Feasibility of up-the-ramp sampling under variable sky for ground-based spectrographs
Many modern near-infrared instruments employ HAWAII-2RG (H2RG) detectors with integration times that can reach 300-600s. Up-the-ramp (UTR) sampling offers advantages over Fowler sampling, including superior cosmic ray rejection and noise reduction, but requires fitting linear ramps from 30-60 reads. Ground-based K-band sky brightness has been reported to vary by 3-10% on timescales of minutes, potentially introducing systematic errors and compromising photometric accuracy. Additionally, UTR data formats involve higher-dimensional FITS files with larger file sizes impacting observatory operations. We present a feasibility study using the GIRMOS Data Simulator with high-fidelity flux budgets and empirical K-band sky variations estimated, for Mauna Kea, from Gemini-NIRI at 10-20s cadence. Using a Monte Carlo approach we assess whether linear ramp fitting remains viable under variable sky conditions, quantify SNRs and systematic biases, and report nightly data volume estimates. Our results show that, in the H-band, the advantages of the UTR readout hold for read-noise-limited targets placed in the inter-line regions, translating into 3-4% savings in observing time. The K-band inter-line regions do not show significant SNR improvement and can even degrade it due to the dominance of shot-noise generated by the thermal emission of the instrument+telescope system. In these regions, cosmic ray rejection recovers $>$ 98% of events with false positive rates below 0.1%, even under high sky variability. Over the sky emission lines, UTR fitting remains possible but its performance is compromised, both by a degradation in SNR and by a high rate of pixels falsely flagged by the cosmic ray rejection algorithm under highly variable sky. These findings address how ground-based conditions affect UTR implementation in near-infrared spectrographs, with GIRMOS as a concrete case of study.