What is the Connecticut Land Cover Map Series?
The Connecticut Land Cover Map Series (Version 2.3) consists of seven dates of land cover data (1985, 1990, 1995, 2002, 2006, 2010 and 2015) created from satellite imagery. Each dataset includes twelve consistently interpreted land cover categories. The data were produced in a way to insure easy comparison, especially for land cover change studies. The land cover data are available in digital format for use in Geographic Information Systems and can be used to produce a wide variety of map products.
Can I mix and match different versions (version 1, version 2.2, version 2.3, etc) of the land cover?
No! Each series was created to be used together. Every date of the Version 2.2 series is different from every date of the Version 2.3 series.
How does the Agricultural Field class compare to other estimates of Agricultural areas, such as the NRI?
Our Agricultural Field category includes areas that are under agricultural uses such as crop production and/or active pasture. It is also likely to include some abandoned agricultural areas that have not undergone conversion to woody vegetation. The USDA NRCS “Natural Resources Inventory” is similar, in that is based on Landsat imagery, but since the NRI is a national survey it is created by statistical sampling, with a larger likelihood of error than CLEAR’s processing of each and every pixel. CCL and NRI data are nonetheless fairly compatible, if you make allowances for differences in how the major land cover categories are defined. The USDA Agricultural Census data is collected via field office reports, mostly gathered via survey of farmers, and thus is not comparable to our research methods. Also, these data include all land included on farmsteads, not only agricultural fields but also forest, developed lands, and other land uses; thus it predictably quotes larger acreage statistics for comparable time periods.
Two change maps? I’m confused. What are the two change maps showing?
The purpose of these maps is to show areas that went from one of the “undeveloped” land cover classes (other grasses, agricultural field, deciduous forest, coniferous forest, non-forested wetland, tidal wetland, etc.) to the “developed footprint” classes of developed or turf & grass. Turf & grass is included because, most of the time, it is closely associated with development.
The two types of change map both show the same areas, as defined above, that have changed from undeveloped to developed during the 1985 – 2015 period. However, the color schemes of the two maps show slightly different aspects of the change - what it went to (change to) and what it came from (change from).
The Change to map is colored based on what the area became.
|Developed before 1985|
|Turf & Grass before 1985|
|Change to Developed between 1985 and 2015|
|Change to Turf & Grass between 1985 and 2015|
The Change from is colored based on what the area was.
|Developed before 1985|
|Turf & Grass before 1985|
|Agricultural Field to Developed between 1985 and 2015|
|Agricultural Field to Turf & Grass between 1985 and 2015|
|Forest* to Developed between 1985 and 2015|
|Forest* to Turf & Grass between 1985 and 2015|
|Other Classes to Developed between 1985 and 2015|
|Other Classes to Turf & Grass between 1985 and 2015|
*Forest is a combination of the classes deciduous forest, coniferous forest, forested wetland and utility right-of-way. The Right-of-way class was created by hand digitizing through forest areas only. Therefore, it is appropriate to lump these linear areas with forest for the 8-class series.
How do you calculate change?
In this kind of research, there are three ways to measure change as described in the table below. The percent change reported in the data tables the Your Town and Your Basin page is the Relative rate of change. This means that if there is a small area to begin with (little area in time 1, or 1985), then the relative rate is likely to be high (or high percentage). If there is a large area to being with (lots of area in time 1, or 1985), then the relative rate is likely to be low (a small percentage).
T stands for a time period, so T1 would be an earlier date than T2. In most data tables on this website, T1 is 1985 and T2 is 2006.
|Measure of land cover change||Unit||Calculation||Significance|
|Absolute change||area, such as acres||(acres T2 – acres T1)||Allows aggregation of total areal change across the same geographic areas (such as towns or basins)|
(noted as "change in percent")
|(% area T2 - % area T1)||Allows comparison between areas (such as towns or basin); relates to land cover indicators|
|Relative rate of change||percent
(noted as "percent change")
|(area T2 - area T1) / area T1||Gives feel for how quickly land cover is changing relative to 1985 baseline, within and between the same geographic area|
Why can’t I see my house on the land cover maps?
The land cover maps were produced by interpreting Landsat satellite images, which have a ground resolution of 30 meters or approximately 100 feet. At this resolution, the satellite sensor "sees" areas that are about a quarter of an acre in size. Within any quarter acre, there may be a number of different landscape features and what the satellite ends up "seeing" is the largest feature or the largest group of similar features. Chances are your house, especially in forested rural areas, will be "dwarfed" by the surrounding trees. This will cause the area to be classified as forest rather than developed. However, if your house is in a neighborhood where houses, driveways, sidewalks, etc. are the predominant landscape features, then it will fall into a "developed" land cover class. Thus, what gets mapped depends on what the predominant landscape features are.
How were the land cover data created?
The land cover data were produced from Landsat satellite images using a computer application called image processing. Landsat images are made up of millions of small squares called pixels. Each pixel represents an area on the ground that measures 30 meters by 30 meters. For each pixel, the Landsat image records the amount of reflected energy in 6 narrow bands of the electromagnetic spectrum (red, green and blue visible light, a near-infrared and two mid-infrared bands). Because landscape features reflect light differently, we can use reflectance data to identify areas of deciduous forest, coniferous forest, water, etc. This is where the image processing software is so helpful. Rather than you or me having to analyze data for each pixel, the software can do it much faster, and with the help of image analysts, can assign pixels to land cover classes based on differences and/or similarities in reflectance values.
Is the land cover data correct?
Of course we'd like to think that the datasets are error free but the reality is that there are misclassification errors. A great deal of time was spent to avoid and correct errors. A variety of ancillary data sources such as USGS topographic maps and orthophotos, were used to aid the classification process. An accuracy assessment is under way so that we can quantitatively assess the accuracy of each land cover map. However, as of December 2008, we have not yet completed the accuracy assessment.
Why not use more detailed satellite imagery to make more detailed maps?
There are several reasons why we chose Landsat imagery for the Land Cover Project.
- A Landsat image is about 185 kilometers on each side and one image covers almost the entire state. One image, acquired on a cloud free day, would provide the bulk of the data for each of the five dates.
- Landsat imagery is relatively inexpensive or free.
- 30-meter resolution Landsat imagery was first collected in 1982 making it possible to create historic land cover data for change studies.
High resolution datasets were not used for the regional analysis for these reasons:
- Unmanageable file size.
- Too many scenes. Images captured from high resolution satellites have much smaller footprints (area covered on the earth) than Landsat images. It would be difficult to create a seamless dataset from so many images.
- Limited historical archive. Commercial high resolution imagery has only been available since the late 1990's. The limited archive would prohibit meeting the project objective - to derive a time series of land cover over a 20+ year period.
- Likelihood of seasonal and temporal variability. Because so many scenes would be required to cover the study area, it is likely that images would be captured on many different days, introducing artifacts such as variations in vegetation phenology and atmospheric effects (haze).
- Classification Methodology. Classification techniques applied to the Landsat images in this project do not work well on high resolution imagery. New techniques and software would need to be used. Currently, the software does not adequately handle extremely large datasets.
- High Cost. High resolution satellites are operated commercially and not by the government, they are expensive, especially for the huge number of images necessary to cover the area.
- Project Objectives. The objective of the project was to create a regional assessment. Detail available in high resolution images actually makes it more difficult to provide regional information.
Why doesn’t the data go back farther in time?
The first Landsat satellite was launched in 1972. However, it acquired imagery at 80-meter resolution, which is too coarse for this project, and in different bands of the electro-magnetic spectrum. Since we wanted to create land cover data sets that were comparable, we need to use spatially and spectrally consistent imagery. The first Landsat satellite to acquire 30-meter resolution imagery was launched in 1982.
Does the land cover data show land use?
No. With Landsat imagery it is possible to consistently and accurately determine what is on the land's surface but it's not possible to determine how the land is being used. For example, at 30-meter resolution, a group of homes, a group of farm buildings or a group of small offices may reflect light similarly. It's possible to classify the group as developed land but it's difficult, if not impossible, to determine reliably how the buildings are being used.
Does the land cover show all increases in developed land?
The land cover data certainly show an increase in developed land between any two time periods. However, it is likely that isolated or small development, especially within forested areas, may be missed due to the spatial resolution of the Landsat imagery. Therefore, we consider this to be a conservative measure of development. It's also worth mentioning that for the developed land cover class, the land cover data only show the actual developed potion of any particular site. In other words, if a 100 acre parcel is developed as a business park, only that portion that includes buildings, parking lots, roads etc. will be depicted as developed. Any vacant portion of such a site would be classified based on the predominant land cover - forest, non-forested wetland, turf & grass, etc.
Where can I get a copy of the land cover data?
>Any of the land cover data can be downloaded from the download portion of this website as Imagine raster files. All the data are in Connecticut State Plane Coordinates, NAD 83 and are in units of feet. To help CLEAR researchers understand how the land cover data are being used and to notify users via e-mail of new derived data products, we are requesting that you register prior to downloading the data. A simple easy-to-complete on-line form is provided for this purpose.
What is a HUC 12 watershed?
The maps and information included in this project (except for the summary maps) are organized by the watershed level known as Hydrologic Unit Code 12, or “HUC-12,” the smallest of the available basins delineated by the US Geological Survey. There are 195 HUC-12 basins within the study area, averaging about 30 square miles (19,256 acres) in area. They range from just over 3.5 square miles (2346 acres) to 62 square miles (39,655 acres).
HUCs Explained: The United States Geological Survey created a hierarchical system of hydrologic units originally called regions, sub-regions, accounting units, and cataloging units. Each unit was assigned a unique Hydrologic Unit Code (HUC). As of 2010 there are six levels in the hierarchy, represented by hydrologic unit codes from 2 to 12 digits long, called regions, subregions, basins, sub-basins, watersheds, and sub-watersheds.
Was an Accuracy Assessment completed?
Yes, the results of the accuracy assessment including the error matrices can be found on the Accuracy Assessment page.