Carbon stocks variability in Agro Ecosystems along an Altitudinal Gradient: a case study of Taita Hills, Kenya

Abstract

This study quantifies the effect of land cover change and seasonality on soil organic carbon and carbon dioxide emissivity. It takes to account the coupled inter-relationships with other ecological factors such as temperature and moisture. Next, the study assesses how topographic and ecological factors drive spatial soil nutrient stock variations and quantifies the observations required to discriminate stock detection in the mountain ecosystem. Thereafter, the study derives and evaluates modeling frameworks that integrate remote sensing, geography information systems and field measured ecological data to maximally explain soil nutrient stocks variation from terrain and seasonal dimensions. In order to address the objectives and answer research questions, activities described in this thesis combined advanced tools with renowned (geo) statistical methods. The results present a simple yet effective approach to establish baseline soil gas emissions and nutrient stocks, taking into account limitations posed by terrain accessibility and resources availability.The results of a one year chamber based soil CO2 sampling investigation within the study transect show seasonal and spatial soil CO2 emission patterns were most significantly explained by rainfall and land surface temperature patterns within the five land cover types assessed. Specifically, forest and agroforestry land use situated from 1400 to 2200 m contributed to the highest mean monthly CO2 fluxes compared to the shrub and cereal croplands mainly below 1400 m elevation. Similarly, the mean monthly soil CO2 relationship with ambient temperature indices were highly variable below 1400 m elevation compared to transect areas beyond this range. Higher spatial and temporal soil CO2 variability was derived in regression models combining altitude to either land surface temperature or rainfall compared to those solely using altitude. Soil organic carbon (soil OC) and total nitrogen (TN) stocks assessments show suppressed but positive linear relationship between altitude and either soil OC (R2= 0.30; p-value < 0.05) or TN (R2 = 0.35; p-value < 0.05) that varies within altitude categories. Moreover, nutrient stocks were comparable and lower in croplands and agro-forestry systems in contrast to nutrient rich natural land cover systems. Altitude, soil temperature and soil water were significant controls for soil OC and TN stocks explaining > 30 % and > 80 % variation in the low and high altitude ranges respectively. Detection of carbon and nitrogen stock varied with altitudinal ranges and depended on innate soil nutrient stocks. Derived landscape position and terrain ruggedness classification schemes were used to assess spatial soil OC and TN stocks and revealed subtle differences between land surface and intrinsic soil properties. Landscape position explained lower plot soil OC and TN stocks variation (CV < 0.5) compared to terrain ruggedness (CV > 0.5). Bulk density was a dominant soil OC predictor in the landscape position scheme, with valley (r2 = 0.74) and plateau (r2 = 0.77) models explaining higher variation by including slope and soil moisture. Finally, mixed soil OC and TN stocks patterns were revealed in the conventional wet (March-April-May, MAM and October-November-December, OND) and dry (January-February, JF and June-July-August-September, JJAS) seasonal evaluation. Significant inter-seasonal mean monthly soil OC and TN stocks variations were observed in maize and forest but were absent in avocado and shrub land cover plots. Seasonal mean monthly soil % C and % N concentrations revealed an increasing trend from low to high altitude categories, with large interseasonal coefficients of variation. The pattern is revealed for instance, in the more than 50 % change in soil % C concentration from MAM to JJAS seasons at 1300 - 1800 and 1800 - 2300 m elevation ranges.Soil OC stock revealed the highest statistical seasonal co-relationship with daytime and nighttime land surface temperature, soil water filled pore space and soil pH. Prediction models i.e soil C % predicted using Inverse Distance Weighting and using ordinary kriging, soil C % co-kriged with soil pH and with WFPS, compared favorably in their seasonal MAM (from 0.5 to 12 %) and JJAS (from 0.5 to 14 %)predictions. However, the models predicted varied inter-seasonal changes (from -5 to 1 % C) within different areas of the study transect. The study concludes that altitude driven land cover and topographic micro-climates contributed differentially to seasonal - spatial soil CO2 fluxes and nutrient heterogeneity in the Taita Hills. The baselines established in this study can be adopted for other environments bearing similar land cover and altitudinal characteristics within East Africa Afromontaine ecosystems. The framework(s) used for this study can similarly be adopted for comparative evaluation, and can be improved through use of rapidly advancing high resolution digital elevation models. The results from this study are an useful input to national carbon inventory exercises. They can also serve as a guide to design of rehabilitation, land health surveillance and soil fertility improvement options for use by smallholders, land