Development of Land Use and Land Cover (LULC) Data for the
Houston-Galveston Area Council (H-GAC) Region
The image acquisition, processing, classification and accuracy assessment protocols that are used to produce the satellite based (LandSat 8) land use land cover dataset (LULC) using Remote Sensing (RS) techniques for the Houston-Galveston Area Council (H-GAC) planning region are described in this document. The data development is conducted by the Community and Environmental Planning Department (C&E) staff at H-GAC.
Land cover data is an important element in understanding the dynamics of the H-GAC region and applying for various analytical projects that are not limited to regional and county water quality planning and watershed analysis, land use and socioeconomic modeling applications, large area resource management planning, educational purposes for students and citizens, broad scale evaluation of impact analysis and change detection. Generally, H-GAC uses two sources of national land cover datasets for its geospatial analysis and visualization purposes. They are National Oceanic and Atmospheric Administration Coastal Change Analysis Program (NOAA C-CAP) and National Land Cover Database (NLCD) from United States Geological Services (USGS). These two datasets usually release in every five-year interval and take about 3-5 years’ time to release a current year’s data. For example, the most recent release of LULC dataset contains 2011 data and it was released in 2014. The next dataset will be 2016 data and it is expected to release in 2018/19. Therefore, H-GAC has decided to develop in-house LULC datasets using remote sensing techniques and satellite data on an annual basis to have most up-to-date information for the purposes of regional geospatial analysis and modeling. Requirement for this approach was to develop LULC dataset that includesland classes compatible with the two national datasets with an acceptable level of accuracy (Kappa greater than 70%) in annual basis using LandSat 8 satellite data.
The goal of this effort is tocreate an accurate land cover data that is current and easily integrated with other H-GAC’s data management and analytical tools. This effort will determine suitable land cover categories that most accurately represent the diverse nature of the H-GAC region. The categories developed will be based on concerns, issues, and activities that face the integrity of the area waterbodies, developments, agriculture, wetlands, vegetation and other environmental factors. The end product will be a regional data set showing the general distribution of land cover throughout the region.
- Geographic Area of Interest
The Area of Interest for this project includes the 13 county region of H-GAC and the additional areas (of San Jacinto and Grimes Counties) that comprise the four assessment basins for the Clean River Program (CRP) Region. The four basins are the Trinity-San Jacinto Coastal Basin, San Jacinto River Basin, San Jacinto-Brazos Coastal Basin and the San Bernard segments of the Brazos-Colorado Coastal Basin. The total area to be included in this data development is roughly 12,500 square miles or 8,000,000 acres.
- Expected Uses
The demand for current land cover data on a recurring basis is growing, especially in the areas of the region that are experiencing rapid change in land cover characteristics. Some potential applications for the data include:
- Locating various land cover types within a geographic boundary
- Determining acreage and/or percentage of each land cover type within a geographic boundary (for ex; a watershed)
- Detectingthe historical land cover changes and applying in predictive models
- Identify new developments and evaluate growth patterns of the region
- Applying in impact analysis such as water and air quality impacts fromurbanization
- Associating water quality characteristics with land cover types and spatial patterns.
- Land Cover Classification Scheme
The exact list of land cover types in the final products are not defined. However, H-GAC’s intention is to develop the datasets with classes consistent with NOAA C-CAP classification scheme. At this level H-GAC is planning to develop the datasets with 10 major land classes which may include multiple NOAA C-CAP classes in one H-GAC class. The target land cover classes were decided from level 1 of the NOAA C-CAP classification scheme.The C-CAP classification system is hierarchical, reflects ecological relationships, and focuses on land cover classes that can be discriminated primarily from satellite remote sensor data. It was adapted and designed to be compatible with other nationally standardized classification systems, especially the US Geological Survey (USGA), Environmental Protection Agency (USEPA) and the Fish & Wildlife Service. Each of the classes and subclasses listed are described in greater detail below.
3.1Developed, High Intensity (1) - contains significant land area is covered by concrete, asphalt, and other constructed materials. Vegetation, if present, occupies < 20 percent of the landscape. Constructed materials account for 80 to 100 percent of the total cover. This class includes heavily built-up urban centers and large constructed surfaces in suburban and rural areas with a variety of land uses.
3.2Developed, Medium Intensity (2) - contains area with mixture of constructed materials and vegetation or other cover. Constructed materials account for 50 to 79 percent of the total area. This class commonly includes multi- and single-family housing areas, especially in suburban neighborhoods, but may include all types of land use.
3.3Developed, Low Intensity (3) - contains areas with a mixture of constructed materials and substantial amounts of vegetation or other cover. Constructed materials account for 21 to 49 percent of total area. This subclass commonly includes single-family housing areas, especially in rural neighborhoods, but may include all types of land use.
3.4Developed, Open Space (4) - contains areas with a mixture of some constructed materials, but mostly managed grasses or low-lying vegetation planted in developed areas for recreation, erosion control, or aesthetic purposes. These areas are maintained by human activity such as fertilization and irrigation, are distinguished by enhanced biomass productivity, and can be recognized through vegetative indices based on spectral characteristics. Constructed surfaces account for less than 20 percent of total land cover.
3.5Cultivated (5) - contains areas intensely managed for the production of annual crops. Crop vegetation accounts for greater than 20 percent of total vegetation. This class also includes all land being actively tilled.
3.6Pasture/Hay and Grasslands (6) – This is a composite class that contains both Pasture/Hay lands and Grassland/Herbaceous
3.6.1Pasture/Hay - contains areas of grasses, legumes, or grass-legume mixtures planted for livestock grazing or the production of seed or hay crops, typically on a perennial cycle and not tilled. Pasture/hay vegetation accounts for greater than 20 percent of total vegetation.
3.6.2Grassland/Herbaceous - contains areas dominated by grammanoid or herbaceous vegetation, generally greater than 80 percent of total vegetation. These areas are not subject to intensive management such as tilling, but can be utilized for grazing.
3.7Forest (7) – This is a composite class that contains all three forest land types, and shrub lands.
3.7.1Deciduous Forest - contains areas dominated by trees generally greater than 5 meters tall and greater than 20 percent of total vegetation cover. More than 75 percent of the tree species shed foliage simultaneously in response to seasonal change.
3.7.2Evergreen Forest - contains areas dominated by trees generally greater than 5 meters tall and greater than 20 percent of total vegetation cover. More than 75 percent of the tree species maintain their leaves all year. Canopy is never without green foliage.
3.7.3Mixed Forest - contains areas dominated by trees generally greater than 5 meters tall, and greater than 20 percent of total vegetation cover. Neither deciduous nor evergreen species are greater than 75 percent of total tree cover. Both coniferous and broad-leaved evergreens are included in this category.
3.7.4Scrub/Shrub - contains areas dominated by shrubs less than 5 meters tall with shrub canopy typically greater than 20 percent of total vegetation. This class includes tree shrubs, young trees in an early successional stage, or trees stunted from environmental conditions.
3.8Barren Lands (8)–This class contains both barren lands and unconsolidated shore land areas
3.8.1Barren Land - contains areas of bedrock, desert pavement, scarps, talus, slides, volcanic material, glacial debris, sand dunes, strip mines, gravel pits, and other accumulations of earth material. Generally, vegetation accounts for less than 10 percent of total cover.
3.8.2Unconsolidated Shore - includes material such as silt, sand, or gravel that is subject to inundation and redistribution due to the action of water. Substrates lack vegetation except for pioneering plants that become established during brief periods when growing conditions are favorable.
3.9Wetlands (9) – this is a composite class that contains all the palustrine and estuarine wetland land types
3.9.1Palustrine Forested Wetland - includes tidal and non-tidal wetlands dominated by woody vegetation greater than or equal to 5 meters in height, and all such wetlands that occur in tidal areas in which salinity due to ocean-derived salts is below 0.5 percent. Total vegetation coverage is greater than 20 percent.
3.9.2Palustrine Scrub/Shrub Wetland - includes tidal and non-tidal wetlands dominated by woody vegetation less than 5 meters in height, and all such wetlands that occur in tidal areas in which salinity due to ocean-derived salts is below 0.5 percent. Total vegetation coverage is greater than 20 percent. Species present could be true shrubs, young trees and shrubs, or trees that are small or stunted due to environmental conditions.
3.9.3Palustrine Emergent Wetland (Persistent) - includes tidal and non-tidal wetlands dominated by persistent emergent vascular plants, emergent mosses or lichens, and all such wetlands that occur in tidal areas in which salinity due to ocean-derived salts is below 0.5 percent. Total vegetation cover is greater than 80 percent. Plants generally remain standing until the next growing season.
3.9.4Estuarine Forested Wetland - includes tidal wetlands dominated by woody vegetation greater than or equal to 5 meters in height, and all such wetlands that occur in tidal areas in which salinity due to ocean-derived salts is equal to or greater than 0.5 percent. Total vegetation coverage is greater than 20 percent.
3.9.5Estuarine Scrub / Shrub Wetland - includes tidal wetlands dominated by woody vegetation less than 5 meters in height, and all such wetlands that occur in tidal areas in which salinity due to ocean-derived salts is equal to or greater than 0.5 percent. Total vegetation coverage is greater than 20 percent.
3.9.6Estuarine Emergent Wetland - Includes all tidal wetlands dominated by erect, rooted, herbaceous hydrophytes (excluding mosses and lichens). Wetlands that occur in tidal areas in which salinity due to ocean-derived salts is equal to or greater than 0.5 percent and that are present for most of the growing season in most years. Total vegetation cover is greater than 80 percent. Perennial plants usually dominate these wetlands.
3.10Water (10) – This is a composite class that contains open water, and both palustrine and estuarine aquatic beds
3.10.1Open Water - include areas of open water, generally with less than 25 percent cover of vegetation or soil.
3.10.2Palustrine Aquatic Bed - includes tidal and non-tidal wetlands and deep water habitats in which salinity due to ocean-derived salts is below 0.5 percent and which are dominated by plants that grow and form a continuous cover principally on or at the surface of the water. These include algal mats, detached floating mats, and rooted vascular plant assemblages. Total vegetation cover is greater than 80 percent.
3.10.3Estuarine Aquatic Bed - includes tidal wetlands and deep water habitats in which salinity due to ocean-derived salts is equal to or greater than 0.5 percent and which are dominated by plants that grow and form a continuous cover principally on or at the surface of the water. These include algal mats, kelp beds, and rooted vascular plant assemblages. Total vegetation cover is greater than 80 percent.
5. Sensor and Imagery Requirements
The preferred satellite sensor system for this LULC data development is Landsat 8.LANDSAT 8 satellite has two main sensors: The Operational Land Imager (OLI) and the Thermal Infrared Sensor (TIRS). OLI collects images using nine spectral bands in different wavelengths of visible, near-infrared, and shortwave light to observe a 185 kilometer (115 mile) wide swath of the Earth in 15-30meter resolution covering wide areas of the Earth's landscape while providing sufficient resolution to distinguish features like urban centers, farms, forests and other land uses. TIRS collects images using two spectral bands from wavelengths of thermal radiation of earth surface. TIRS information are mainly used to support to track how land and water are being used. TIRS is a useful tool for managing water consumption.The following table provides the list of bands in LANDSAT8.
Band Name / Bandwidth (µm) / Resolution (m)Band 1 – Coastal Aerosol / 0.43 - 0.45 / 30
Band 2 - Blue / 0.45 - 0.51 / 30
Band 3 - Green / 0.53 - 0.59 / 30
Band 4 - Red / 0.63 - 0.67 / 30
Band 5 - Near Infrared (NIR) / 0.85 - 0.88 / 30
Band 6 - Short Ware Infrared (SWIR) 1 / 1.57 - 1.65 / 30
Band 7 - Short Ware Infrared (SWIR) 2 / 2.11 - 2.29 / 30
Band 8 - Panchromatic / 0.50 - 0.68 / 15
Band 9 - Cirrus / 1.36 - 1.38 / 30
Band 10 - TIRS 1 / 10.6 - 11.19 / 100
Band 11 - TIRS 2 / 11.5 - 12.51 / 100
In order to have accurate classification on natural lands, summer months (preferably early summer) are considered a suitable timeframe for our broad classification needs. However, a second set of imageries from winter/early spring months (leaf-off season) will be used in order to enhance the classification of develop land classes such as high, medium and low intensity developed lands.The acquired imagery will possess little or no cloud cover, with a maximum percent of cloud cover of 5-10%. Based on the size of a LandSatimage, our geographic area of interest will require four LandSatscenes for complete spatial coverage.The four scenes cover from two LandSat paths and two rows (i.e. path 25 and row 39, path 25 and row 40, path 26 and row 39, path 26 and row 40).
The LandSat 8 contains 11 bands total ranging from visual to thermal infrared wavelengths of the electromagnetic spectrum (see above table). Based on the reflectance characteristics, only seven bands will be used (Coastal band, Red, Green and Blue Visible bands, Near Infrared, and two short wave infrared bands) during the classification process for LULC data development.
6. Image Processing Techniques and Accuracy Assessment
The most significant effort of this data development will be devoted to image classification and the accuracy assessment of the classification. Prior to the classification process, several preprocessing steps has to be completed. Four Landsat 8 scenes should be required to cover the study area. Two dates for each Landsat scene, representing both the summer and leaf-off seasons, will be acquired from USGS ( An attempt will be made to obtain cloud free imagery, however, if clouds are present they will be masked to prevent confusion during the image classification. Each of the scenes will be registered to a USGS 7.5 minute quadrangle base map with an allowable error of 0.5 pixels. Multi-date imagery from both the winter season and summer season will be incorporated into the image classification to provide additional spectral reflectance data on phenological changes in vegetation growth and moisture variations that occur from season to season. Multi-temporal information on the spectral reflectance of our land cover classes should allow us to separate land cover such as different forest covers and capture spectral reflectance of residential areas that are normally hidden under deciduous canopy cover during the summer.
Reflectance values of dark and bright areas in the images, such as water bodies and high intensity developed areas will be used in a regression to calibrate the overall brightness of the image scenes. This method will preserve variation resulting from seasonal changes in phenological growth of vegetation while eliminating differences in spectral reflectance that may result from sensor offset and gain and variation in scene illumination. Each scene will then be input into a tasseled cap transformation algorithm to produce an image of greenness, brightness, and moisture. The tasseled cap transformation is a data reduction technique that isolates variation in spectral reflectance that is the result of physical structures in the image scene. The resulting greenness, brightness, and moisture bands for the two corresponding scenes will be combined to form a single multi-temporal dataset for input into the unsupervised and supervised image classification.
The Image Processing Techniques (see the flowchart) will involve an iterative process, using a hybrid, unsupervised and supervised classification methodology. Initiated with an Unsupervised Classification (UC), a common image processing routine (ISODATA) will serve primarily as an exploratory procedure to determine homogenous areas for the location of training sites for the supervised classification and to uncover dominant spectral signatures in the imagery.
Generation of training sites is the key component of developing accurate LULC data. In this analysis the training sites will be developed from expertized digitization of homogeneous sample areas using high resolution aerial imagery (6 inch and 1ft resolution H-GAC’s biannual aerial imageries), H-GAC’s parcel based land use data, and most recent other land use and land cover data types such as NOAA C-CAP and NLCD datasets. Overlaying all these datasets in GIS environment is one of most efficient ways to identify the unique samples of each land use and land cover categories. In order to have an accurate classification, a Region Of Interest file (ROI file) with a homogeneous distribution of training sample is essential. A 0.1 decimal degree square grid will be overlaid on each imagery as the guide to collect samples. The ROI will be developed to have at least one training site for each land class in one grid cell. This ROI development is usually performed in ArcGIS environment.