Modeling soil respiration: Seasonal variability and drivers in pine and broad-leaved forests of the lower Himalayas

• Published in: Trees, Forests and People (Online) Vol. 20; p. 100804
• Main Authors: Sivaranjani, S., Sarkar, Mriganka Shekhar, Panwar, Vijender Pal, Pandey, Rajiv, Mishra, Arun Pratap, Rathnayake, Upaka
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.06.2025
• Subjects: Environmental variables; Pinus roxburghii; Shorea robusta; Soil respiration; Soil respiration; Pinus roxburghii; Environmental variables; Shorea robusta
• ISSN: 2666-7193; 2666-7193
• DOI: 10.1016/j.tfp.2025.100804

•Soil respiration rates differ significantly between Sal and Chir-Pine forests across seasons.•Soil moisture is the primary driver of Rs in Sal, while rainfall dominates in Chir-Pine.•The findings improve carbon pool projections, which are important for climate change and forest management. Soil respiration (Rs) is the largest source of carbon dioxide emissions from terrestrial ecosystems. While numerous studies have examined its environmental controls, significant knowledge gaps remain regarding the complex interactions between biotic and abiotic factors regulating Rs. These uncertainties hinder the accuracy of model predictions, limiting our ability to assess ecosystem carbon dynamics under changing environmental conditions. This study hypothesizes that, soil properties, microclimatic and environmental variables influence Rs, with variations across forest types. To explore this, the study aims to quantify Rs in two distinct forests and predict its relationship with environmental, microclimatic, and soil characteristics in S. robusta and P. roxburghii forests in the lower Indian Himalayas. Initially, we collected field data containing soil respiration, soil properties and environmental factors. The ANOVA analysis revealed that Rs rates across different seasons in Sal (F = 100.9, P < 0.05) and Chir-Pine forests (F = 49.89, P < 0.05) were found significantly different. Subsequently, we employed machine learning techniques with various training strategies to improve model accuracy and analyze the relationship between soil respiration and environmental factors. The RF machine learning algorithm was applied to estimate the relationship between Rs and other properties. The results showed that Random Forest model in Sal Forest achieved the lowest RMSE (2.11) and MAE (1.38), suggesting it had the best predictive performance than the others. The most influential parameter influencing Rs rates in Sal was Soil moisture, followed by Soil Temperature and Rainfall. Similarly, Chir-Pine Forest also performed best in the RF model with the lowest RMSE (1.455) and MAE (1.011), as well as the highest R2 value (0.363). In Chir-Pine, the most influential parameter was RF followed by ST and SM. The present study concluded that combining forest-specific properties with climatic parameters may provide more robust predictions of Rs. The findings will enable the precise future accounting of temporal and spatial changes in carbon pools and atmospheric CO2 concentrations and their evolving trajectories concerning species composition in forests under climate change.