Humans have profoundly altered phosphorus (P) cycling across scales. Agriculturally driven changes (e.g., excessive P‐fertilization and manure addition), in particular, have resulted in pronounced P accumulations in soils, often known as “soil legacy P.” These legacy P reserves serve as persistent and long‐term nonpoint sources, inducing downstream eutrophication and ecosystem services degradation. While there is considerable scientific and policy interest in legacy P, its fine‐scale spatial heterogeneity, underlying drivers, and scales of variance remain unclear. Here we present an extensive field sampling (150‐m interval grid) and analysis of 1438 surface soils (0–15 cm) in 2020 for two typical subtropical grassland types managed for livestock production: Intensively managed (IM) and Semi‐natural (SN) pastures. We ask the following questions: (1) What is the spatial variability, and are there hotspots of soil legacy P? (2) Does soil legacy P vary primarily within pastures, among pastures, or between pasture types? (3) How does soil legacy P relate to pasture management intensity, soil and geographic characteristics? and (4) What is the relationship between soil legacy P and aboveground plant tissue P concentration? Our results showed that three measurements of soil legacy P (total P, Mehlich‐1, and Mehlich‐3 extractable P representing labile P pools) varied substantially across the landscape. Spatial autoregressive models revealed that soil organic matter, pH, available Fe and Al, elevation, and pasture management intensity were crucial predictors for spatial patterns of soil P, although models were more reliable for predicting total P (68.9%) than labile P. Our analysis further demonstrated that total variance in soil legacy P was greater in IM than SN pastures, and intensified pasture management rescaled spatial patterns of soil legacy P. In particular, after controlling for sample size, soil P was extremely variable at small scales, with variance diminished as spatial scale increased. Our results suggest that broad pasture‐ or farm‐level best management practices may be limited and less efficient, especially for more IM pastures. Rather, management to curtail soil legacy P and mitigate P loading and losses should be implemented at fine scales designed to target spatially distinct P hotspots across the landscape.