Farm typology and spatial variability of selected soil fertility parameters on selected small-holder farms in KwaZulu-Natal province, South Africa.

Abstract

Diversity of resource endowment, soil and climatic conditions may affect the level of management and productivity and soil fertility in small-holder farms. The objectives of this study were to (i) develop farm typologies, (ii) assess fertility gradients, and (iii) map spatial variability of soil fertility in small-holder farms of uMbumbulu and Msinga regions in KwaZulu-Natal province, South Africa. To obtain data for the identification of farm typologies, a detailed open-ended questionnaire was used with topics including socio-economic attributes, local crops grown, soil amendments, management practices, labour, crop residue management, farmers perceptions and production constraints. The questionnaire was administered to fifty farmers per region. The data which had Kaiser-Meyer-Olkin (KMO) measure values of 0.67 and 0.51 for uMbumbulu and Msinga respectively, qualified for Principal component analysis (PCA). Three PCs which had significant eigenvalues of >1, provided key factors that determine the farm typologies, namely land size, livestock ownership, income from farming and external income. Multiple correspondence analysis (MCA) and cluster analysis were used to analyse quantitative and qualitative data, and variables and aggregate farms into clusters according to production, socioeconomics, and demographics. Three farm topologies were identified, namely (i) resource-endowed farms which have large land and profit from farming (type I), (ii) the middle-resourced group (type II), which is neither poor nor rich, and (iii) Poor resource groups (type III) with limited to no resources at all and have small land holdings and minimum profits from farming. For fertility gradients and mapping, soils were sampled from 0 – 20 cm depth, using a sampling interval of 5×5m and analysed for fertility parameters. There were no fertility gradients observed between homefields and outfields for both sites. Mapping was done only in uMbumbulu site with descriptive statistics (mean, standard deviation, covariance, skewness, and kurtosis) tested for normality to be used for kriging, and only the spherical model was tested in this study using R-Studio. For geo-statistics (Lag size, sill, and nugget) for semiviriograms produced was done using ArcMap-GIS as well as the maps. For type I farms the spatial dependency was strong (< 25%) for most variables tested (pH, total carbon, calcium, magnesium, potassium, and Clay %), while type III had a variety of spatial dependency from pH and clay % were weak (<75%), Ca and total carbon moderate (25-75%) to phosphorus, magnesium, potassium, and acid saturation strong (<25%). Overall implications of these maps can be very useful in targeting specific areas of poor or rich fertility and fertiliser recommendation, which is more economically viable to small-holder farmers to put in what is needed.