overestimate reported catches, but the bias is still within the bounds of expected variation from probability sampling. Representativeness varied greatly across species and years, and there was evidence that the estimators traditionally used for unreported catches may be introducing bias due to assumptions being unmet. These results provide support for the development of improved estimators and consideration of a more conservative estimation of uncertainty. Applying a cluster-based estimator that better describes true variations between sampled vessels produces a more realistic, albeit more uncertain estimate of unreported catches. This is also the case for additional uncertainty incurred from converting numbers of fish to biomass, which must use an additional modelling step due to a lack of information on fish weights. The current methodology for estimating discards in coastal fisheries is restricted by the fishery-level data that is used for extrapolating estimated discard rates. However, current developments in mandatory reporting requirements suggest that future model-based approaches could improve discard estimates. Therefore, an exploratory model was fitted to the sampling data to identify potentially important variables that explain variations in discarding. This model can then inform the variable selection in a future model-based approach when fishery-level data collection is improved. The estimation methodologies presented in this thesis form the basis of a national routine for estimating unreported catches in Norwegian fisheries. Quantifying the bias of estimators and accounting for additional, important sources of uncertainty provides a standardised design-based estimator for unreported catches in Norwegian fisheries. Predictive performance is now supported by quantitative evidence and further improvements have been identified to optimise estimators in the future such as accounting for rare occurrences and size-based estimates. Furthermore, the lessons learnt throughout this doctoral research highlight the importance of creating a standardised framework for estimating unreported catches. This ensures that improvements are centralised rather than being hidden within individual case studies.