Monitoring Agricultural in using the Vegetation Drought Response Index (VegDRI) Catherine Champagne 1,Jesslyn Brown 2, Tsegaye Tadesse 3, Trevor Hadwen 1, Andrew Davidson 1, and Richard Warren 1. 1 Agroclimate, Geomatics and Earth Observation Division, and Agri-Food Canada 2 Earth Resources Observation and Science Center, U.S. Geological Survey, Sioux Falls SD, USA 3 National Drought Mitigation Center, University of Nebraska, Lincoln NB, USA Poster ID 81191: GC13H: Sustainable Global Agricultural Production Monitoring Practices and Methods II Posters Impact of Data Inputs on VegDRI Model Development The Vegetation Drought Response Index (VegDRI) • Drought conditions vary tremendously from place to place and week to week. Accurate drought monitoring is • Cubist models were trained two ways: (1) Using only essential to understand a drought, its progression and potential effects, and to provide information to support Canadian climate data and (2) using Canadian and U.S.

drought mitigation decisions. Monitoring is improved by integrating information that is timely and region specific to 2001 climate data identify where and when they are happening. • Comparisons of (1) and (2) showed agreement over • The Vegetation Drought Response Index or VegDRI is a hybrid drought monitoring and mapping tool that integrates 71% of the pixels, with 93% within one VegDRI category satellite observations of vegetation status and climate data with information on land cover, soil characteristics, and and 98% within two categories other environmental factors. This allows for more precise identification of anomalies in satellite-measured vegetation • Canada has far fewer climate stations with long health that are directly related to drought 2002 historical records; this could be improved by gap filling • Developed by the U.S. Geological Survey (USGS)'s Earth Resources Observation and Science (EROS) Center and historical stations using gridded climate data sets the National Drought Mitigation Center (NDMC), VegDRI reveals vegetation conditions as plants respond to solar • Discontinuities in national data sets were found between energy, and other limiting factors. Researchers used integrated VegDRI products to produced detailed the two countries (soils data in particular), and these VegDRI maps that show levels of drought stress on vegetation across the conterminous United States. With a should be minimized to have a seamless and consistent relatively high degree of spatial detail, VegDRI maps support near-real-time monitoring of drought effects at state 2008 mapping across borders and county levels. These maps combine the higher spatial resolution of the satellite data (<1 square kilometer) with the sparser climate station information to provide a detailed picture of drought impacts on vegetation. Validation of VegDRI Maps for Canadian Agricultural Regions Late June Late July Late August Late June Late July Late August Developing a VegDRI for Canadian Agriculture • In 2013, a pilot study was conducted along the Canada-US border areas to assess seamless integration of data from Canada and the US for VegDRI 2001 • Climate data from 881 stations with long historical records in border regions of Canada and the US were used (77 in Canada, 804 in the US) to calculate two drought indices: the Self Calibrated Palmer Drought Severity Index (PDSI) and the Standardized Precipitation Index (SPI) . A 36 –week SPI was chosen after evaluation of SPI for numerous time scales since the provided the best predictive accuracy. • Satellite data from the Advanced Very High Resolution Radiometer (AVHRR) sensor were used to calculate bi- 2002 weekly composited Normalized Difference Vegetation Index (NDVI) over a 20 year period from 1989 – 2008. Two metrics were calculated from the time series NDVI: the Start of Season Anomaly (SOSA) and the Seasonal Greeness (SG) . • A database was built using the dynamic climate and satellite derived data in additional to biophysical data (soil available water holding capacity, land cover, , ecozones) to train a model for VegDRI in Canada using a Classification and Regression Tree (CART) method in Cubist software. 2008 • Models classify each pixel into one of eight VegDRI categories, ranging from Extreme Drought, through to Normal,

through to Extreme Moist (Figure 1). Southern Southern Ontario

Top: Southern Alberta region in 2001 shows accelerating drought impacts over the season resulting from a dry winter Top: In 2001 in Southern Ontario, conditions were mixed during the growing season, with wet conditions in which led to low soil moisture reserves, lack of germination due to dryness and high winds led to reseeding of fields in early spring, and drier conditions emerging in mid to late summer. This is evident in the increased drought early June and some late June rains. Low precipitation in July and pest infestations led to high crop losses, low stress seen in areas later in the season. irrigation water reserves; crop yields were 20-40% below average; dry fall led to early harvest Middle: Crop reports describe above normal precipitation in the spring of 2002, particularly in eastern Middle: Southern Alberta region in 2002 had a cool wet spring, flooding in southern region, some fields left unseeded Ontario. Conditions became hot and dry as the summer progressed. Maps show some localized impacts of due to excess moisture. Dry conditions led to crop deterioration in July, which is not evident in the VegDRI maps. Cool hot weather, which is consistent with crop reports. damp conditions in September with average yields. Damage from long term drought not evident in maps. Bottom: In 2008 in Ontario conditions were generally favourable with persistent wet weather and high yields Bottom: Southern Alberta region in 2008: Cool weather in early June, with adequate moisture reserves for seed across the region. Maps show excess moisture, which appears to be incorrect since the moisture in this germination; some crop damage mid season due to hail storms; average yields; crops 10-14 days behind normal due case was beneficial to crop growth. Some disease related to moisture was reported, but this did not have a to cool spring, yields 29% above 10 year average. Maps show primarily normal conditions, consistent with crop reports significant impact on crop yields. Maps were validated against provincial crop reports that summarized seasonal climate, soil moisture and impacts. Next Steps • VegDRI model training and testing will be extended into Canada south of 60 °N using NDVI composites from the Moderate-resolution Imaging Spectroradiometer (MODIS) satellite compiled at weekly time steps, and an increased number of climate station with historical data records in Canada (~900 stations in Canada and ~800 in VegDRI Categories the United States) • New models will be trained using climate and satellite data which cover a period from 2000 – 2014. Challenges • Data on climate and soils is much less available north of 60 °N. These areas are largely uninhabited boreal forest and tundra Fig. 1 Overview of VegDRI model training and development Pilot Area for VegDRI Modelling with MODIS NDVI • The shorter growing season and prevalence of requires unique methods to calculate satellite variables such as Start and End of Season