Soil carbon stocks and greenhouse gas fluxes quantification in selected smallholder farmers' land utilization types of Siaya county, Kenya

Abstract

The up-surging population in sub-Saharan Africa (SSA) has led to the conversion of forests to agricultural land leading to greenhouse gas (GHG) emissions. The resilient land utilization types are key in soil carbon sequestration. There is a vast data gap for the National and regional greenhouse gas (GHG) budget from different smallholders’ land utilization types in Kenya and sub-Saharan Africa (SSA). This study aimed to quantify carbon stock and greenhouse gas fluxes from different land utilization types (LUT) in Siaya. The LUTs considered in the study were agroforestry M (agroforestry with Markhamia lutea), sole sorghum, agroforestry L (agroforestry with Leucaena leucocephalaI), sole maize, and grazing land replicated thrice. Soil samples were collected at a depth of 0-5, 5-10, 10-20, and 20-30cm from different LUTs to determine soil bulk density, organic carbon (SOC) concentration, and carbon stock. PROC ANOVA was used to determine the significant difference in soil bulk density, SOC %, SOC stock. Additionally, GHG data were subjected to analysis of variance (ANOVA) using SAS 9.4 software. Before analysis, the normality of soil GHG fluxes was tested using the Shapiro-Wilk test. DeNitrification-DeComposition (DNDC) model was also used to simulate GHG gases. Soil bulk density varied significantly (p<0.05) across the LUTs and soil depths with a range of 1.30 and 1.60gcm-3 under Agrofestry M and grazing land, respectively, at 0-5cm depth. A significant difference (p<0.0001) in SOC concentration was observed with high SOC concentration under Agroforestry M of 30.14gCkg-1 at 0-5cm depth than all the other treatments and low SOC (8.4g Ckg-1) in sole maize. Soil organic carbon stocks significantly (p<0.0001) varied across LUTs and depths. There was high carbon stock in agroforestry M (19622 kg C ha-1) and grazing land (20069.7kgCha-1) at 0-5 cm. Soil GHG fluxes significantly varied across the LUTs methane p<0.05, Carbon diode p=0.05, and nitrous oxide p=0.05. The cumulative methane fluxes ranged from -0.35 kg CH4-C ha-1 in grazing land highest -1.05 kg CH4-C ha-1 sole maize. Low soil CO2 emissions under sole maize, 6510 kg CO2-C ha-1, and the highest under grazing land were observed, 14401 kg CO2-C ha-1. The results showed the lowest soil N2O fluxes under grazing land, 0.69 kg N2O-Nha-1, and the highest under agroforestry L 2.48 kg N2O-N ha-1. The model showed a high degree of fit in simulating daily soil temperature, soil moisture, and soil N2O emissions. The model depicted good results during simulation of soil moisture; root mean square error (RMSE) <5, 2% < normalized root means square error (nRMSE) <15.54%, 0.86< modelling efficiency (NSE) (NSE) <0.99, 0.03< coefficient of determination (R2) <0.97 and (index of agreement) d < 0.99. Daily soil temperature; 0.08 <RMSE< 1.33, from 0.3%< nRMSE< 5.9%, from 0.27< NSE< 0.99, from 0.12< R2 <0.99. daily soil N2O; 0.002 <RMSE< 0.006, 0.45% <nRMSE< 2.48%, NSE=0.99, from 0.5< R2 <0.9, 0.98 <d< 0.99. The DNDC model showed relatively good results in simulating soil moisture, temperature, and N2O fluxes. The model showed relatively fitted N2O emissions peaks following the precipitations across all the LUTs. The model had good to excellent performance in simulating the N2O fluxes. The drivers of soil GHG were soil bulk density, soil organic carbon, soil moisture, clay content, and root production during GHG simulation and estimation. The results thus help fill the gaps in the national and regional data on carbon and emissions budgets.