Recurring outbreaks caused by endemic pathogens present a significant challenge to agricultural systems. Thus, understanding the risk factors involved in fueling the continued outbreaks and pathogen evolution is a priority. Isolate genome sequencing efforts have largely guided our past understanding of the pathogen population structure. However, this approach can overlook the importance of co-occurring pathogenic genera, species, or even lineages of the same pathogenic species in shaping disease dynamics. Here, we aimed to monitor pathogen population dynamics at a finer resolution, tap into the genetic variation existing and emerging within and across fields, and understand the determinants of this diversity in theXanthomonas-tomato pathosystem. Using strain-resolved metagenomics, we found that pathogen heterogeneity with multiple co-occurring lineages is common, although accompanied by differential lineage dynamics and that higher disease severity is associated with higher pathogen diversity. Considering these observations, we used response-specific regression models to investigate the roles of environmental variables on driving these differential dynamics. We find that climatic fluctuations can modify the endemic disease risk and that the pathogen adapts to these climatic shifts by maintaining diversity of co-occurring lineages, each with a varying fitness contribution. We identified signatures of seasonal adaptation by monitoring genome-wide allele frequencies in pathogen. The observation of seasonal oscillations in allelic frequencies depicted evidence for fluctuating selection contributing to the patterns of genetic variation. We also identified positively selected loci under parallel evolution such as type VI secretion system genes and TonB-dependent receptors, which may explain the nature of selection pressures experienced by the pathogen. The findings from this study reveal fitness strategies adopted by endemic pathogens and how pathogens can evolve under the changing climate. Our high-resolution combinatorial approach exploiting a series of sequence data and metadata types and analysis tools, is general enough to finely investigate eco-evolutionary dynamics of pathogens at large scales in diverse case-studies concerning plant health, but also animal and human health.