CLIP – Linkages between Land Cover and Regional Climate Jiaguo Qi presenting on behalf of the LC Group (Alphabetically: Jianjun Ge, David Lusch, Joseph Maitima Jennifer Olson, Robin Reid, Nathan Torbick, Jing Wang, Lijian Yang) Michigan Sate University Qi, Lusch, Reid, Maitima, Olson, Torbick, Ge

LC Objectives To assess land use/cover change impact on regional climate To better characterize LC as input to regional climate model (RAMS) To quantify the LC requirement for regional climate modeling. Specifically to assess the effect of land cover classification accuracy on regional climate simulation. This links the human land use to the regional climate To assess the degree of LCC impact on regional climate Categorical changes Continuous changes

Data LU/C Biophysical Variables Scales GLC2000, Africover, IGBP, MODIS LAI, EVI, DEM, Albedo, LST, and Precip. etc. Scales Continental – Case Study/Field-Observations

Nairobi LULC products range in data sources, objectives, classification methods

Assessments and Developing LULC Scheme Multiple LULC assessments and evaluations Strengths and weaknesses Videography over selected ecological gradients Q-statistic (uses LAI as ‘evaluator’) Develop LULC Schemes for models Create hybrid CLIP Cover LU from Africover, LC from GLC2000 Crosswalk to LEAF2, biophysical variables

LU/C Videography Assessment Sample Points Aquatic grasslands Rainfed herbaceous Bare rock Urban LU/C Videography Assessment Sample Points

LAI from MODIS June 2001 March 2001 Sept 2001 Dec 2001

Climate Model In this study, Regional Atmospheric Modeling System (RAMS) Version 4.3 is used (Pielke et al., 1992). It is fully three-dimensional, nonhydro-static; includes interactive nested grid capabilities, supports various radiation, initialization and boundary condition schemes.

Methods RAMS MODEL Simulations RS Biophysical Parameters LULC Land – Atmosphere Feedback Crosswalk to Predetermined Defaults Albedo Leaf Area Index Range of Leaf Area Index Vegetation Fraction Range of Vegetation Fraction RS Simulations

Land Cover Conversion or Classification Accuracy

Experiment Design Classification error was added to original GLC2000 dataset at random locations and by random predominant (five) classes, which was increased from 0% to 50% at 5% interval RAMS was run 11 times with different amount of errors in land cover. These runs are called R00, R05, R10, …., R50.

Land Cover Change – Continuous Change

Leaf Area Index LAI in RAMS was further improved by incorporating directly monthly 1km MODIS LAI images. LAI values were reduced to 50% at 5% rate. 11 runs were conducted. Kain-Fritsch without interior nudging were used.

March 15th RAMS new MODIS RAMS default LAI values (MODIS, RAMS, and MODIS/Interpolation). The default in this slide has already used the GLC2000. What if we use the original OGE?

The difference between simulated results from original (0% reduction in LAI) vs those that were reduced at difference percentage. Note that this is the absolute value differences (no negative numbers!)

Differences of two time series (maximum differences of the two time series)

Phenology at 0o, 5oNorth and 5o South

THANK YOU Questions?

DEM