Aquatic ecosystems are facing a growing number of human-induced stressors and the need to implement more biomonitoring to assess the ecological status of water bodies is eminent. This dissertation aims at providing tools to reduce the costs and improve the accuracy of freshwater benthic macroinvertebrate biomonitoring. To improve the cost-e ciency, we consider automated classi cation and develop a novel classi er suitable for complex macroinvertebrate image data. To enhance the accuracy of macroinvertebrate biomonitoring, we study the statistical properties of the Percent Model A nity index crucial to current Finnish biomonitoring and the factors a ecting these statistics. Finally, we perform a simulation study to analyze how di erent biological indices are a ected by misclassi cations in automated identi cation of macroinvertebrates.