1. Variations of tree-ring width in oaks of the upland study forest are closely related to precipitation and PDSI for the previous year (July-June). Thus, tree-ring width is a direct indicator of current climatic conditions that affect growth of wood in the body of the tree.
2. The initial year of a drought results in a slight decline for NDVI values in the upland forest, whereas the bottomland forest experiences a slight increase for NDVI values. The latter is a result of greater understory growth infilling dry hollows and potholes. The second year of a drought leads to significant declines of NDVI values for both upland and bottomland forests. Minimal NDVI values are achieved the year after the end of a drought.
3. Variations in Landsat-derived NDVI values lag one or two years behind climatic events, because forest canopy (leaf coverage) depends on food stored in trees from previous years. Thus, NDVI values represent antecedent climatic events. Forest canopy, as portrayed by NDVI values, does not relate to current climatic factors, nor do NDVI values correspond to tree-ring growth for the same period.
4. Most year-to-year changes in NDVI values occurred on the margins of the two study forests, while forest interiors remained relatively stable during the period of study. Changes along forest boundaries are presumably the results of microclimatic stress for both forests, as well as impact of Missouri River flooding on the bottomland forest. Human activites also may have affected some forest edges.
5. NDVI values from 1987 to 1997 are slightly down in both study forests and the whole study scene. This is probably a reaction to decreased precipitation throughout the study period compared to the long-term average.
6. Residual values for the TM band 5:4 ratio are a reflection of changes in leaf moisture. The band 5:4 residual values are tied to both drought and flood episodes, so their climatic significance is difficult to interpret. The upland forest displays consistently higher band 5:4 ratios than does the bottomland forest. This relationship could be useful for identifying and classifying similar forests in other locations.
7. The upland study forest has a rough canopy in which meter-sized shadows are widely scattered. These shadows are not visible in TM images, but may likely depress individual band values. Because band ratioing (NDVI and 5:4 ratio) tends to remove such variations in brightness, and canopy roughness changed little during the period of study, the influence of canopy roughness is considered insignificant for this study.
8. Controlled burning of the bottomland prairie (April 2000) had a marked impact along the margin of the bottomland forest. Cottonwood trees on the forest edge were killed or injured by the fire; these trees were probably more susceptible to fire damage because of drought conditions.
Kansas Geological Survey
Web version December 2001
http://www.kgs.ku.edu/Current/2001/aber/aber12.html
email:lbrosius@kgs.ku.edu