skip to main content

Title: Community Workflows to Advance Reproducibility in Hydrologic Modeling: Separating Model‐Agnostic and Model‐Specific Configuration Steps in Applications of Large‐Domain Hydrologic Models

Despite the proliferation of computer‐based research on hydrology and water resources, such research is typically poorly reproducible. Published studies have low reproducibility due to incomplete availability of data and computer code, and a lack of documentation of workflow processes. This leads to a lack of transparency and efficiency because existing code can neither be quality controlled nor reused. Given the commonalities between existing process‐based hydrologic models in terms of their required input data and preprocessing steps, open sharing of code can lead to large efficiency gains for the modeling community. Here, we present a model configuration workflow that provides full reproducibility of the resulting model instantiations in a way that separates the model‐agnostic preprocessing of specific data sets from the model‐specific requirements that models impose on their input files. We use this workflow to create large‐domain (global and continental) and local configurations of the Structure for Unifying Multiple Modeling Alternatives (SUMMA) hydrologic model connected to the mizuRoute routing model. These examples show how a relatively complex model setup over a large domain can be organized in a reproducible and structured way that has the potential to accelerate advances in hydrologic modeling for the community as a whole. We provide a tentative blueprint of how community modeling initiatives can be built on top of workflows such as this. We term our workflow the “Community Workflows to Advance Reproducibility in Hydrologic Modeling” (CWARHM; pronounced “swarm”).

more » « less
Award ID(s):
1835569 1928369
Author(s) / Creator(s):
 ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  
Publisher / Repository:
DOI PREFIX: 10.1029
Date Published:
Journal Name:
Water Resources Research
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Krzhizhanovskaya, Valeria V. ; Závodszky, Gábor ; Lees, Michael H. ; Dongarra, Jack J. ; Sloot, Peter M. ; Brissos, Sérgio ; Teixeira, João (Ed.)
    The HydroFrame project is a community platform designed to facilitate integrated hydrologic modeling across the US. As a part of HydroFrame, we seek to design innovative workflow solutions that create pathways to enable hydrologic analysis for three target user groups: the modeler, the analyzer, and the domain science educator. We present the initial progress on the HydroFrame community platform using an automated Kepler workflow. This workflow performs end-to-end hydrology simulations involving data ingestion, preprocessing, analysis, modeling, and visualization. We demonstrate how different modules of the workflow can be reused and repurposed for the three target user groups. The Kepler workflow ensures complete reproducibility through a built-in provenance framework that collects workflow specific parameters, software versions, and hardware system configuration. In addition, we aim to optimize the utilization of large-scale computational resources to adjust to the needs of all three user groups. Towards this goal, we present a design that leverages provenance data and machine learning techniques to predict performance and forecast failures using an automatic performance collection component of the pipeline. 
    more » « less
  2. Abstract

    Many have argued that datasets resulting from scientific research should be part of the scholarly record as first class research products. Data sharing mandates from funding agencies and scientific journal publishers along with calls from the scientific community to better support transparency and reproducibility of scientific research have increased demand for tools and support for publishing datasets. Hydrology domain‐specific data publication services have been developed alongside more general purpose and even commercial data repositories. Prominent among these are the Hydrologic Information System (HIS) and HydroShare repositories developed by the Consortium of Universities for the Advancement of Hydrologic Science, Inc. (CUAHSI). More broadly, however, multiple organizations have been involved in the practice of data publication in the hydrology domain, each having different roles that have shaped data publication and reuse. Bibliographic and archival approaches to data publication have been advanced, but both have limitations with respect to hydrologic data. Specific recommendations for improving data publication infrastructure, support, and practices to move beyond existing limitations and enable more effective data publication in support of scientific research in the hydrology domain include: improving support for journal article‐based data access and data citation, considering the workflow for data publication, enhancing support for reproducible science, encouraging publication of curated reference data collections, advancing interoperability standards for sharing data and metadata among repositories, developing partnerships with university libraries offering data services, and developing more specific data management plans. While presented in the context of CUAHSI's data repositories and experience, these recommendations are broadly applicable to other domains.

    This article is categorized under:

    Science of Water > Methods

    more » « less
  3. Obeid, I. ; Selesnik, I. ; Picone, J. (Ed.)
    The Neuronix high-performance computing cluster allows us to conduct extensive machine learning experiments on big data [1]. This heterogeneous cluster uses innovative scheduling technology, Slurm [2], that manages a network of CPUs and graphics processing units (GPUs). The GPU farm consists of a variety of processors ranging from low-end consumer grade devices such as the Nvidia GTX 970 to higher-end devices such as the GeForce RTX 2080. These GPUs are essential to our research since they allow extremely compute-intensive deep learning tasks to be executed on massive data resources such as the TUH EEG Corpus [2]. We use TensorFlow [3] as the core machine learning library for our deep learning systems, and routinely employ multiple GPUs to accelerate the training process. Reproducible results are essential to machine learning research. Reproducibility in this context means the ability to replicate an existing experiment – performance metrics such as error rates should be identical and floating-point calculations should match closely. Three examples of ways we typically expect an experiment to be replicable are: (1) The same job run on the same processor should produce the same results each time it is run. (2) A job run on a CPU and GPU should produce identical results. (3) A job should produce comparable results if the data is presented in a different order. System optimization requires an ability to directly compare error rates for algorithms evaluated under comparable operating conditions. However, it is a difficult task to exactly reproduce the results for large, complex deep learning systems that often require more than a trillion calculations per experiment [5]. This is a fairly well-known issue and one we will explore in this poster. Researchers must be able to replicate results on a specific data set to establish the integrity of an implementation. They can then use that implementation as a baseline for comparison purposes. A lack of reproducibility makes it very difficult to debug algorithms and validate changes to the system. Equally important, since many results in deep learning research are dependent on the order in which the system is exposed to the data, the specific processors used, and even the order in which those processors are accessed, it becomes a challenging problem to compare two algorithms since each system must be individually optimized for a specific data set or processor. This is extremely time-consuming for algorithm research in which a single run often taxes a computing environment to its limits. Well-known techniques such as cross-validation [5,6] can be used to mitigate these effects, but this is also computationally expensive. These issues are further compounded by the fact that most deep learning algorithms are susceptible to the way computational noise propagates through the system. GPUs are particularly notorious for this because, in a clustered environment, it becomes more difficult to control which processors are used at various points in time. Another equally frustrating issue is that upgrades to the deep learning package, such as the transition from TensorFlow v1.9 to v1.13, can also result in large fluctuations in error rates when re-running the same experiment. Since TensorFlow is constantly updating functions to support GPU use, maintaining an historical archive of experimental results that can be used to calibrate algorithm research is quite a challenge. This makes it very difficult to optimize the system or select the best configurations. The overall impact of all of these issues described above is significant as error rates can fluctuate by as much as 25% due to these types of computational issues. Cross-validation is one technique used to mitigate this, but that is expensive since you need to do multiple runs over the data, which further taxes a computing infrastructure already running at max capacity. GPUs are preferred when training a large network since these systems train at least two orders of magnitude faster than CPUs [7]. Large-scale experiments are simply not feasible without using GPUs. However, there is a tradeoff to gain this performance. Since all our GPUs use the NVIDIA CUDA® Deep Neural Network library (cuDNN) [8], a GPU-accelerated library of primitives for deep neural networks, it adds an element of randomness into the experiment. When a GPU is used to train a network in TensorFlow, it automatically searches for a cuDNN implementation. NVIDIA’s cuDNN implementation provides algorithms that increase the performance and help the model train quicker, but they are non-deterministic algorithms [9,10]. Since our networks have many complex layers, there is no easy way to avoid this randomness. Instead of comparing each epoch, we compare the average performance of the experiment because it gives us a hint of how our model is performing per experiment, and if the changes we make are efficient. In this poster, we will discuss a variety of issues related to reproducibility and introduce ways we mitigate these effects. For example, TensorFlow uses a random number generator (RNG) which is not seeded by default. TensorFlow determines the initialization point and how certain functions execute using the RNG. The solution for this is seeding all the necessary components before training the model. This forces TensorFlow to use the same initialization point and sets how certain layers work (e.g., dropout layers). However, seeding all the RNGs will not guarantee a controlled experiment. Other variables can affect the outcome of the experiment such as training using GPUs, allowing multi-threading on CPUs, using certain layers, etc. To mitigate our problems with reproducibility, we first make sure that the data is processed in the same order during training. Therefore, we save the data from the last experiment and to make sure the newer experiment follows the same order. If we allow the data to be shuffled, it can affect the performance due to how the model was exposed to the data. We also specify the float data type to be 32-bit since Python defaults to 64-bit. We try to avoid using 64-bit precision because the numbers produced by a GPU can vary significantly depending on the GPU architecture [11-13]. Controlling precision somewhat reduces differences due to computational noise even though technically it increases the amount of computational noise. We are currently developing more advanced techniques for preserving the efficiency of our training process while also maintaining the ability to reproduce models. In our poster presentation we will demonstrate these issues using some novel visualization tools, present several examples of the extent to which these issues influence research results on electroencephalography (EEG) and digital pathology experiments and introduce new ways to manage such computational issues. 
    more » « less
  4. Abstract

    Soil microbial communities play critical roles in various ecosystem processes, but studies at a large spatial and temporal scale have been challenging due to the difficulty in finding the relevant samples in available data sets as well as the lack of standardization in sample collection and processing. The National Ecological Observatory Network (NEON) has been collecting soil microbial community data multiple times per year for 47 terrestrial sites in 20 eco‐climatic domains, producing one of the most extensive standardized sampling efforts for soil microbial biodiversity to date. Here, we introduce the neonMicrobe R package—a suite of downloading, preprocessing, data set assembly, and sensitivity analysis tools for NEON’s newly published 16S and ITS amplicon sequencing data products which characterize soil bacterial and fungal communities, respectively. neonMicrobe is designed to make these data more accessible to ecologists without assuming prior experience with bioinformatic pipelines. We describe quality control steps used to remove quality‐flagged samples, report on sensitivity analyses used to determine appropriate quality filtering parameters for the DADA2 workflow, and demonstrate the immediate usability of the output data by conducting standard analyses of soil microbial diversity. The sequence abundance tables produced byneonMicrobecan be linked to NEON’s other data products (e.g., soil physical and chemical properties, plant community composition) and soil subsamples archived in the NEON Biorepository. We provide recommendations for incorporatingneonMicrobeinto reproducible scientific workflows, discuss technical considerations for large‐scale amplicon sequence analysis, and outline future directions for NEON‐enabled microbial ecology. In particular, we believe that NEON marker gene sequence data will allow researchers to answer outstanding questions about the spatial and temporal dynamics of soil microbial communities while explicitly accounting for scale dependence. We expect that the data produced by NEON and theneonMicrobeR package will act as a valuable ecological baseline to inform and contextualize future experimental and modeling endeavors.

    more » « less
  5. . Granting agencies invest millions of dollars on the generation and analysis of data, making these products extremely valuable. However, without sufficient annotation of the methods used to collect and analyze the data, the ability to reproduce and reuse those products suffers. This lack of assurance of the quality and credibility of the data at the different stages in the research process essentially wastes much of the investment of time and funding and fails to drive research forward to the level of potential possible if everything was effectively annotated and disseminated to the wider research community. In order to address this issue for the Hawai'i Established Program to Stimulate Competitive Research (EPSCoR) project, a water science gateway was developed at the University of Hawai‘i (UH), called the ‘Ike Wai Gateway. In Hawaiian, ‘Ike means knowledge and Wai means water. The gateway supports research in hydrology and water management by providing tools to address questions of water sustainability in Hawai‘i. The gateway provides a framework for data acquisition, analysis, model integration, and display of data products. The gateway is intended to complement and integrate with the capabilities of the Consortium of Universities for the Advancement of Hydrologic Science's (CUAHSI) Hydroshare by providing sound data and metadata management capabilities for multi-domain field observations, analytical lab actions, and modeling outputs. Functionality provided by the gateway is supported by a subset of the CUAHSI’s Observations Data Model (ODM) delivered as centralized web based user interfaces and APIs supporting multi-domain data management, computation, analysis, and visualization tools to support reproducible science, modeling, data discovery, and decision support for the Hawai'i EPSCoR ‘Ike Wai research team and wider Hawai‘i hydrology community. By leveraging the Tapis platform, UH has constructed a gateway that ties data and advanced computing resources together to support diverse research domains including microbiology, geochemistry, geophysics, economics, and humanities, coupled with computational and modeling workflows delivered in a user friendly web interface with workflows for effectively annotating the project data and products. Disseminating results for the ‘Ike Wai project through the ‘Ike Wai data gateway and Hydroshare makes the research products accessible and reusable. 
    more » « less