Britta Anderson 12/15/15 Watershed Modeling with GIS and Remote Sensing How do GIS & RS Support Watershed Modeling? A watershed model is a simplified, mathematical representation of the hydrologic cycle within a given watershed. Models are useful analytical tools for water quality scientists and managers who seek to understand a watershed s hydrologic response to physical changes within and across the landscape. Hydrologic simulation models are often complex and require that the user understands the spatial, conceptual and temporal design of the models they choose to use. Geographic Information System (GIS) and Remote Sensing (RS) technology support watershed modeling by efficiently storing and manipulating spatial data, providing an interface between the data and a model s conceptual design, and delivering a way to simulate past, present and future responses of a watershed to various environmental changes. Spatial data is perhaps the most important element of watershed modeling and it is often collected using GIS and RS technology. The significance of this component is made apparent by the simple fact that as water flows downhill across the landscape, it interacts with various hydrologic, biologic and physical properties spatially distributed on or below the earth s surface. Such properties include, among other things, vegetation patterns, impervious surfaces, soil types, terrain slopes and stream networks. Climactic data, such as precipitation, snow melt and atmospheric temperature information are also considered critical components of simulating the hydrologic cycle. By nature, GIS and RS technologies offer the most efficient and accurate method for collecting, organizing and digitally storing the spatial distribution of physical properties within a watershed, information which was historically collected by hand, recorded on maps and preserved in file cabinets. GIS software supports a model s conceptual design by creating an interface linking mathematical algorithms to a spatial database. Through this linkage, GIS-based models extract information from the database and create overlay maps to mathematically correlate various spatial layers to one another. The correlation is then analyzed by algorithms coded into the GIS-based software. Most hydrologic models can be designed to have either empirical or physical based algorithm inputs. Empirical simulations rely more heavily on processed or synthesized data, whereas physically based simulations use data acquired from in situ measurements. This is one reason why it is important that the GIS-based spatial data and software interface reflects the model s conceptual design. Lastly, GIS and RS technology support the temporal component of watershed modeling by monitoring physical changes across the landscape. These changes may include the seasonal flow of streams, vegetation growth patterns, increases or decreases in the quantity of a resource, or even future development plans. These physical changes are detected via GIS or RS, inventoried within the GIS database and utilized to simulate past, present or future deviations within the watershed. Depending on the type of the data and design parameters, temporal model outputs range from hourly, daily, monthly, seasonally or may even produce continuous results. Data Models Used Most Extensively The use of a GIS-based watershed model depends largely the designer s intended application of the system. Models are generally classified as either lumped or spatially distributed. A lumped model assumes that there is little to no spatial heterogeneity throughout the watershed and physical attributes of subbasins are empirically averaged. Distributed models, on the other hand, assume greater spatial variability in hydrologic, biologic and physical properties. Although distributed models are more data intensive, they are more extensively used because they allow the user more freedom to manipulate the types and distribution of certain variables. As mentioned previously, watershed model require both algorithmic modules and spatial data inputs to accurately predict hydrologic responses to changes across the landscape. Algorithmic modules are used to analyze and process spatial variables to derive watershed characteristics from a known dataset. Many government agencies and research institutions are involved in the creation and revision of modules. Examples of GIS-based modules used extensively include, but are Page 1 of 6
certainly not limited to, the Geographic Resource Analysis Support System (GRASS), created by the US Army Construction Engineering Research Laboratory, ARC/INFO GIS from the Environmental Systems Research Institute (ESRI), Hydrologic Engineering Center (HEC)-GeoHMS, Topographic Parameterization (TOPAZ) and Watershed Modeling System (WMS) software. These modules, or a combination of modules, can be incorporated into watershed models for research and/or commercial purposes. Hydrologic models are being continuously developed and are individually designed to use certain modules and specific types of spatial data. As with modules, many government agencies and research institutions are involved in the creation and revision of these tools. Some of the most extensively used GIS-based models include, but are not limited to, HEC River Analysis System (HEC-RAS) and the Hydrologic Modeling System (HMS), both designed by the U.S. Army Corps of Engineers, Système Hydrologique European (SHE), created by several European research institutes, the Soil and Water Assessment Tool (SWAT) produced by the USDA, and TOPMODEL designed by the University of Lancaster in the U.K. Most Valuable Data Sources Most GIS-based hydrology models require basic elevation, soils, land cover, and climactic data inputs. Digital Elevation Models (DEMs) from the US Geological Survey (USGS) are a critical data source. Digitized topographic information can be stored in grid (raster), triangulated irregular network (TIN) and/or contour-based (vector) format. DEM s provide topographic information for delineating watersheds, such as land slope, drainage divides, catchment areas, stream lengths and channel connectivity. The U.S. Department of Agriculture (USDA) Natural Conservation Service (NRCS) has built an extensive database of the distribution of soils at both state and county scales. These databases are referred to as State Soil Geographic Dataset (STATSGO) and the Soil Survey Geographic database (SSURGO), respectively. STATSGO and SSURGO provide a reference for soil properties such as moisture content/capacity, salinity, porosity, depth to water table or aquifers and soil erodibility factors that play a key role in water distribution throughout a watershed. Land cover and climactic data are two variables often associated with remote sensing. This technology has the capability to provide large scale data on precipitation, land cover, soil moisture, snowmelt and surface or atmospheric temperature. Satellite systems such as SPOT, MODIS, Landsat, LiDAR and Side Looking Airborne Radar (SLAR) technology are now especially reliable sources for classifying land cover within a watershed. Remote sensing can help watershed simulations predict the amount of water lost through evapotranspiration from vegetation, runoff volume and peak flows from impervious surfaces. Furthermore, remotely sensed spatial data provides a simplified record of long-term temporal changes throughout the watershed. It should be noted, however, that collecting climactic data via remote sensing, is still an emerging technology. For example, while evidence exists that quantifying precipitation with radar is reasonably accurate, other studies advocate that in situ rain gauges collect more realistic data (Price et.al.). Analytical Procedures and Modeling Applications GIS-based watershed models are advantageous for quantifying hydrologic responses to predicted changes within a watershed. Data analysis and management decisions, however, must go beyond the immediate data produced by simulation tools. Stake-holders such as researchers, water resource managers and policy officials are held accountable for analyzing data under the correct context of the model and using the results to make informed decisions. On a technical level, one of the most time-consuming stages of watershed modeling is the initial collection and configuration of data. For example, spatial data may only exist in map form, resulting in a labor intensive step of digitizing the information. Other datasets may require extensive interpolation methods to convert data into a usable form or to improve its resolution. Likewise, preparing and calibrating algorithms within simulation modules can require complex statistical analysis to evaluate the accuracy of simulation outputs. Page 2 of 6
Analyzing the results produced by GIS-based watershed models can help solve many environmental and human health problems. For instance, models can help protect drinking water resources by calculating the future accumulation or transport of terrestrial contaminants. Modelling may also be applied to floodplain management by forecasting the impacts of flood events on infrastructure or natural flow regimes. Other models are being applied to ecohydrology and are capable of predicting how land use or climate change will effect natural resources that both humans and wildlife depend on. Whatever the application, it is important that hydrologists have a thorough understanding of the data structure and the conceptual basis of a hydrologic model so that the simulated outputs are analyzed correctly. Future of GIS & RS in Watershed Modeling The future of GIS and RS in watershed modeling appears promising however there is no doubt room for improvement. Due to the data intensive nature of GIS-based simulations, most scientists agree that the number one limiting factor is the collection and processing of various spatial data. The need for data manipulation is not always straight forward and becomes a common source of error that users must be attentive to. Additionally, although remote sensing offers promising potential, it is still an evolving technology whose capabilities require further research. Regardless, GIS-based watershed modeling has become an indispensable tool for hydrologists. GIS and remote sensing technology support watershed modeling by efficiently storing and manipulating spatial data, providing an interface between the data and a model s conceptual design, and delivering a way to simulate past, present and future responses of a watershed to various environmental changes. This technology in continuously developing, providing hydrologists with a better understanding of watershed hydrology and also creating more effective solutions to water quality problems, ultimately protecting human and environmental health. Annotated Bibliographies Band, L. E. (1986). Topographic partition of watersheds with digital elevation models. Water resources research, 22(1), 15-24. The objective of this paper is to introduce an algorithmic module designed to extract and construct stream networks and drainage basins from raster elevation files (Digital Elevation Models (DEM)). In other words, the author provides a method for converting DEMs from raster to vector format to more accurately depict the spatial distribution of watershed characteristics, such as drainage divides, stream lines and subbasin polygons. The author provides a brief review of previous studies, justifies the need for a new model to correct computation errors, and discusses the techniques his algorithm is based upon, such as concave and convex pixels. It is brought to the reader s attention that the main limitation of existing techniques is a lack of connectivity between polylines when they are extracted from DEM pixel segments. The author emphasizes that, despite these mistakes, previous algorithmic modules are quite valuable and states how each method is incorporated into his system. The author s model is unique in that it integrates a maximum descent algorithm instructing the model to drain pixels based on elevation or until another stream segment is encountered. This algorithm also codes a set of topology rules linking stream networks to drainage pixels, thus ensuring connectivity throughout the watershed delineation. This paper is valuable because it provides a solution to connectivity errors in watershed modeling that are keeping DEMs from being utilized to their fullest potential. As a result, hydrologists can now design models that more accurately simulate drainage and stream networks. DeVantier, B. A., & Feldman, A. D. (1993). Review of GIS applications in hydrologic modeling. Journal of Water Resources Planning and Management, 119(2), 246-261. In this paper, DeVantier et.al. summarize the development of Geographical Information System (GIS) based hydrologic models for watershed analysis purposes. The paper discusses the structure of spatial data storage, conceptual designs of current hydrologic models, and existing evidence of the application of GIS to hydrologic analysis. In the beginning of the paper that the authors are somewhat skeptical of the accuracy of GIS-based modeling. They acknowledge the value of GIS Page 3 of 6
systems for storing digital data that would otherwise be depicted on paper maps, however they emphasize the complexity of collecting and processing the data required to run a hydrologic model using GIS. Nevertheless, the authors review basic GIS fundamentals such as data types and data handling approaches, such as Digital Elevation Models (DEMs), Triangular Irregular Network (TIN) interpolation methods, and emerging remote sensing technologies. A major portion of the paper is dedicated to describing the types of models used in hydrologic analysis such as lumped parameter models and physics based models. This discussion was valuable because it provided a technical overview of how spatial data inputs are manipulated before the simulation occurs. The authors follow-up this discussion by providing examples of modeling results produced by the Hydrologic Engineering Center (HEC), such as forecasting the impacts of watershed runoff (i.e. erosion, flooding, transportation of contaminants, etc.). Overall this paper was valuable because of its pragmatic approach to the design, functionality and application of GIS-based technology to hydrologic modeling. It reminds the reader that current GIS-models are still evolving, and although the benefits are clear, hydrologists should be cautious about the accuracy of these complex models until GIS becomes an established analytical tool. Garbrecht, J., Ogden, F. L., DeBarry, P. A., & Maidment, D. R. (2001). GIS and distributed watershed models. I: Data coverages and sources. Journal of Hydrologic Engineering, 6(6), 506-514. In this paper, Garbrecht et.al. acknowledge that the recent emergence of Geographic Information Systems (GIS) spatial data and hydrologic models has made it challenging for engineers to evaluate their options and make informed decisions on the correct way to use these resources. The authors focus almost exclusively on the types, sources and the structure of various spatial data and emphasize the importance of formatting information so it can be processed by GIS-based watershed models. GIS fundamentals, such as raster and vector format, and interpolation methods, are introduced first, followed by watershed specific data sources such as Digital Elevation Models (DEMs). In fact, the authors discuss DEMs to a great extent, explaining how the DEMs can be manipulated to reliably identify stream networks, drainage divides, and slopes among other attributes. The most valuable content of the paper is an overview of information that is absolutely necessary to operate a watershed model, including soil, precipitation, temperature, land cover and topography data. This component is valuable because it provides an initial resource for a beginner audience. To a lesser extent, the authors discuss remote sensing and predict that this developing technology will likely become an indispensable resource in the future. In conclusion, the paper was well written and complied with its initial intention of simplifying GIS concepts and providing an overview of data sources to a general audience. Houser, P. R., Shuttleworth, W. J., Famiglietti, J. S., Gupta, H. V., Syed, K. H., & Goodrich, D. C. (1998). Integration of soil moisture remote sensing and hydrologic modeling using data assimilation. Water Resources Research, 34(12). In this paper, Houser et.al. present a methodology for assimilating distributed fields of soil moisture into TOPMODEL-based Land Atmosphere Transfer Scheme (TOPLATS), a semi-distributed hydrologic model. The authors explain that remote sensing of soil moisture is inherently limited by soil types, elevation, vegetation cover and temporal variables. Likewise, hydrologic models are capable of calculating soil moisture but are themselves limited by model structure and associated parameters. Therefore, the objective of the paper is to solve both of these problems at once by designing an assimilation method that more accurately predicts the interaction between atmospheric soil moisture content and hydrologic processes. The methodology involved the collection of six 160 square-kilometer push broom microwave radiometer (PBMR) images of a watershed in southeast Arizona. The remotely sensed data was processed and entered into a TOPLATS model modified with a statistical correction assimilation method. The results suggest that the adjusted model replicates known soil-moisture data and that the model calculates realistic soil-moisture patterns within the watershed. Although the statistical analysis section of the paper is critical to justifying the methodology, the most interesting part of the paper describes the technology utilized to capture spatial data throughout the watershed. The PBMR instruments derived soil moisture content by collecting brightness temperature data. Brightness temperature was observed to decrease during and immediately after a rainfall event. The most valuable part of the paper is the conclusion, where the authors express limitations to their module; although it is computationally efficient, the methodology it is limited by the amount and complexity of data that must be collected to procure accurate results. Page 4 of 6
Ogden, F. L., Garbrecht, J., DeBarry, P. A., & Johnson, L. E. (2001). GIS and distributed watershed models. II: Modules, interfaces, and models. Journal of Hydrologic Engineering, 6(6), 515-523. This paper, by Ogden et.al., is a continuum of GIS and Distributed Watershed Models: I: Data Coverages (Garbrecht et.al.). The authors assume that the audience is familiar with GIS fundamentals and the hydrologic data sources required by most watershed modeling systems. As such, the intent of this paper is to introduce to engineers more advanced concepts of the application of spatial data to Geographic Information System (GIS) based watershed models. The authors begin by introducing GIS modules used for hydrologic data processing within watershed models themselves. Brief descriptions of each module provide the audience with a general understanding to the design and capability of each program in addition to the agency or company that originally coded the software. The best section of the paper discusses watershed models that use the GIS modules and geospatial data to forecast the impact of changing hydrologic variables on watershed characteristics and water quality. This component is valuable because the authors provided an unbiased review of the scope, functionality, advantages and disadvantages of the most commonly used watershed models. Therefore, the audience becomes capable of choosing a model to best suit their needs. The authors end the paper by discussing current and future developments in GIS-based watershed modeling in addition to actions the audience can take to successfully manage their own spatial databases. Overall this paper was well written and achieved its objective of presenting a more technical view of the current application and limitations of GIS-based watershed modeling tools. Price, K., Purucker, S. T., Kraemer, S. R., Babendreier, J. E., & Knightes, C. D. (2014). Comparison of radar and gauge precipitation data in watershed models across varying spatial and temporal scales. Hydrological Processes, 28(9), 3505-3520. The purpose of this paper, written by Price et.al., is to compare the accuracy of a precipitation data collected by a radar system (Multisensor Precipitation Estimator (MPE or Stage IV Next-Generation Radar)) to tipping-bucket rain gauge data within a semi-distributed watershed model (Soil and Water Assessment Tool (SWAT)). The authors objective attempts to provide more conclusive evidence that one data source is more accurate over the other, as many existing studies have produced unclear or conflicting results. As such, the authors chose to simulate streamflow by applying SWAT to four nested North Carolina watersheds over five temporal events. Stream flow was simulated twice for each watershed, once with gauge precipitation data and once with radar precipitation data, while keeping all other model inputs constant. The study found that the accuracy of each system depends on the temporal scale (i.e. daily, weekly, monthly, etc. rainfall), the size of the watershed and the distribution of rain gauges within or near the watershed in question. More specifically, the authors found that radar results were more accurate during large rainfall events whereas rain gauges were more accurate during small scale rain events. It was interesting that they suggested the best solution could be using a combination of the two technologies, such as correcting radar data by using rain gauge data. The best section of the paper was the conclusion, where the authors discussed that hydrologists must understand that the use of either technology requires that they are familiar with the spatial and temporal scales of certain modeling inputs in order the achieve accurate results. To do this, the authors outlined 4 recommendations for further research to understand the relationship between watershed models and precipitation data: Model structure, user defined spatial discretion, storm events and precipitation storm structure. Reference Summary Band, L. E. (1986). Topographic partition of watersheds with digital elevation models. Water Resources Research, 22(1), 15-24. Brooks, Kenneth N., Ffolliott, Peter F., & Magner, Joseph A. (2012). Tools and Emerging Technologies. In Hydrology and the Management of Watersheds (pp. 489-511). Oxford, UK: Blackwell Publishing. Page 5 of 6
DeVantier, B. A., & Feldman, A. D. (1993). Review of GIS applications in hydrologic modeling. Journal of Water Resources Planning and Management, 119(2), 246-261. Daniel, E. B., Camp, J. V., LeBoeuf, E. J., Penrod, J. R., Dobbins, J. P., & Abkowitz, M. D. (2011). Watershed modeling and its applications: A state-of-the-art review. Open Hydrology Journal, 5(2). Garbrecht, J., Ogden, F. L., DeBarry, P. A., & Maidment, D. R. (2001). GIS and distributed watershed models. I: Data coverages and sources. Journal of Hydrologic Engineering, 6(6), 506-514. Houser, P. R., Shuttleworth, W. J., Famiglietti, J. S., Gupta, H. V., Syed, K. H., & Goodrich, D. C. (1998). Integration of soil moisture remote sensing and hydrologic modeling using data assimilation. Water Resources Research, 34(12). Kite, G. W., & Kouwen, N. (1992). Watershed modeling using land classifications. Water Resources Research, 28(12), 3193-3200. Ogden, F. L., Garbrecht, J., DeBarry, P. A., & Johnson, L. E. (2001). GIS and distributed watershed models. II: Modules, interfaces, and models. Journal of Hydrologic Engineering, 6(6), 515-523. Price, K., Purucker, S. T., Kraemer, S. R., Babendreier, J. E., & Knightes, C. D. (2014). Comparison of radar and gauge precipitation data in watershed models across varying spatial and temporal scales. Hydrological Processes, 28(9), 3505-3520. Srinivasan, R., Ramanarayanan, T., Arnold, J., & Bednarz, S. (1998). LARGE AREA HYDROLOGIC MODELING AND ASSESSMENT PART II: MODEL APPLICATION 1. JAWRA Journal of the American Water Resources Association, 34(1), 91-101. Page 6 of 6