Climate change is increasing the frequency and duration of spring precipitation events in the US Midwest, overwhelming tile drains and causing intermittent flooding events. These events, when they occur in agricultural fields, reduce vegetation cover as crops intolerant of flooding die or fail to thrive. Identifying the spatial and temporal extent of ponding, however, remains a challenge. One approach is to identify these areas with Landsat imagery near-infrared (NIR) and short-wave infrared (SWIR) bands. Vegetation is highly reflective in NIR, while soil is highly reflective in SWIR. Using classification algorithms on mid-growing season imagery, areas where flooding events have occurred in agricultural fields can be identified and quantified. This analysis focused on the 2013 growing season in the US Corn Belt, a particularly wet year which suggests an upper limit for flooding damage. Cultivated areas were analyzed with a trained Classification and Regression Tree (CART) algorithm in Google Earth Engine.
Results/Conclusions
The classifier algorithm of Landsat 8 imagery from the 2013 growing season in the Corn Belt indicates that 13.4% of cultivated land is classified as soil, meaning that either flooding damage was severe enough to denude those portions of the fields or that the soil was saturated to a point where planting was severely delayed or abandoned. 43.3% of pixels were classified as mixed soil-vegetation, indicating reduced canopy cover and likely occurring from wet soils compromising plant growth. The classification analysis indicates that in a wet year, over half of cultivated land cover may be negatively affected by increased precipitation. The new hydrological regime under climate change is also likely to affect the net greenhouse gas flux balance of the region. Further work will include drone mapping of flooding events at higher spatial and temporal scale than what is possible through satellite remote sensing products.