I have always been amazed by the complexity and diversity of ecological communities, which is a wonderful dream for a passionate field ecologist, but a nightmare for a theoretician trying to find mathematical solutions to the eco-puzzle. For several years, I conducted experimental and natural gradient field research on how human land use and landscape structure impact and interact with community composition and diversity in the agricultural landscapes of North-Eastern Germany. Over the last few years, I have shifted into theoretical and computational ecology to elucidate and explain general patterns of ecosystem structure and function, focusing on networks of predator-prey interactions expressed as complex food webs . I am using this framework to explore how food web structure and dynamics interact to promote species' persistence (survival) and network resilience to perturbations such as keystone species loss or nutrient enrichment. The food web structures studied include trophic interactions (who eats whom), spatial structure (who eats whom at which place) and body size distributions (big fish eat small fish or small parasitoid on large host). When compared to random network structures, it appears that these trophic, spatial and allometric aspects of food web structure fit together amazingly well to provide higher chances of persistence for species and thus maintain the overwhelming diversity and complexity found in natural ecosystems. Very recently, I started studying the interplay between keystone species and network structure in preserving ecological diversity with my colleagues Eric Berlow and Neo Martinez. We found that when embedded in a broader network context than traditionally studied, keystone effects are characterized by compensatory context-dependency that dampens natural variation in top-down or bottom-up forces to promote resilience of complex ecosystems to such perturbations. In addition to working with food webs, I am interested how we can empirically measure species richness and other characteristics of complex communities in the field. This work includes evaluation of estimation methods using simulated in silico landscapes.