Wednesday, November 7, 2007
328-4

Spatiotemporal Mapping of Total Carbon Stock in Agroforestry Systems of Sub-Saharan Africa.

Antonio Querido1, Russell Yost1, Sibiry Traore2, Mamadou D. Doumbia3, Richard Kablan1, Hamidou Konare1, and Abdramane Ballo3. (1) Tropical Plant and Soil Sciences, University of Hawaii, 3190 Maile Way St. John 102, Honolulu, HI 96822, (2) ICRISAT, Bamako, Mali, (3) Institut d'Economie Rurale, BAMAKO, BP: 262, Mali

The Sudano-sahelian region supports a diverse agro-ecosystem comprising plant species essential for the livelihood of humans as well as for the global carbon cycle.  This region of the African continent remains among the least studied terrestrial ecosystems.  Severe droughts coupled with unsustainable exploitation of woody plants as well as overgrazing have contributed to accelerated land degradation and marginalization of a substantial part of the region.  Nevertheless, these agroecosystems have potential for sequestration and storage of carbon dioxide from the atmosphere to help offset global CO2 emissions.  The buildup of soil organic carbon also improves the quality of the soils, increases productivity and sustainability and hence increases food security for Africans.  The objective of this study was to map the spatial and temporal distribution of organic soil carbon in mixed agro-forestry systems in Mali-West Africa and provide an estimate of the soil carbon along with its reliability expressed as variance, mean squared error (MSE) and mean error (ME). Soil samples, collected from two depths (0-20 and 20-40cm) over a 6 year period, were analyzed for carbon content and other properties. The Bayesian Maximum Entropy (BME) spatiotemporal geostatistics procedure (Christakos, 2000, 2002), was used to integrate soil carbon content (0-20cm) with class interval data (clay, carbon 20-40cm) for improved prediction.  The prediction methods were compared using bias (ME) and prediction error (MSE).  The MSE showed that cokriging using 20-40cm data (MSE=0.0225) failed to improve the estimate of soil carbon (0-20cm) when compared with ordinary kriging (MSE=0.0176).  Simple and multivariate BME methods yielded a more reliable prediction (MSE=0.0161, 0.0150), respectively. BME predictions were characterized by less bias (0.029, 0.0033), respectively than kriging and cokriging (0.0407, 0.0602) respectively.  The results of repeated measures analysis using SAS PROC MIXED indicated a significant increase in soil organic carbon over time.