Pedosphere 27(5): 877--889, 2017
ISSN 1002-0160/CN 32-1315/P
©2017 Soil Science Society of China
Published by Elsevier B.V. and Science Press
| A novel evolutionary genetic optimization-based adaptive neuro-fuzzy inference system and geographical information systems predict and map soil organic carbon stocks across an Afromontane landscape |
Kennedy O. WERE1,2 , Dieu TIEN BUI3, Øystein Bjarne DICK1, Bal Ram SINGH4 |
1Department of Mathematical Sciences and Technology, Norwegian University of Life Sciences, P. O. Box 5003, NO-1432 Ås (Norway)
2Kenya Agricultural and Livestock Research Organisation, Kenya Soil Survey, P. O. Box 14733-00800, Nairobi (Kenya)
3Department of Business Administration and Computer Science, Faculty of Arts and Sciences, University College of Southeast Norway, NO-3800 Bøi Telemark (Norway)
4Department of Environmental Sciences, Norwegian University of Life Sciences, P. O. Box 5003, NO-1432 Ås (Norway) |
| Corresponding Author:Kennedy O. WERE |
| ABSTRACT |
| Soil organic carbon (SOC) pool has the potential to mitigate or enhance climate change by either acting as a sink, or a source of atmospheric carbon dioxide (CO2) and also plays a fundamental role in the health and proper functioning of soils to sustain life on Earth. As such, the objective of this study was to investigate the applicability of a novel evolutionary genetic optimization-based adaptive neuro-fuzzy inference system (ANFIS-EG) in predicting and mapping the spatial patterns of SOC stocks in the Eastern Mau Forest Reserve, Kenya. Field measurements and auxiliary data reflecting the soil-forming factors were used to design an ANFIS-EG model, which was then implemented to predict and map the areal differentiation of SOC stocks in the Eastern Mau Forest Reserve. This was achieved with a reasonable level of uncertainty (i.e., root mean square error of 15.07 Mg C ha-1), hence demonstrating the applicability of the ANFIS-EG in SOC mapping studies. There is potential for improving the model performance, as indicated by the current ratio of performance to deviation (1.6). The mapping also revealed marginally higher SOC stocks in the forested ecosystems (i.e., an average of 109.78 Mg C ha-1) than in the agro-ecosystems (i.e., an average of 95.9 Mg C ha-1). |
| Key Words: artificial neural networks, carbon sequestration, climate change mitigation, digital elevation model, digital soil mapping, Eastern Mau Forest Reserve, fuzzy logic |
| Citation: Were, O., Tien, B., Bjarne, D. and Ram, S. 2017. A novel evolutionary genetic optimization-based adaptive neuro-fuzzy inference system and geographical information systems predict and map soil organic carbon stocks across an Afromontane landscape. Pedosphere. 27(5): 877-889. |
|
View Full Text
|
|
|
|
