The MODIS global land cover product (MOD12) consists of two suites of scientific data sets (SDS): MOD12Q1 (global land cover), and MOD12Q2 (global land cover dynamics). The MOD12Q1 SDS suite currently includes of a set of five internally consistent map layers depicting different global land cover classifications. The MOD12Q2 SDS suite provides a set of seven layers that characterize the land surface phenology of ecosystems at global scale. Both are produced locally at Boston University at a spatial resolution of 1-km, and are available for 2001-2004 at annual and biannual time-steps for MOD12Q1 and MOD12Q2, respectively.
In the past two years, significant progress has been made in regards to improving the quality of the MOD12Q1 algorithm. In addition to significant overall improvement to classification quality, key achievements include substantial augmentation and refinement of our land cover site database, and improved representation of global urban areas and wetlands. In regards to MOD12Q2, we have evolved our algorithm from what was essentially a research tool to one that is operational and able to produce globally reliable results. For both products we have performed extensive accuracy assessment and validation. Finally, we have implemented a suite of code changes in preparation for collection 5 processing, which is scheduled to begin soon. Key among these changes will be a shift from 1-km to 500-m data sets.
In this proposal, we describe two complementary sets of tasks. The first will consist of EOS Instrument Specific Algorithm Refinement and Cal/Val Activities, and will focus exclusively on MODIS. To this end, we propose a set of algorithm refinement and maintenance tasks that will support and improve upon the current MOD12 product for the next three years. Activities will include production of collection 5 maps at 500-m spatial resolution on an annual basis, and targeted research activities designed to address key outstanding weaknesses in the algorithms that are currently used to produce MOD12Q1 and MOD12Q2. The second objective will consist of Algorithm Refinement and Cal/Val for ESDRs activities, and will include a set of research and algorithm development tasks in support of developing Earth Science Data Records for both land cover and land surface phenology. Key objectives will be to augment the existing MOD12Q1 classification suite to include the FAO Land Cover Classification System (which has become a de-facto international standard), development of methods to distinguish classification errors from areas of real land cover change, and the use of higher temporal resolution data to estimate land surface phenology. As part of these activities we will explore data fusion approaches using multiple sensors to improve ESDR quality.
