Monitoring and Predicting General Evaluation: Predicted vs. Observed SSG
Vegetation Condition Using Climate, What is VegOut? Figure 3
Satellite, Oceanic, and Biophysical Data The Vegetation Outlook (VegOut) is a new experimental tool that provides future outlooks of vegetation conditions (seasonal Vegetation Condition
greenness) based on an analysis of: 1. climate-based drought index data (PDSI & SPI); 2. satellite-based vegetation condition Extreme stress
Tsegaye Tadesse1, Brian D. Wardlow1, and Jae H. Ryu1 information (standardized seasonal greenness from NDVI); 3. biophysical characteristics (e.g., land cover type, ecoregion Severe stress
Drought Mitigation Center, School of Natural type, irrigation status, and soil available water capacity); and 4. oceanic indicators (e.g., Multivariate El Niño/Southern Fair (near normal)
Resources, University of Nebraska-Lincoln, NE 68583-0988 Oscillation index, MEI). Good vegetation
Very good vegetation
Out of season
The complexity of drought characteristics and the diverse temporal and
spatial climate-vegetation interactions make monitoring drought impacts on (a) 2-week outlook (a) Observed SSG
vegetation very challenging. Improved meteorological observations and
new analytical methods coupled with recent advances in satellite-based Figure 3: (a) Two-week Vegetation
remote sensing offer great potential to improve our ability to monitor the Outlook (VegOut) map that predicted
impact of drought on vegetation. Such information can be utilized in
drought early warning systems. In addition, recent studies have found
SSG values for the bi-weekly period
significant improvements in seasonal climate predictions when ocean- Figure 1 ending on September 4, 2006; (b) bi-
atmosphere relationships are considered, and such teleconnection weekly SSG observed for the period
information should be integrated into drought-related vegetation condition ending September 4, 2006; (c) a
monitoring and prediction. difference map comparing the
Traditional climate or satellite-based vegetation index (VI) data has predicted vs. observed SSG values
formed the basis for most drought monitoring tools for vegetation. (i.e., VegOut minus the observed
However, new methods that integrate both types of data to leverage the Legend
strengths of both approaches have the capability to provide more accurate Under predict
Note: If the difference is < 1 standard deviation (SD), it is
and reliable information regarding drought-related vegetation conditions. Over predict classified as similar; otherwise it was labeled as “under
Recent studies have shown that data mining techniques are effective for (c) Difference map predict” if the SD was < -1 or “over predict” if the SD was >
integrating diverse, large, and often complex data sets and identifying +1.
hidden patterns within these data to investigate complex relationships
among many variables related to phenomena such as drought. Data mining Current and Future Works
techniques can be used to analyze climate, satellite, and biophysical data in • At present, the VegOut uses rule-based regression tree models that are generated
an effort to assess the current drought stress on vegetation and also to by identifying relationships between satellite-derived vegetation conditions,
predict future conditions based on historical patterns in these data. climatic drought indices, oceanic indices, and other biophysical data.
In this study, a new approach for identifying and predicting the spatio- • Alternative modeling techniques including association rules and neural networks
temporal patterns of drought and its impact on vegetation is presented. A are being investigated to compare with the current VegOut models. Ensemble
regression tree modeling technique was applied to a 17-year time-series techniques that base predictions on the results from multiple data mining
record of climate and satellite-based VI data and other biophysical techniques are also under consideration.
information (e.g., soil characteristics and land cover type) to identify • In addition, new inputs into the current VegOut models are also being
historical relationships and patterns among these variables that are similar investigated in an effort to provide more accurate predictions of future vegetation
to currently observed conditions, which are then used to predict the general conditions. The current VegOut research is focusing on the development of 2-, 4-,
vegetation conditions at several time steps into the future (i.e., 2-, 4-, and 6- and 6-week vegetation outlooks in the U.S. Great Plains, but expansion of VegOut
weeks in advance). This new drought monitoring tool is called the to other areas of the U.S. is planned in the near future.
Vegetation Outlook (VegOut).
• Researchers are selecting the best predictive variables, using higher correlation
and integrating the best climate and/or oceanic variables that correlate with
VegOut maps are produced using rule-based regression tree models that Figure 2
were generated to identify similar historical relationships (patterns) in space Figure 2. Two-week Vegetation Outlooks vegetation condition to produce an improved drought monitoring tool (VegOut).
and time between satellite-derived vegetation conditions, climate-based (VegOut), which predict SSG values, are • VegOut information will be provided to enhance the U.S. Drought Monitor.
drought indices, oceanic indices, and biophysical data. The data used to presented for: • Spatio-temporal drought monitoring and predictive information will be provided
produce the VegOut maps include Standardized Seasonally integrated through a web-based client-server delivery system to agricultural producers and
satellite vegetation Greenness (SSG); climate drought indices such as the (a) spring (period 11: May 21 – June 3), decision makers, and a fully operational, web-based drought decision support
Standardized Precipitation Index (SPI) and Palmer Drought Severity Index system is being developed.
(PDSI), oceanic indices that include the Southern Oscillation Index (SOI), (b) mid-summer (period 16: July 30 – August
Multivariate ENSO index (MEI), Pacific Decadal Oscillation (PDO), and 12) , and • Semi-operational maps are planned for the 2008 growing season
Atlantic Multi-decadal Oscillation (AMO); and biophysical parameters such
as land cover type, available soil water capacity, percent irrigated farm (c) fall (period 18: August 8 – September 9) Summary
land, and ecoregion. Because the models can be applied iteratively with phases of the 2006 growing season. • The VegOut is a new drought monitoring tool that provides outlooks of
input data from previous time periods, the method can be used to predict general vegetation conditions.
vegetation conditions later in the growing season based on information Observed SSG values and patterns for • VegOut integrates climate information and satellite-based observations of
about prior conditions in the year. An overview of the VegOut methodology periods 10 (early growing season: May 7 – current vegetation conditions with oceanic index data and other biophysical
and examples of the regional-scale VegOut maps are presented and future
work tasks are highlighted.
20), 11, 16, and 18 are presented in (e) information about the environment to produce 1-km resolution maps of
through (g), respectively. projected general vegetation conditions.
This work was funded by the United States Department of Agriculture Risk
Management Agency’s Partnership Agreement 05-IE-0831-0228.
For further information contact:
Dr. Tsegaye Tadesse Dr. Brian Wardlow
National Drought Mitigation Center National Drought Mitigation Center
Climate Prediction Applications Science Workshop (CPASW)
University of Nebraska-Lincoln University of Nebraska-Lincoln
March 4 - 7, 2008 Chapel Hill, North Carolina, USA. Telephone: (402) 472-3383 Telephone: (402) 472-6729
Email: firstname.lastname@example.org Email: email@example.com