Lagrangian observation techniques of the ocean have experienced rapid 
advancement in the last few decades. Isopycnal RAFOS floats make it possible to 
accurately track water parcel movements, and it is particularly useful in 
dynamic frontal regions where enhanced mixing and water property exchange can 
take place along tilting isopycnals that significantly intersect geopotential 
surfaces. Equipped with pressure, temperature and dissolved oxygen sensors, the 
floats, as roving current meters and hydrographers, can map out the ocean 
general circulation and property distributions on isopycnal surfaces. As 
Lagrangian particle followers, the floats provide water mass pathways and data 
for particle dispersion and turbulent mixing studies.

Isopycnal floats have been deployed in the Gulf Stream, North Atlantic Current 
(NAC), and Subpolar Front (SPF) regions in separated programs. In particular, 
significant numbers of the floats were deployed in pairs or triplets with the 
aim to study relative dispersion, especially in the recently completed ACCE 
(Atlantic Climate Change Experiment) field program. We present single particle 
and relative dispersion statistics from the above observations. Specifically, we 
discuss the mean and eddy decomposition, shear versus turbulent dispersion, the 
pdf (probability density function) distribution of the turbulent flow, and the 
laws governing the single and relative dispersion at different stages (i.e. 
different time and spatial scales). We discuss our results for different 
regions. For example, eddy kinetic energy (EKE) generally decreases from the 
Gulf Stream downstream to the Iceland Basin, but locally enhanced EKE was found 
over the Mid Atlantic Ridge, and south and west of the Rockall Plateau where the 
NE-flowing NAC/SPF water encounters the Mediterranean water. Compared to the 
Newfoundland Basin, the Iceland Basin has a longer integral time scale but 
smaller turbulent diffusivity for single particle dispersion. We also point out 
that inadequate mean-eddy separation and/or data limitation could result in 
different conclusions.