Studies of DNA methylation in Arabidopsis have rapidly advanced from the analysis of a single reference accession to investigations of large populations. The goal of emerging population studies is to detect differentially methylated regions (DMRs) at the genome-wide scale, and to relate this variation to gene expression and phenotypic diversity. Whole-genome bisulfite sequencing (WGBS-seq) has established itself as a gold standard in DNA methylation analysis due to its high accuracy and single cytosine measurement resolution. However, scaling up the use of this technology for large population studies is currently not only cost prohibitive but also poses nontrivial bioinformatic challenges. If the end-point of the study is to detect DMRs at the level of several hundred base pairs rather than at the level of single cytosines, low-resolution array-based methods, such as MeDIP-chip, may be entirely sufficient. However, the trade-off between measurement accuracy and experimental/analytical practicality needs to be weighted carefully. To help make such experimental choices, we conducted a side-by-side comparison between the popular dual-channel MeDIP-chip Nimblegen technology and Illumina WGBS-seq in two independent Arabidopsis lines. Our analysis shows that MeDIP-chip performs reasonably well in detecting DNA methylation at probe-level resolution, yielding a genome-wide combined false-positive and false-negative rate of about 0.21. However, detection can be susceptible to strong signal distortions resulting from a combination of dye bias and the CG content of effectively unmethylated genomic regions. We show that these issues can be easily bypassed by taking appropriate data preparation steps and applying suitable analysis tools. We conclude that MeDIP-chip is a reasonable alternative to WGBS-seq in emerging Arabidopsis population epigenetic studies.