In western Kenya, soil nutrient depletion is one of the main problems that has led to declining crop yield. Agricultural intensification through the judicious application of fertilizers has been considered amongst mitigation options for these smallholder farming systems with an average land size of less than 3.0 ha. The blanket fertilizer recommendations used in this region, have led to poor response to the fertilizer applied and low nutrient use efficiency. These recommendations do not take into account the spatial variability occurring at the local level across the smallholder landscape. Furthermore, the methods used to diagnose soil nutrient constraints are inefficient, because they do not take into account the spatial extent to which the nutrient deficiencies occur. Digital Soil Mapping (DSM) technique and the Population-Based Farm Survey (PBFS) approach are promising strategies that can help address this problem though they have not been fully exploited for smallholder farming systems. The main objective of this study was to develop and test nutrient management strategies that could be used to improve fertilizer recommendation using the DSM technique and the PBFS approach. The approach was tested to provide site-specific nutrient diagnostics and provide management recommendations in heterogeneous smallholder farming systems. First, evaluation of Fertilizer Response (FR) – a response ratio, and Agronomic Nutrient Use Efficiency (NAE) was conducted using fertilizer trial data. Meta-analysis technique was employed to identify key factors that influence FR and N-AE in smallholder farming systems. The results indicated soil, climate, and management factors could explain only small amounts (< 30 %) of variation in FR and N-AE. Soil pH, phosphorus (P), texture, and rainfall had significant (P <0.001), but low levels of power in explaining variation in FR and N-AE. This implied that strategies for refining the blanket fertilizer recommendations should include soil-based information, but soil testing needs to be accompanied by nutrient response trials. Secondly, the utility of using the DSM technique was explored, to determine the optimum scale of using digital soil maps, relevant to nutrient management for maize farming systems. A farm survey was conducted and data on soil properties; soil pH, Soil Organic Carbon (SOC), Total Nitrogen (TN), xiv Potassium (K), Phosphorus (P), Cation Exchange Capacity (CEC), Calcium (Ca), and Magnesium (Mg), Grain Yield (GY) and Plant Biovolume (BV) were collected. Data on the soil properties and crop responses (GY and BV) were analyzed using Step-wise Multiple Linear Regression (SMLR) analysis and geostatistical techniques. The results showed high variability in GY, with 32 % of the observed variation being accounted for by the underlying soil properties. SOC was identified as the key driver of crop response to fertilizer application in the study area. Moderate spatial dependencies for SOC with an effective distance of 523 m were observed. The lower nugget value (0.0542) was indicative of short-distance spatial variability in soil properties. A threshold scale of 250 m was proposed, below which, local growing conditions within the study area were captured, implying that a soil nutrients map with a resolution < 250 m would capture the local variability. Lastly, a sampling approach on a population-based survey of smallholder maize fields was tested to diagnose soil nutrient constraints rather than the conventional agronomic trials. Soil test values were established using Cate-Nelson Analysis (1978) for NPK, which were used to define cases on nutrient constraints. In these study, three aspects are considered; evaluation of FR and N-AE to guide nutrient management strategies, the use of DSM techniques to provide fertilizer recommendations at a refined spatial scale, and utility of PBFS for diagnosis of nutrient limitations in smallholder farming systems. The main finding of the study includes: (i) FR and N-AE were highly variable in smallholder maize fields of western Kenya, (ii) SOC was the key soil factor that captured local spatial variability on farms. Thus, 250 m was the optimum soil sampling distance for nutrient management based on the spatial range of SOC. This study demonstrated that soil nutrient maps are useful tools, which can be implemented in strategies aimed at a refined fertilizer recommendation across SSA. The utility of DSM and the new PBFS approach has the potential for providing site-specific diagnostics to guide nutrient management decisions. Successfully developing such an integrated soil-based diagnostic system is warranted, and the wider application will be instrumental for refining fertilizer recommendation across maize smallholder agroecosystem systems.

show more