Two major foundations of ecosystem science are comparative analyses and long-term studies. Here, we explore the capacity of long-term and comparative data to predict lake characteristics (LCs). We ask if a variable is best explained by neighboring lakes (NL; comparative data), lake history (LH; temporal data), or by some combination of the two. To answer this question, we used 22+ years of data collected in seven lakes as part of the North Temperate Lakes Long-Term Ecological Research (NTL-LTER) program to predict August epilimnetic depth (Epi), average summer chlorophyll a concentration (Chl), and average annual dissolved organic and inorganic carbon, total nitrogen and phosphorus, calcium (Ca), and sulfate concentrations. Lake characteristics were predicted by NL using linear regression and by a LH using autoregressive time series analysis (AR). Lake history and NL characteristics were combined in an AR analysis with NL included as covariates. We compared models on the basis of their prediction errors (model residual variance [MRV]). MRV ranged from 0.0001 – 1. Ca was the most predictable characteristic (lowest MRV) and Chl and Epi the least predictable. MRV from models that predicted LCs from NL were typically an order of magnitude smaller than models that predicted LCs from LH. Less than 50% of the variables could be predicted from LH. Combining NL characteristics and LH resulted in similar MRV as NL alone. Lakes with similar physical and chemical characteristics were good predictors of each other. For example, dystrophic bogs often predicted annual characteristics of other dystrophic bogs. Additionally, upstream lakes predict characteristics in downstream lakes to which they are hydrologically linked. Overall, our analysis of the effectiveness of comparative and temporal data clearly demonstrates the value of data from NL when predicting annual lake characteristics.