NetMap's Technical Help Guide

GIS Data:  Advantages and Limitations

GIS Data: Advantages and Limitations
During the last decade there has been a proliferation of geospatial data in natural resource management including in the disciplines of forestry, fishery management, geology, geomorphology, hydrology, wildfire and climate change (Miller 2003, Wing and Bettinger 2008). Geographical information system (GIS) data and associated model output are only as good as the remote sensing methods from which they are derived (e.g., aerial photographs, satellite imagery, laser altimetry, field surveys, digitizing, etc.). Important attributes about GIS data include their spatial (three dimensional) resolution (90 m, 10 m, and <10 m), accuracy, and precision. In addition, GIS information derived from predictive numerical models is also only as good as the model and the data that go into it.
 
GIS data used in natural resource management can include hillslope gradients, aspects, stream networks, stream gradients, vegetation and other watershed features. In general, across the western U. S., 10-m digital elevation models (DEMs) are used within GIS-based numerical models to derive these and other watershed attributes such as slope stability, debris flow potential, and channel and fish habitat characteristics (Benda et al. 2007, Burnett et al. 2007). Forest growth models (FVS, Zelig, ORGANON) that use plot- scale field data are used to create predictions about stand structure over time. These model predictions, as well as others that use a single year’s remote sensed data on stand structure, can be used to forecast the recruitment of wood to channels (using yet other models), and those predictions can be used to predict changes to fish habitat quality and abundance.
 
It is important to remember that GIS raster or cell-based data are relatively ‘coarse grained’ which means that data, such as vegetation type, is represented by square cells with sides of length, for example 90 m or 10 m with cell areas of 8100 m2 or 100 m2, respectively. These types of data are not accurate down to a more human scale of meters (e.g., while standing in the field); an exception is GIS information that utilizes sub-meter resolution LIDAR. Forest data at coarse scales are generalized, or averaged, and thus GIS information of forest structure will be only accurate in an averaged sense. Nevertheless, this type of coarse-grained information could be used effectively to plan timber harvest and or forest restoration activities across a large watershed over the next 50 years.
 
Another type of GIS information is vector (line) data such as stream channels that are derived either from digitizing paper (USGS) maps or from numerical models that use DEMs and roads (typically digitized from paper maps or aerial photographs). The accuracy of stream lines depends on the accuracy of the original map product (such as U.S.G.S. 1:24,000-scale topographic maps or the resolution of DEMs). If channel network extraction models are used (Miller and Burnett 2007), the accuracy of the delineated channel network will be much better using 10-m versus 90-m digital elevation data. Similarly, the stream attributes so derived (e.g., gradient, floodplain width, orientation, etc.) will also depend on the DEM resolution and on the robustness of the numerical model itself.  For example, if the delineated channel segments are 100 m in length, then the predicted channel gradients will be an average over that length scale.
 
The spatial accuracy of road lines is dependent on the care with which the locations of roads were digitized from maps or photos. The attributes that are extracted from roads, such as road gradients and drainage points, are also dependent on the digitizing accuracy.
 
Given the necessary coarse grain and, thus, approximate nature of most GIS data and numerical model predictions, the relative difference among values (whether grid cells, lines, points or polygons) is likely more accurate compared to the absolute value of any single data point. For instance, predictions of slope stability typically reveal a large range of failure potential across a watershed. The value of any site specific prediction (pixel scale) provides only a rough approximation of reality (because of model limitations and uncertainty in governing parameters). The relative difference between areas of high and low instability, however, can provide a more accurate accounting of hillslope stability (or erosion potential) across a watershed and this type of knowledge is suitable for planning purposes.
 
Watershed- to landscape-scale GIS information about topography, stream networks, forest vegetation, erosion potential, and aquatic habitat has provided an unprecedented ability to consider entire watersheds (and landscapes) in the implementation of forestry and fishery management (Spies and Johnson 2007), and also to quantitatively forecast outcomes, including cumulative effects of forest practices (Dunne et al. 2001). Prior to advanced GIS, numerical models, and computer technology, this capability did not exist. Given the limitations of GIS information and associated numerical models, but also the advantages of these information systems, it is important to ask the following question: How do resource managers and analysts apply geospatial data and models in their day-to-day work? 
 

Created with help of Dr.Explain

Copyright Earth Systems Institute 2013