Computer inference of network of ecological interactions from sampling data.

Both direct and indirect ecological interactions may occur in an ecosystem with large numbers of taxa. Traditional food web technique is a popular tool to measure the quality and health of the environment. Much of works must be done before constructing a food web for an ecosystem especially with many taxa. This food web is generally specific for some ecological interactions and fixed for a set of given species. It is therefore not an effective method for dynamic and prompt assessment of environment. Ecological interactions and their interactive intensity may be detected by sampling biological taxa in the field and by detecting various between-taxa distances or similarities. Network may clearly exhibit the complex interactions among biological taxa. Statistic tests on various distance or similarity measures and computer designs are required to infer the {network. We develop an algorithm and software to infer the network of direct or indirect ecological interactions in ecosystem. It is a prompt and effective tool in monitoring and assessment of the environment. A redundant network may be inferred and drawn by computer based on the statistic tests on sampling data or the pathway information given in HTML file. Dominant taxa may be found in the network. In total of 16 distance and similarity measures, including Euclidean distance, Manhattan distance, Pearson correlation, partial correlation, point correlation, linkage coefficients, Jaccard coefficient etc., are provided to detect taxa pairs with significant parametric or nonparametric similarities, based on randomization tests and ordinary statistic tests. Criteria to use distance and similarity measures are discussed.