In heart failure (HF), dysregulated cardiac ryanodine receptors (RyR2) contribute to the generation of diastolic Ca2+ waves (DCWs), thereby predisposing adrenergically stressed failing hearts to life-threatening arrhythmias. However, the specific cellular, subcellular, and molecular defects that account for cardiac arrhythmia in HF remain to be elucidated. Patch-clamp techniques and confocal Ca2+ imaging were applied to study spatially defined Ca2+ handling in ventricular myocytes isolated from normal (control) and failing canine hearts. Based on their activation time upon electrical stimulation, Ca2+ release sites were categorized as coupled, located in close proximity to the sarcolemmal Ca2+ channels, and uncoupled, the Ca2+ channel-free non-junctional Ca2+ release units. In control myocytes, stimulation of β-adrenergic receptors with isoproterenol (Iso) resulted in a preferential increase in Ca2+ spark rate at uncoupled sites. This site-specific effect of Iso was eliminated by the phosphatase inhibitor okadaic acid, which caused similar facilitation of Ca2+ sparks at coupled and uncoupled sites. Iso-challenged HF myocytes exhibited increased predisposition to DCWs compared to control myocytes. In addition, the overall frequency of Ca2+ sparks was increased in HF cells due to preferential stimulation of coupled sites. Furthermore, coupled sites exhibited accelerated recovery from functional refractoriness in HF myocytes compared to control myocytes. Spatially resolved subcellular Ca2+ mapping revealed that DCWs predominantly originated from coupled sites. Inhibition of CaMK∏ suppressed DCWs and prevented preferential stimulation of coupled sites in Iso-challenged HF myocytes. These results suggest that CaMK∏-(and phosphatase)-dependent dysregulation of junctional Ca2+ release sites contributes to Ca2+-dependent arrhythmogenesis in HF.