A new toolkit for hydro-climate time series forecasting
A common observation is the occurrence of climate variability from one year to another, causing droughts in some years and excess water or flooding in other. This climate uncertainty affects those that depend on timely annual rainfall resources – people like small subsistence farmers.
Of all the elements of the hydrological cycle, river flow may have the most serious implications for the lives of human beings. Food security depends on the hectares of land irrigated with water from regulated rivers and reservoirs, a good reason for studies related to the planning and management of water resources to focus on flow forecasts. Researchers in this field have managed to collect a considerable number of models for flow prediction, but in many cases the predictions are not the most accurate through time and space. This complication requires the application of various models to better quantify uncertainty and thus make a more informed decision.
The need for hydrological forecasting implies integrating knowledge and experience until a given methodology can be sensibly used: one that meets standards of acceptability and usability. With these objectives in mind, was developed of a new toolkit for forecasting non-linear time series in hydrology and climate (rain, river flow, temperature, humidity, etc) on monthly and seasonal time scales.
PHMet uses climatic variability to forecast hydro-climate time series. It was developed by the National University of Colombia (UNAL) researcher Julian Rojo (download the complete publication at: www.bdigital.unal.edu.co/5493) and uses mathematical models and expert opinions to forecast river flow. Such a hybrid model forecasts features two main components: a regression method, which adjusts the associated variables of the problem to develop the forecast, and a spectral decomposition method used to filter or smooth the signals for cleaner results. The PHMet tool incorporates regression decision-trees to correct flow forecasts, which reduces error by integrating both qualitative information (from experts) and quantitative macro-climatic variables. Forecasts of Pacific Ocean surface temperatures and ENSO conditions can be easily incorporated in the correction flow of the predictions using the appropriate algorithm.
This work was supported by The Inter-American Institute for Global Change Research (IAI), the University Corporation for Atmospheric Research (UCAR), the US National Science Foundation, the Administrative Department of Science, Technology and Innovation of Colombia (COLCIENCIAS) and the Centro de Previsão de Tempo e Estudos Climáticos (CPTEC). Its results have been promising for South-American rivers and hopefully this model will also be applied to other countries around the world.
The PHMet installer can be downloaded at http://julianrojo.weebly.com/phmet.html, where in addition to viewing your own results and methodologies obtained with the tool, the site allows space for discussion, suggestions and feedback on the simulations of other researchers. Thus, through collaboration, users can deliver accurate predictions and values that are the most consistent with the hydro-climatic reality of their countries.