Estimating the soil respiration under different land uses using artificial neural network and linear regression models

TLDR: In this article, the authors used an Artificial Neural Network (ANN) and Linear Regression Methodology (LRM) to predict soil respiration in arid areas of Iran using 150 data points obtained from soil samples collected from the surface to 20 cm of depth under different land use categories.

Abstract: Soil respiration is a biological process in microbes that convert organic carbon to atmospheric CO2. This process is considered to be one of the largest global carbon fluxes and is affected by different physicochemical and biological properties of soil, land use, vegetation types and climate patterns. Soil respiration recently received much attention, and it could be measured in two states basal respiration (BR) and substrate induced respiration (SIR) which together gives a good representation of the general soil microbial activity. The aim of this study was to estimate the BR and SIR of 150 data points obtained from soil samples collected from the surface to 20 cm of depth under different land use categories using the Artificial Neural Network (ANN) and Linear Regression Methodology (LRM). This study is bringing data from an arid area, and there is little information on this issue. Soil samples were chosen from three provinces of Iran, with humid subtropical and semi-arid climate patterns. In each soil sample a variety of characteristics were measured: soil texture, pH, electrical conductivity (EC), calcium carbonate equivalent (CCE), organic carbon (OC), OC fractionation data e.g. light fraction OC (LOC), heavy fraction OC (HOC), cold water extractable OC (COC) and warm water extractable OC (WOC), population of fungi, bacteria, actinomycete, BR and SIR. Our goal was to use the most efficient ANN-model to predict soil respiration with simple soil data and annual precipitation (AP) and mean annual temperature (MAT) and compare it with LRM. Our results indicated that for an ANN model containing all the measured soil parameters (14 variables), the R2 and RMSE values for BR prediction were 0.64 and 0.05 while these statistical indicators for SIR obtained 0.58 and 0.15, respectively; whereas the addition of AP and MAT data to this model (16 variables) caused a decrease in statistical indicators. When the R2 and RMSE values of the BR-ANN and SIR-ANN predicted using an ANN model with only 7 variables (including OC, pH, EC, CCE and soil texture) they were estimated to be 0.66, 0.043 and 0.52, 0.16, respectively. Overall, LRM in comparison to ANN had a lower R2. Therefore, the results show that ANN modeling is a reliable method for predicting soil respiration, even when based on easy to measure data. Our results revealed that highest and lowest BR and SIR were recorded in rice paddy soils and saline lands, respectively. In total, soil respiration (BR: 0.09 vs 0.06 and SIR: 0.46 vs 0.32 mg CO2 g−1 day−1) was higher in agricultural land compared to natural covered land.