AFFILIATION : Mekerere University - Uganda THEME : Handling Spatial Sampling Points Bias in mCROPS Data in Uganda SUMMARY
The principal aim of the mCROPS research project is to develop a low-cost, handheld, in-field-based tool for diagnosing viral crop diseases in cassava crops and a lot of progress has been made in this area. However, the handling of spatial uncertainty has received very little attention. We must study raw mCROPS data relative to sample size and distribution, representativeness, sampling bias, temporal factors, arrangement of samples and edge effects, and collected measures. Using GIS and spatial analysis techniques, we studied whether the sampling of these study sites was biased or not. First, we created a map showing the locations of mCROPS study sites (n = 80,871). This map was used to visualize and analyze spatial patterns, as well as, to detect and measure any potential spatial sampling bias in the raw mCROPS data. Given that potential biases in study sites were observed in linear-like shape and patterns, the cassava crop samples were most likely measured in close proximity to rivers and roads. We also quantified the intensity of sampling of each study site inside the 100 meters and compared it with sampling intensity in farther areas (outside 100 meters). Statistical tests were employed to determine if the spatial distribution of these sampling records were significantly different from random distributions. Our findings show that the location and intensity of collecting were heavily influenced by road and river accessibility. Sampling localities show dense, significant aggregation around and along rivers and roads. Thus, such study sites should be interpreted with caution proportional to the bias found in the raw mCROPS data. We argue that study sites of cassava crop samples setting analyses require (1) statistical tests to detect these biases, and (2) data treatment to reflect cassava crop distribution rather than patterns of collecting effort.