Landuse classes, (“Uso do Solo”) is a critical category in evaluating the landscape.
Landuse classes, (“Uso do Solo”) is a critical category in evaluating the landscape. Vegetation land cover data represent both the classes of vegetation, and then the specific attributes associated with each class. For example, for VIC, the Veg_file. specifies, for each grid cell, a series of properties (rooting depth, LAI, etc). Veg_lib. For each vegetation type, the Veg_lib look-up table specifies the values for a series of attributes. At this scale, the main factor in quality might be the accuracy of the numerical values that assigned to each category. DHSVM has a similar structure.
Several data sources are available. The most recent is IEMA’07. This classification, completed and made available in early 2014, is based on a LANDSAT data for 2007. It includes the 24 primary classes of relevance for planning purposes.. As the most recent and most detailed product available, the IEMA’07 will be considered as the base frame of reference for the project.
Land use, from IEMA 2007
GEOBASE 1997 is an earlier product, based on LANDSAT from 1996 and 1997, thus representing the state at that time. Representing the GEOBASE classifications equivalent to IEMA 2007 makes it possible to see the evolution in landuse.
Landuse comparison, from GEOBASE 1997(6) left relative to reference IEMA 2007 (right)
Espírito Santo State and Minas Gerais
It is important that coverages are the "same" across the entire model domain. The problem is that IEMA/GEOBASE cover only Espírito Santo, while the model domain extends into neighboring states, especially for the Rio Doce. A “consistent” data layer is needed for the VIC modeling. Ideally, there are comparable products available in especially Minas Gerais that would allow extension of the IEMA-based Uso domain across the entire model domain, but at this time such coverage is not available.
Landcover derived from the MODIS satellite does, however, cover the whole region (with attention being paid to capture the heterogeneity of the regional landscape). The MODIS data were derived from: MODIS 2003 (MOD12Q1 Land Cover Product - MODIS/Terra Land Cover 96 Day L3 Global 1 km ISIN Grid - IGBP land cover classification) tiles acquired through the NASA ECHO website: http://www.echo.nasa.gov/index.html. The MODIS tiles were mosaicked for the region Aggregating the native MODIS resolution into the 4-km scale results in a loss if local heterogeneity, but the basic features are retained. The vegetation patterns in the upper right are consistent with the eucalypt signal.
MODIS landcover, by MODIS classifications, at native 1/120o (top) and aggregated into the ~4 km or 1/24o model resolution.
The next question is, how well does MODIS represent “reality, as so defined by IEMA’07? Re-classing IEMA to the classes of MODIS, and comparing visually, the base patterns are certainly maintained. The aggregation of finer into the coarser scales subsumes some of the more local heterogeneity.
Comparison of IEMA 2007 classifications (recombined to MODIS classes)
The possibility of a “merged” product was evaluated, whereby Uso do Solo would be used for Espírito Santo, and MODIS for outside of the state. But vegetation classifications (e.g. primary and secondary forest cover, cut forest, forest regrowth etc) and timing are different, which complicate a merge. Finally, in the spirit of efficiency, the decision was made to base landcover strictly on MODIS for the launch of the modeling. Over the next some months, a classification based on the most updated, combined Uso do Solo can be developed, and substituted later on, if necessary.
Bacias Rio Jucu and Rio Santa Maria da Vitória
Greater land use class specificity is needed for Rios Jucu and Santa Maria da Vitoria than for the state-level VIC coverage. The reference coverage for the can be extracted from the state-level product.
IEMA 2007 classifications for the Rios Jucu and Santa Maria de Vitoria
Selecting Classes for Hydrologic Modeling
It is difficult to get a sense of specific classes from such a map representing all classes, especially to evaluate how representative they are for hydrology modeling. By mapping each class separately, the relative distribution of each can be evaluated, together with the assessment of how (economically) important a particular class might be.
IEMA 2007 mapped by individual class for Rios Jucu and Santa Maria de Vitoria.
From this analysis, 8 primary classes are dominant, and will be the base for modeling.
Primary classes from IEMA 2007, as used for DHSVM modeling for the Rios Jucu and Santa Maria de Vitoria
The next issue is, how has landuse change affected the landscape? The GEOBASE 1997 dataset nominally represents conditions in 1997. The dataset itself has fewer and different classes. Here is the dataset coded to thee equivalent classes for IEMA’07. To compare the 2 products, at least visually, the IEMA’07 classes were set to match the GEOBASE. There would appear to be a significant change over the 2 periods.
Comparison of Uso do Solo, between GEOBASE 1996 and IEMA 2007. Colors are by comparable classes.
Datasets from FpV
In addition to the updated landuse, the Floresta para Vida (FpV) project has related datasets available.
Multi-data layers from the FpV projectMulti-data layers from the FpV project