Visitors to the scenic grasslands of the Norfolk Broads in Eastern England will be enjoying its diverse wildlife and tranquil pastures for years to come thanks to a high-tech conservation program that is restoring traditional farming methods to environmentally sensitive areas of the United Kingdom.
      The conservation program is closely observed and supervised with a land cover monitoring system that bridges the gap between raster and vector data. Raster-based satellite images and vectorized aerial photos are crucial to the program because they provide an accurate and cost-effective means of detecting changes in land conditions and measuring the effectiveness of conservation schemes.
      Integration of raster and vector data in the monitoring system was made possible through use of the ERDAS IMAGINE Vector Module, which is based on the ARC data model. This tool allowed project designers to incorporate land cover information from recent satellite imagery with the base-line data from airphoto vector maps. The ability to integrate the two types of data has improved the economics and overall viability of the monitoring program.

Protecting the Environment
The Ministry of Agriculture, Fisheries and Food (MAFF) in England first began designating regions like the Norfolk Broads as Environmentally Sensitive Areas (ESA) in 1987. Conservation programs were devised because modern farming methods were adversely changing many of the naturally and historically important landscapes in England and Scotland.
      Twenty-two ESAs have since been designated, and farmers within those areas are encouraged with financial incentives to participate in a voluntary preservation program. Participating farmers use less intensive farming methods to protect and enhance the natural value of existing grasslands. In some ESAs, the farmers are even encouraged to convert arable farmland back to its natural grassland state.
      The Norfolk Broads, for example, is a marshy grassland that has long been a major cattle grazing area as well as a popular tourist site. The scenic meadows with their abundant wildlife draw numerous visitors who cruise the lakes and canals in boats.
      Modern improvements in drainage techniques, however, allowed farmers to fill in boundary ditches and cultivate many hectares of what had been poorly drained permanent grassland, explained Jacqueline Slater, a senior consultant with ADAS, the MAFF agency charged with monitoring the conservation program. Change from grassland to arable fields was lessening the appeal of the scenery and diminishing the wildlife.
      Norfolk Broads farmers who participate in the conservation program have returned to a more traditional farming regime with restrictions on drainage and the use of fertilizers, pesticides and herbicides which may harm wildlife. In the first four years of the program, 500 hectares of permanent grasslands have been created or restored in the Broads.

Establishing a Monitoring System
ADAS gauged the success of the conservation schemes by interpreting aerial photography and then creating vector maps from them. One of the primary objectives is to identify land areas within the ESAs that have changed from arable fields back to natural grasslands or vice versa. In 1993, Remote Sensing Applications Consultants (RSAC) of Medstead, Hampshire, was contracted to test the suitability of monitoring ESAs using satellite imagery and to develop a digital map updating system.
      Although each ESA had been photographed from aircraft for creation of a baseline set of vector land cover maps shortly after its designation, raster satellite imagery was found to be a more affordable means of identifying land changes and updating the maps.
      "Satellite imagery costs significantly less than airphotos for large areas," said Robert Brown, an RSAC consultant. "And digital satellite data contains additional information that can be optimized through enhancement techniques."
      Rather than purchase both an image processing system to analyze satellite imagery and a separate vector package to update the maps, RSAC built the monitoring system around ERDAS IMAGINE software equipped with the IMAGINE Vector Module. RSAC chose this software because it needed an image processing package with an open architecture robust enough to perform complex image enhancement functions, but user friendly enough for the non-technical staff to use, explained Brown.
      RSAC was expecting to customize the image processing package so the ADAS staff could perform most analysis procedures in a simple point-and-click interface but found the existing interface provided everything. They did, however, use the IMAGINE Developers' Toolkit to create customized map updating and change analysis utilities.
      The IMAGINE Vector Module is a processing tool developed by ERDAS Inc. of Atlanta, Ga. in cooperation with ESRI of Redlands, Calif., to enable ERDAS users to import vector data for integration, updating and manipulation within the image processing system. Although the user works on the vectors and their attributes in the raster image processing environment, the vectors remain in the ARC format, so there are no file conversion or translation errors.
      ERDAS incorporated many standard ARC/INFO functions into the IMAGINE Vector Module which enable the user to draw and edit vectors and polygons on the screen on top of the raster image. Instead of working with command lines, the user accesses the vector codes with the point-and-click interface of the image processing system.

Monitoring the Broads
"We wanted to keep the monitoring system simple," said Mike Wooding, RSAC's principal consultant, explaining why most of the change detection analysis is conducted visually on the display screen.
      RSAC obtained a variety of Landsat, SPOT and ERS-1 radar imagery acquired over the Norfolk Broads in years following the initial aerial survey. The satellite images, subjected to minor enhancements, were displayed on the image processing screen. Then the IMAGINE Vector Module was used to overlay the baseline vector maps onto the satellite images to see how much the land had changed in the intervening years.
      Visual change analysis was easier and more accurate when certain land cover classes were highlighted and others were masked on the vector map. RSAC found it was most useful to mask out either arable lands or grasslands so the analyst could focus on detecting change in one land cover type at a time.
     For instance, with all arable lands masked on the vector map, the analyst was then looking through the vector polygons only at those areas in the satellite image that had been grasslands at the time of the air photo survey. In theory, all grasslands in the image should have the same general appearance. Grassland areas that had changed, possibly to farmland, had anomalous appearances which were easy to spot on the screen. The same process was then conducted with grasslands masked.
      "We updated the vector maps right on the screen, dividing polygons with vectors and drawing new polygons to represent the changes in land cover," said Wooding.
      Working with the vector tool, the analyst colorized polygons to highlight changes on the vector maps and reassigned land cover attribute information as necessary.

Tapping into ERDAS IMAGINE
Extensive field work was required in the early phases of monitoring to ensure that land changes were being accurately identified. Field checks revealed the change detection procedure was identifying some important aspects of local farming practices. Several areas identified as farm fields in the vector maps later appeared covered with natural grasses in the satellite imagery. But not all of these were permanent grasslands. Many were fields sewn with grass for a year as part of crop rotation.
      This was one of several situations where RSAC needed to tap into the analysis capabilities of the ERDAS IMAGINE image processing system.
      The researchers obtained three ERS-1 radar images which had been acquired over one of the ESAs in three successive autumns from 1991 to 1993. These images were analyzed in a multi-temporal change detection routine to distinguish permanent from temporary grasslands.
      Each radar image was loaded into one of the three RGB color guns in the system for display as a composite on the screen. Since grasslands appear dark in single-band radar imagery and bare farm fields have bright reflectance, the researchers knew permanent grassland would yield a dark signature in the composite image as well. A red, green or blue tone indicated a bare farm field during at least one year, while mixed bright or white tones signaled a field that was bare in all years.

Assessing the Results
ADAS uses the land cover information gathered through the monitoring system to determine which land changes can be credited to their conservation strategies and to estimate which strategies work the best. The Ministry of Agriculture, Fisheries and Food receives this information in regular reports that help to fine tune the program.
      As the monitoring program progresses, RSAC and ADAS will conduct more involved image processing studies with the satellite imagery to differentiate among many grassland types and conditions. This will prove critical in judging the conservation program since many preservation schemes aim at improving the vigor of existing grasslands.

About the Author:
Kevin P. Corbley is a freelance writer specializing in remote sensing, digital mapping, GIS and GPS. He is located in Lakewood, Colo., and can be reached at 303-987-3979.

Back