Thermo-erosional gullies are characteristic features of ice-rich permafrost landscapes. With the Arctic rapidly warming, thermo-erosional gullies are expected to develop more often and expand more rapidly. This has far-reaching consequences, influencing not only the local hydrology through changed drainage pathways, but also leading to ground ice loss, release of greenhouse gases, and changed nutrient and sediment fluxes. The underlying dynamics and processes at gully scales are complex, involving mass-wasting on slopes, erosion, and thaw-related subsidence. They are expected to be influenced by factors such as gully morphology, snow cover, and vegetation. To gain a better understanding of these processes, we used multi-temporal digital surface model (DSM) data. These data were derived photogrammetrically from both uncrewed aerial vehicle (UAV) and airborne images from the modular aerial camera system (MACS), as well as ground-based and airborne lidar data. Our data was acquired at three thermo-erosional gully sites in Northwestern Alaska: two on the Baldwin Peninsula (BAP-B: 2022, 2023, 2024, and 2025; BAP-S: 2021 and 2022) and one on the central Seward Peninsula (CSP-F: 2021, 2022, 2023, and 2024). The BAP-B site is a coastal gully exceeding 10 m incision into the surrounding hillslope terrain. The gullies at BAP-S and CSP-F were formed following lake drainage, recently for BAP-S in 2022 and around 20 years ago for CSP-F. We derived DSMs from the densely overlapping visible-spectrum imagery acquired during drone flights over multiple years using photogrammetric software for structure-from-motion (SfM) image processing, and finely co-registered them. We compared the final DSMs with differential GPS (dGPS) data collected along transects at the two gully sites to evaluate the vertical accuracy of the UAV-derived DSMs. Furthermore, we complemented the time steps with additional DSMs derived from lidar and airborne MACS data by co-registering them to the UAV data. Our results show that UAV-derived DSMs provide reliable elevation data even in difficult tundra settings. They also highlight that ground control points and careful co-registration are vital, particularly for comparisons across multiple time steps. The repeated measurements, including the additional time steps derived from airborne and lidar data, allowed us to detect and monitor the development of a slumping area within the BAP-B gully, and the gully at BAP-S widening. In addition to the gully morphology, we were able to extract the height of early summer snow bank remnants. These datasets provide crucial input for analysing thermo-erosional gully development.