This proposal addresses NOAA-OAR-CPO-2021-2006389, Competition #5 “Innovative Ocean Dataset/Product Analysis and Development for support of the NOAA Observing and Climate Modeling Communities”. Our focus is on assessing observational strategies for the Tropical Pacific Observing System (TPOS) in the context of data assimilation for reanalysis and model initialization. Core elements of the TPOS mission include observing ENSO, advancing scientific understanding of ENSO variability, and identifying effective observational strategies for skillful weather and climate prediction on subseasonal to seasonal (S2S) timescales. To support TPOS, we propose a fundamentally dynamics-based approach to observing system design within a dynamically and kinematically consistent adjoint-based (4DVar) data assimilation framework. Our research proceeds along three main thrusts: (i) We will leverage a now-existing medium-resolution (1/3°) Tropical Pacific Ocean State Estimate (TPOSE) which will be augmented to a higher-resolution, 1/6° product (TPOSE_HR). This state estimate will improve the representation of currents and Tropical Instability Waves (TIWs) and provide a baseline for three study periods, covering an Eastern Pacific El Niño (2015/16), a Central Pacific El Niño or Modoki (2009/10), and an Eastern Pacific La Niña (2010/11). These baseline TPOSE_HRs will enable us to address a number of questions regarding the impact and best practices of assimilating emerging TPOS sensors. Because our adjoint-based framework provides sensitivities to the atmospheric state, we will be able to assess how existing or hypothetical sensors will constrain air-sea fluxes. (ii) In a second thrust, we will conduct dynamical attribution studies for a range of quantities of interest (QoIs) by means of a Green’s functions approach, with the detailed sensitivity kernels provided by the adjoint. The choice of QoIs is guided by TPOS2020 goals, in particular with respect to ENSO monitoring, understanding, and prediction. Recognizing that ENSO diversity is governed by a complex interplay of dynamical processes and that TPOS serves to inform a diverse range of weather and climate phenomena, we will assess the degree to which sensitivity patterns for diverse QoIs to be tested exhibit redundancies. Sensitivity maps, interpreted as “heat maps” quantify variables and regions that impact the QoI as a function of lag time. These maps provide valuable information regarding predictable anomaly propagation and instrument placement. (iii) In a third thrust, use of derivative information will be further refined by quantifying how existing or hypothetical observational assets may reduce uncertainties in the QoIs chosen under (ii). This will be achieved within the context of Hessian-based uncertainty quantification (UQ) and optimal observing network design. The ability to conduct such calculations sets our framework apart from conventional studies based upon observing system [simulation] experiment (OS[S]E) approaches.