From Pixels to Mobile Pastoral Households: Scale Mismatch, Exposure Measurement, and Inference in Mongolian Rangeland Research

Lkhamdulam Ganbat , Uyanga Ganbat

Spatially explicit Earth-observation products and machine-learning models increasingly support
detailed monitoring of vegetation productivity and land-surface dynamics across Mongolia. Yet
the spatial unit observed by a satellite, the area actually grazed by a mobile herd, the
administrative unit used in official statistics, and the household unit used in socioeconomic
surveys are rarely identical. This methodological study formalizes that scale mismatch as a
source of exposure-measurement and inferential error in linked rangeland–livelihood research.
It develops a pixel-to-household framework that separates five objects: remotely sensed
environmental signals, ecological interpretation, seasonal grazing locations, mobility-weighted
household exposure, and livestock or household outcomes. The framework identifies four
recurrent mismatches—pixel-to-pasture, residence-to-grazing-area, temporal aggregation, and
administrative-boundary mismatch—and specifies alternative procedures for constructing
exposure weights using GPS tracks, participatory mapping, seasonal-location surveys, livestock
unit weighting, and probabilistic exposure surfaces. Using Mongolian rangeland research as the
application domain, the study shows why strong predictive agreement with a remotely sensed
productivity proxy does not by itself establish household-level exposure or causality. The
resulting design standards emphasize explicit spatial and temporal support, structured
validation, exposure uncertainty, and sensitivity to alternative linkage rules. The framework is
intended as a technical design tool for remote-sensing, ecological, livestock, and socioeconomic
studies in mobile pastoral systems