Climate variability represents a growing challenge for livestock systems; however, its indirect economic effects remain insufficiently understood, particularly in data-scarce contexts. This study evaluates whether satellite-derived bioclimatic indices propagate into short-term variability of livestock-related sales from a digital agriculture perspective. Weekly commercial records from two geographically proximate livestock branches in Ecuador were integrated with meteorological data provided from NASA POWER to compute the Temperature Humidity Index (THI). A basal temperature index, defined as a four-week moving average of THI, and a corresponding thermal anomaly were derived in order to represent both cumulative and short-term thermal conditions. Linear time series models incorporating exogenous variables (ARIMAX) and a non-linear machine learning approach (Random Forest) were employed using lagged climatic and economic features. The results showed that linear models had limited explanatory capacity, indicating that short-term sales variability was primarily driven by market dynamics and logistical processes rather than direct climatic forcing. While the Random Forest model achieved better predictive performance, this was mainly due to its ability to capture systemic inertia and autoregressive structure in the sales series; climatic variables only provided a secondary, indirect signal. These findings highlight the value of artificial intelligence in identifying weak and delayed climate-related patterns in aggregated commercial indicators and support of satellite-based climate data in market-level decision making in livestock supply chains where animal-level measurements are unavailable.
Aguirre-Munizaga et al. (Tue,) studied this question.