Background. The major capsid protein VP2 of canine parvovirus type 2 (CPV-2) continues to diversify into antigenic variants that shape diagnostic performance and vaccine relevance. Whether this diversification is driven chiefly by geographically structured lineage expansion, by site-specific molecular adaptation, or by recombination has not been resolved using a single, internally consistent global dataset that also applies rigorous multiple-testing correction and formal population-genetic structure testing. Objectives. We sought to reconstruct the global phylogenetic structure of publicly available complete CPV-2 genomes, test temporal signal and geographic population structure formally, test directly for recombination and for site- and branch-specific selection pressure on VP2 with appropriate correction for multiple testing, characterize population-genetic and haplotype structure, and place candidate residues in the structural context of the solved CPV capsid. Methods. Ninety-five complete CPV-2 genomes from 18 countries (1993-2025) were aligned (MAFFT) and used to infer a maximum-likelihood phylogeny (IQ-TREE2, ModelFinder, 1000 ultrafast bootstrap/SH-aLRT replicates). Temporal signal was assessed by TreeTime root-to-tip regression. The VP2 open reading frame was extracted computationally and validated against the original 1978 prototype sequence resolved by X-ray crystallography (PDB 2CAS). Recombination (GARD) and codon-level selection (FEL, SLAC, FUBAR, MEME, BUSTED, aBSREL) were tested directly in HyPhy 2.5.59, with Benjamini-Hochberg false discovery rate (FDR) correction applied to all site-wise p-values. Population-genetic statistics (Tajimaʹs D, Fu and Liʹs D*/F*, haplotype and nucleotide diversity, Hudsonʹs Fst) were computed in EggLib and scikit-allel, both overall and stratified by genotype and country. Genotype-country association was tested with Pearsonʹs chi-square, a Fisher-Freeman-Halton exact test, and PERMANOVA; isolation-by-distance was tested with a Mantel test. Candidate residues were mapped onto the secondary-structure annotations of the solved capsid structure. Major Results. The phylogeny resolved fully monophyletic, single-country clades for Iraq and Nigeria, three independent Brazilian clusters, and a broadly paraphyletic Chinese population. Root-to-tip regression showed a weak temporal signal (R^2=0.09; rate=3.98x10^-4 substitutions/site/year), consistent with multiple asynchronous introductions rather than one clonally evolving lineage. Genotype distribution (CPV-2c 43%, CPV-2a 35%, CPV-2b 22% of 93 genotyped sequences) was significantly associated with country (chi-square=47.5, p<0.0001; exact test p=2x10^-5; Cramerʹs V=0.59) and confirmed by PERMANOVA (pseudo-F=11.60, p=0.0001). Pairwise Fst was moderate to large both between genotypes (0.19-0.45) and between countries (0.17-0.72); a Mantel test found no significant isolation-by-distance signal (r=0.11, p=0.78), consistent with long-distance introduction rather than gradual diffusion. GARD found no recombination breakpoint improving on a no-breakpoint baseline for VP2. FEL and SLAC identified zero codons under pervasive positive selection against a background of strong purifying selection (dN/dS=0.15). FUBAR and MEME flagged codons 297 and 324 as candidates before correction; after Benjamini-Hochberg FDR correction, neither survived at q<=0.1 (q=0.107 and q=0.95), so this site-level signal is suggestive rather than confirmed. Gene-wide evidence of episodic diversifying selection remained significant (BUSTED, p=0.0064), and aBSREL identified exactly two significant branches, both belonging to the monophyletic Iraqi clade. Population-genetic analysis showed high haplotype diversity (Hd=0.99) against modest nucleotide diversity (pi=0.007/site) with significantly negative Tajimaʹs D (-1.998) and Fu and Liʹs D*/F* (-4.60/-4.19), reproduced within each genotype subset. Biological Interpretation. Multiple, convergent lines of evidence - phylogenetic clade structure, weak clock signal, large Fst with no isolation-by-distance, and significant PERMANOVA partitioning - independently support a model in which global CPV-2 population structure is dominated by discrete, geographically and temporally heterogeneous introduction events rather than by clock-like clonal spread or gradual geographic diffusion. Gene- and branch-level evidence supports episodic diversifying selection localized to a specific introduction event, but site-level attribution to particular codons is corroborating rather than independently conclusive once multiple-testing correction is applied. Conclusions. This integrated phylogenomic, temporal, population-genetic, selection, and structural analysis supports geographically structured introduction-and-expansion as the dominant mode of contemporary CPV-2 diversification, identifies VP2-297 and VP2-324 as candidate but not confirmed sites of diversifying selection, and provides a fully reproducible computational framework for future CPV genomic surveillance.