Objective By comparing the responses of respiration at different observation scales to environmental factors, especially the similarities and differences in temperature and water responses, we aimed to understand the impact mechanism of multi-scale respiration and improve the simulation of cross observation scale respiration models.

Method This study selected a typical Artemisia ordosica community in Yanchi, Ningxia of northwestern China as a research site, and conducted continuous observation of Artemisia ordosica leaf, soil, and ecosystem respiration (i.e., R_l, R_s and R_e) characteristics from May to October, 2022 using fixed plot measurement and in situ continuous monitoring. By fitting nonlinear and linear equations to the relationship between respiration and temperature and moisture content, we can better understand the seasonal dynamic characteristics of multi-scale respiration and its environmental influencing factors.

Result (1) The study found that during the observation period, the maximum value of R_l was 5.96 μmol/(m²·s), which occurred in July and was primarily regulated by temperature (R² = 63.5%) with a temperature sensitivity (Q₁₀) of 1.48. The maximum values of both R_s and R_e occurred in August, at 2.94 μmol/(m²·s) and 4.07 μmol/(m²·s), respectively, and their seasonal changes were regulated by moisture (R² of 44.4% and 50.9%, respectively), with Q₁₀ values of 1.23 and 1.08. (2) The explanatory power of temperature moisture bivariate empirical model for R_l, R_s and R_e was limited compared with the univariate model, with an average R² increase of 0.09, 0.05, and 0.02, respectively. (3) Water availability was the key factor influencing whether the temperature sensitivity of different observation scale respiration tends to be consistent. When soil moisture conditions were poor (relative extractable soil water, W_RE < 0.4), there were significant differences in the response of R_l, R_s and R_e to temperature, with Q₁₀ values of 1.34, 0.63 and 0.84, respectively; when the soil moisture conditions were sufficient (W_RE ≥ 0.4), the response of R_l, R_s and R_e to temperature tended to be consistent, with Q₁₀ values ≈ 1.8.

Conclusion Our study emphasizes the differences in the regulatory factors of seasonal changes in respiration at different observation scales, while the bivariate model has a limited role in improving the accuracy of respiration simulation at different observation scales. Fully considering the differences in observation scale and water conditions is the key to accurately simulating respiration in arid or semiarid areas in the future.