Abstract
The choice for specific parameter estimation methods is often more dependent on its availability than its performance. We developed SPOTPY (Statistical Parameter Optimization Tool), an open source python package containing a comprehensive set of methods typically used to calibrate, analyze and optimize parameters for a wide range of ecological models. SPOTPY currently contains eight widely used algorithms, 11 objective functions, and can sample from eight parameter distributions. SPOTPY has a model-independent structure and can be run in parallel from the workstation to large computation clusters using the Message Passing Interface (MPI). We tested SPOTPY in five different case studies to parameterize the Rosenbrock, Griewank and Ackley functions, a one-dimensional physically based soil moisture routine, where we searched for parameters of the van Genuchten-Mualem function and a calibration of a biogeochemistry model with different objective functions. The case studies reveal that the implemented SPOTPY methods can be used for any model with just a minimal amount of code for maximal power of parameter optimization. They further show the benefit of having one package at hand that includes number of well performing parameter search methods, since not every case study can be solved sufficiently with every algorithm or every objective function.
Highlights
Ecological models are often very complex and contain many parameters that need to be optimized prior to model application
We included the Monte Carlo (MC), Latin Hyper Cube Sampler (LHS) [3] and Robust Parameter Estimation (ROPE) [5] methods that belong to the first group of stochastic probabilistic methods
We found that the rather simple MC and LHS often performed worse when searching the exact global optimum, but give reliable results under equifinality, like it is the case for our hydrological model build with Catchment Modelling Framework (CMF)
Summary
Ecological models are often very complex and contain many parameters that need to be optimized prior to model application. [2,3,4,5,6]), often published without access to the source code They are widely accepted to determine the values of non-measureable parameters for a model [7]. Many of the methods have been established as part of the parameterization problem in hydrological modeling as early as in the 1990s [8,9]. The application of these methods has become more widespread in other ecological disciplines and PLOS ONE | DOI:10.1371/journal.pone.0145180. The application of these methods has become more widespread in other ecological disciplines and PLOS ONE | DOI:10.1371/journal.pone.0145180 December 17, 2015
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
