Michael P. Finn

Geospatial data resampling and resolution effects on watershed modeling: A case study using the agricultural non-point source pollution model

Abstract.Researchers have been coupling geographic information systems (GIS) data handling and processing capability to watershed and water-quality models for many years. This capability is suited for the development of databases appropriate for water modeling. However, it is rare for GIS to provide direct inputs to the models. To demonstrate the logical procedure of coupling GIS for model parameter extraction, we selected the Agricultural Non-Point Source (AGNPS) pollution model. Investigators can generate data layers at various resolutions and resample to pixel sizes to support models at particular scales. We developed databases of elevation, land cover, and soils at various resolutions in four watersheds. The ability to use multiresolution databases for the generation of model parameters is problematic for grid-based models. We used database development procedures and observed the effects of resolution and resampling on GIS input datasets and parameters generated from those inputs for AGNPS. Results indicate that elevation values at specific points compare favorably between 3- and 30-m raster datasets. Categorical data analysis indicates that land cover classes vary significantly. Derived parameters parallel the results of the base GIS datasets. Analysis of data resampled from 30-m to 60-, 120-, 210-, 240-, 480-, 960-, and 1920-m pixels indicates a general degradation of both elevation and land cover correlations as resolution decreases. Initial evaluation of model output values for soluble nitrogen and phosphorous indicates similar degradation with resolution.

Remote Sensing of Soil Moisture Using Airborne Hyperspectral Data

Landscape assessment of soil moisture is critical to understanding the hydrological cycle at the regional scale and in broad-scale studies of biophysical processes affected by global climate changes in temperature and precipitation. Traditional efforts to measure soil moisture have been principally restricted to in situ measurements, so remote sensing techniques are often employed. Hyperspectral sensors with finer spatial resolution and narrow band widths may offer an alternative to traditional multispectral analysis of soil moisture, particularly in landscapes with high spatial heterogeneity. This preliminary research evaluates the ability of remotely sensed hyperspectral data to quantify soil moisture for the Little River Experimental Watershed (LREW), Georgia. An airborne hyperspectral instrument with a short-wavelength infrared (SWIR) sensor was flown in 2005 and 2007 and the results were correlated to in situ soil moisture values. A significant statistical correlation (R 2 value above 0.7 for both sampling dates) for the hyperspectral instrument data and the soil moisture probe data at 5.08 cm (2 inches) was determined. While models for the 20.32 cm (8 inches) and 30.48 cm (12 inches) depths were tested, they were not able to estimate soil moisture to the same degree.

Implications of Web Mercator and Its Use in Online Mapping

Online interactive maps have become a popular means of communicating with spatial data. In most online mapping systems, Web Mercator has become the dominant projection. While the Mercator projection has a long history of discussion about its inappropriateness for general-purpose mapping, particularly at the global scale, and seems to have been virtually phased out for general-purpose global-scale print maps, it has seen a resurgence in popularity in Web Mercator form. This article theorizes on how Web Mercator came to be widely used for online maps and what this might mean in terms of data display, technical aspects of map generation and distribution, design, and cognition of spatial patterns. The authors emphasize details of where the projection excels and where it does not, as well as some of its advantages and disadvantages for cartographic communication, and conclude with some research directions that may help to develop better solutions to the problem of projections for general-purpose, multi-scale Web mapping. Les cartes interactives en ligne sont devenues un moyen populaire de communiquer au moyen de données spatiales. Dans la plupart des systèmes de cartographie en ligne, la projection de Mercator sur le Web est devenue la projection dominante. La projection de Mercator soulève depuis longtemps des discussions sur son caractère inapproprié en cartographie générale, particulièrement à l’échelle de la planète, et elle semble avoir à peu près disparu des cartes imprimées à l’échelle mondiale d’usage général, mais on a constaté un regain de popularité de la projection de Mercator sur le Web. Cet article présente une théorie sur la façon dont la projection de Mercator sur le Web s’est généralisée pour les cartes en ligne et sur ce que cela pourrait signifier pour l’affichage des données, les aspects techniques de la production et de la distribution de cartes, la conception et la cognition des tendances spatiales. Les auteurs mettent en évidence des détails sur les aspects où la projection excelle et sur ceux où elle n’excelle pas, ainsi que certains de ses avantages et inconvénients pour la communication cartographique. Ils concluent par des pistes de recherche qui peuvent aider à trouver une meilleure solution au problème des projections destinées à la cartographie générale à échelles multiples sur le Web.

Projecting Global Datasets to Achieve Equal Areas

Scientists routinely accomplish global modeling in the raster domain, but recent research has indicated that the transformation of large areas through map projection equations leads to errors. This research attempts to gauge the extent of map projection and resampling effects on the tabulation of categorical areas by comparing the results of three datasets for seven common projections. The datasets, Global Land Cover, Holdridge Life Zones, and Global Vegetation, were compiled at resolutions of 30 arc-second, ½ degree, and 1 degree, respectively. These datasets were projected globally from spherical coordinates to plane representations. Results indicate significant problems in the implementation of global projection transformations in commercial software, as well as differences in areal accuracy across projections. The level of raster resolution directly affects the accuracy of areal tabulations, with higher resolution yielding higher accuracy. If the raster resolution is high enough for individual pixels to approximate points, the areal error tends to zero. The 30-arc-second cells appear to approximate this condition.

A program for handling map projections of small-scale geospatial raster data

Scientists routinely accomplish small-scale geospatial modeling using raster datasets of global extent. Such use often requires the projection of global raster datasets onto a map or the reprojection from a given map projection associated with a dataset. The distortion characteristics of these projection transformations can have significant effects on modeling results. Distortions associated with the reprojection of global data are generally greater than distortions associated with reprojections of larger-scale, localized areas. The accuracy of areas in projected raster datasets of global extent is dependent on resolution. To address these problems of projection and the associated resampling that accompanies it, methods for framing the transformation space, direct point-to-point transformations rather than gridded transformation spaces, a solution to the wrap-around problem, and an approach to alternative resampling methods are presented. The implementations of these methods are provided in an open source software package called MapImage (or mapIMG , for short), which is designed to function on a variety of computer architectures.

Shapes on a plane: evaluating the impact of projection distortion on spatial binning

ABSTRACT One method for working with large, dense sets of spatial point data is to aggregate the measure of the data into polygonal containers, such as political boundaries, or into regular spatial bins such as triangles, squares, or hexagons. When mapping these aggregations, the map projection must inevitably distort relationships. This distortion can impact the reader’s ability to compare count and density measures across the map. Spatial binning, particularly via hexagons, is becoming a popular technique for displaying aggregate measures of point data sets. Increasingly, we see questionable use of the technique without attendant discussion of its hazards. In this work, we discuss when and why spatial binning works and how mapmakers can better understand the limitations caused by distortion from projecting to the plane. We introduce equations for evaluating distortion’s impact on one common projection (Web Mercator) and discuss how the methods used generalize to other projections. While we focus on hexagonal binning, these same considerations affect spatial bins of any shape, and more generally, any analysis of geographic data performed in planar space.

Recent literature in cartography and geographic information science

Ayat, B., Y. Yuksel, and B. Aydogan. 2013. “Black Sea Wave Energy Atlas from 13 Years Hindcasted Wave Data.” Renewable Energy: An International Journal 57: 436–447. Brosnan, K. A. 2013. “Napa Valley Historical Ecology Atlas: Exploring a Hidden Landscape of Transformation and Resilience.” Environmental History 18 (4): 796–798. Faenza, L., and V. Lauciani. 2013. “The ShakeMap Atlas for the City of Naples, Italy.” Seismological Research Letters 84 (6): 963–972. Krishnan, N., and D. Morris. 2012. “Mapping Adaptation Opportunities and Activities in an Interactive Atlas.” AMBIO – A Journal of the Human Environment 41: 90–99. Schaffner, F. C. 2012. “Seabird Breeding Atlas of the Lesser Antilles.” The Auk 129: 795–796.

Effects of Sampling Interval on Spatial Patterns and Statistics of Watershed Nitrogen Concentration

This study investigates how spatial patterns and statistics of a 30 m resolution, model-simulated, watershed nitrogen concentration surface change with sampling intervals from 30 m to 600 m for every 30 m increase for the Little River Watershed (Georgia, USA). The results indicate that the mean, standard deviation, and variogram sills do not have consistent trends with increasing sampling intervals, whereas the variogram ranges remain constant. A sampling interval smaller than or equal to 90 m is necessary to build a representative variogram. The interpolation accuracy, clustering level, and total hot spot areas show decreasing trends approximating a logarithmic function. The trends correspond to the nitrogen variogram and start to level at a sampling interval of 360 m, which is therefore regarded as a critical spatial scale of the Little River Watershed.

Data Layer Integration for The National Map of the United States

The integration of geographic data layers in multiple raster and vector formats, from many different organizations and at a variety of resolutions and scales, is a significant problem for The National Map of the United States being developed by the U.S. Geological Survey. Our research has examined data integration from a layer-based approach for five of The National Map data layers: digital orthoimages, elevation, land cover, hydrography, and transportation. An empirical approach has included visual assessment by a set of respondents with statistical analysis to establish the meaning of various types of integration. A separate theoretical approach with established hypotheses tested against actual data sets has resulted in an automated procedure for integration of specific layers and is being tested. The empirical analysis has established resolution bounds on meanings of integration with raster datasets and distance bounds for vector data. The theoretical approach has used a combination of theories on cartographic transformation and generalization, such as Topfer’s radical law, and additional research concerning optimum viewing scales for digital images to establish a set of guiding principles for integrating data of different resolutions.

An Assessment of the Effects of Cell Size on AGNPS Modeling of Watershed Runoff

This study investigates the changes in simulated watershed runoff from the Agricultural NonPoint Source (AGNPS) pollution model as a function of model input cell size resolution for eight different cell sizes (30 m, 60 m, 120 m, 210 m, 240 m, 480 m, 960 m, and 1920 m) for the Little River Watershed (Georgia, USA). Overland cell runoff (area-weighted cell runoff), total runoff volume, clustering statistics, and hot spot patterns were examined for the different cell sizes and trends identified. Total runoff volumes decreased with increasing cell size. Using data sets of 210-m cell size or smaller in conjunction with a representative watershed boundary allows one to model the runoff volumes within 0.2 percent accuracy. The runoff clustering statistics decrease with increasing cell size; a cell size of 960 m or smaller is necessary to indicate significant high-runoff clustering. Runoff hot spot areas have a decreasing trend with increasing cell size; a cell size of 240 m or smaller is required to detect important hot spots. Conclusions regarding cell size effects on runoff estimation cannot be applied to local watershed areas due to the inconsistent changes of runoff volume with cell size; but, optimal cells sizes for clustering and hot spot analyses are applicable to local watershed areas due to the consistent trends.