Distance- and Phylogeny-based Pathogen Transmission in Forest Communities -- Department of Plant Pathology
Zhongwen Tang is working for Dr. Karen Garett to simulate the spread of disease in mapped plant communities. He runs these simulations with a statistics program called 'R'. The additional computing power that Beocat provides helps him get his work done substantially faster.
Prashanth Boddhireddy's Research
Prashanth is doing plant breeding simulations with DNA-level selections on populations. His work requires a large amount of memory and processing power, and Beocat delivers on that front.
I am running extensive simulations of plant breeding population development. The objective is to test strategies for DNA-level selection during inbreeding, in order to arrive at population structures with improved statistical power to detect genetic loci controlling quantitative traits ("QTLs"). I am simulating thousands of populations with different levels of factors and parameters involved in the simulation, and a statistical method called composite interval mapping is used to evaluate the sensitivity of QTL detection and the precision of QTL location. These simulations are run under Java 1.5.
KSU-WRAPS (Watershed Restoration and Protection Strategies)
Amirpouyan Nejadhashemi is currently working to parallelize code to do watershed modeling analysis. After everything is properly parallelized, he will be using the cluster to do the analyses quicker than before.
Dist-EPIC -- Department of Agronomy
Wei Jin and Matthew Miller have developed a distributed version of EPIC, a tool for analyzing the yields of different crop production methods and tools. The distributed version of EPIC runs on Beocat and cuts the estimated simulation time from 100 years to something far more manageable.
Applying Maize and Water Use Parameters for a Crop Simulator (EPIC)
SM Welch1, W Jin2, M. Miller3, D. Andresen4
Crops form the foundation of the human-natural system in the High Plains. Crop culture inputs, crop and forage yields, and their utilization fuel the regional economy. Plants exert major influence on the aquifer. Irrigation to meet transpiration needs comprises 95+ % of groundwater use in some areas. Calculation of the surface crop water balance, yields, and management practice impacts, and the resulting economics is a key requirement when formulating coupled models of the human/natural system. EPIC (Environmental Policy Integrated Climate) comprises a model that integrates data on climate, soils, crops, cultural practices, and production decisions to estimate cropping system outcomes. Three major processes are represented: (i) phenological development; (ii) dry matter production and partitioning to plant tissues, resulting in growth; and (iii) economic yield. It reproduces the results of irrigation, fertilization, tillage, variety selection, alternative production calendars, etc. It also includes an economic component for evaluating outcomes and optimizing management. Because its original focus was erosion-related, EPIC can simulate decade-scale or longer intervals. It is a parcel-based model and therefore links naturally to the object-based organization that conceptually underpins both AEM and GIS. These features have suited EPIC to a far broad range of applications.
The first step in this project is to parameterize EPIC so as to faithfully represent operating conditions in western Kansas. Given the size of extant databases, the computational requirements for this task is potentially enormous - over a century of CPU time, according to our earliest estimates. Clearly, some form of distributed or parallel computation is required. The other aspect of interest, from the computational side, is the mixed nature calculation: our control algorithm is written as a SciLab script, while EPIC is a standalone FORTRAN binary program. Our solution has been to write a communications component in C++. This component broadcasts a population of trial solutions to worker nodes via the Message Passing Inter-face (MPI) for evaluation by EPIC. The results are collated at the master node provide input to the next iteration of the genetic algorithm (written in SciLab) used for optimization. Using a combination of C++ and MPI permits communication and evaluation to be executed efficiently, while allowing the rapid prototyping and visualization capabilities of Scilab to minimize overall development time of critical control algorithms.
1Dept. Agronomy, Professor
2Dept. Agronomy, Postdoctoral researcher
3Dept. Computer and Information Science, Graduate Research Assistant
4Dept. Computer and Information Science, Assoc. Professor