The x-ray imaging performance of an indirect flat panel detector (I-FPD) is intrinsically limited by its scintillator. Random fluctuations in the conversion gain and spatial blur of scintillators (per detected x-ray) degrade the detective quantum efficiency (DQE) of I-FPDs. These variations are often attributed to depth-dependence in light escape efficiency and spatial spread before detection. Past investigations have used theoretical models to explore how scintillator depth effects degrade DQE(f), however such models have not been validated by direct measurements. Recently, experimental methods have been developed to localize the depth of x-ray interactions in a scintillator, and image the light burst from each interaction using an ultra-high-sensitivity optical camera. This approach, referred to as depth-localized single x-ray imaging (SXI), has enabled direct measurements of both depth-dependent and fixed-depth variations in scintillator gain and spatial resolution. SXI has been used recently to measure depth-dependence in the average gain and modulation transfer function (MTF) of columnar CsI:Tl, which is the scintillator-of-choice for medical I-FPDs. When used in a depth-dependent cascaded linear system model, these SXI measurements accurately predict the presampling MTF(f) of CsI:Tl-based I-FPDs as measured using the slanted-edge method. However, such calculations underestimate the CsI:Tl noise power spectrum (NPS), and thereby overestimate its DQE when compared to conventional measurements. We hypothesize that some of this discrepancy is caused by fixed-depth variations in CsI:Tl spatial resolution, which are not considered in current models. This work characterizes these variations directly using depth-localized SXI and examines their impact on scintillator DQE(f).