SATELLITE - DERIVED INFORMATION ON SNOW COVER FOR AGRICULTURE APPLICATIONS IN UKRAINE

SATELLITE - DERIVED INFORMATION ON SNOW COVER FOR AGRICULTURE APPLICATIONS IN UKRAINE Peter Romanov University of Maryland, College Park, USA Abstract This paper demonstrates how NOAA interactive satellite-based maps of snow cover can be used in the assessment of unfavorable agricultural conditions in Ukraine The focus was on two events, the extensive winterkill in winter of 2002-2003 and the drought in the early 2007 Both events had a strong adverse effect on the yield and on the production of major grain crops The analysis of NOAA daily snow maps has revealed an extremely short duration of snow in Ukraine in winter of 2006-2007 This is indicative of lower winter-time precipitation that contributed to the soil dryness in spring 2007 To identify potential crop freeze damage we have estimated the minimum temperature of snow-free land surface from snow charts combined with satellite land surface temperature retrievals Temperatures below -180C indicating potential winterkill were observed in the Central and Eastern Ukraine in December 2002 Introduction For over 35 years satellite observations have been used to monitor the global snow cover distribution and snow pack properties In the last decade a substantial increase in the number of satellite sensors for snow monitoring, as well as enhanced sensors capabilities and image analysis techniques have led to a noticeable improvement in the derived snow cover products This concerns the accuracy of the maps, their spatial resolution, coverage and the update frequency The urge to fully utilize satellite observing capabilities, in particular, their multispectral measurements and high spatial resolution, stimulated the development of automated algorithms to identify snow cover in satellite imagery and to generate maps of snow cover distribution Current operational automated systems derive snow cover from satellite observations in the visible and infrared spectral bands (e.g., Hall et al, 2002; Romanov et al., 2003), from observations in the microwave (Kongoli et al., 2004; Derksen et al., 2003), or use synergy of the two techniques (Foster et al., 2008; Romanov et al., 2000) The accuracy of snow identification and mapping by automated snow detection algorithms is high, often exceeding 90%, however the algorithm performance depends on the topography, vegetation cover, weather conditions, snow pack properties and other environmental factors (e.g Simic et al., 2004) As a result the accuracy of automated snow maps changes with the surface type and varies throughout the year Relatively short lifespan of satellite sensors, especially back in 1980s and 1990s, differences in their spectral bands and spatial resolution, degradation of sensitivity of some sensors with time and satellite orbital drift complicate developing consistent long-term snow datasets from these data Identification of snow in satellite imagery by visual inspection of satellite imagery is the oldest snow cover mapping technique Since 1972 this approach has been routinely used at NOAA to generate weekly maps of snow and ice distribution in the Northern Hemisphere In 1999 a computer-based Interactive Multisensor Snow and Ice Mapping System (IMS) was implemented to facilitate the image analysis by human analysts (Ramsay,1998) This allowed to improve the spatial resolution of the maps from 180 km to 24 km and to start daily snow mapping In 2004 the IMS system was upgraded and spatial resolution of the product was further increased to km (Helfrich et al., 2007) IMS maps of snow and ice cover are considered as the primary NOAA snow cover product and are incorporated in all global and mesoscale operational numerical weather prediction models run by NOAA National Centers for Environmental Prediction (NCEP) With over 35 years of continuous snow cover monitoring, NOAA interactive snow maps present a unique source of information for global climate change studies (Frei and Robinson, 1994) High spatial resolution and daily updates of IMS maps also make them potentially useful for various environmental and practical applications at regional and local scale In Ukraine information on the snow cover distribution and extent is of major importance for agriculture The length of snow season ranges from several weeks in the very south of the country to several months in the north Water accumulated in the snow pack and released through the snowmelt is critical for the winter crops development in early spring Snow pack is also an important factor preventing the frost and freeze damage of winter crops Most information on the snow cover distribution in Ukraine is obtained from ground-based meteorological stations Satellite observations of snow can complement surface observations by providing spatially detailed and frequent in time information on the snow cover distribution and seasonal change In this paper we examine daily IMS snow cover maps over Ukraine for ten winter seasons, from 1999-2000 to 2008-2009 The objective of the work was to demonstrate the efficacy of these maps for agriculture applications and in particular their usefulness in the analysis of spring-time drought conditions and for identification of areas of potential freeze damage of winter grains The primary focus was on the winter season of 20022003 when a severe winterkill destroyed about 55% of winter wheat in Ukraine and on the 2006-2007 winter season, which preceded the intensive drought in spring and summer 2007 Data In this study we have used NOAA daily snow and ice cover maps for the Northern Hemisphere generated within the IMS system Original IMS maps have polar stereographic projection with every pixel in the map corresponding to the land surface labeled as “snow-free land” or “snow” Pixels over water are classified as “ice-free water” or “ice” In contrast to automated snow products, IMS maps not contain “undetermined” pixels If persistent cloud cover prevents from reliably delineating snowcovered areas the analyst makes an intelligent guess regarding the possible change in the snow cover distribution beneath the clouds or retains the snow cover distribution from the previous day product IMS snow and ice charts for ten winter seasons (November to March) from 1999-2000 to 2008-2009 were acquired from the National Snow and Ice Data Center (NSIDC) For convenience all daily maps were regridded to a latitude-longitude projection with a grid cell size of 0.050 or about 5km The portion of the map within 43 -530 N and 190-410E completely covering Ukraine was extracted and saved Figure presents an example of the original IMS daily snow cover map for the Northern Hemisphere and a portion of this map over the study area in the latitude-longitude projection Fig IMS snow and ice cover map for January 31, 2009 (left) and its portion over the study area regridded to the latitude-longitude projection (right) Snow is shown in white, ice is light gray, snow-free land surface is medium-gray and ice free water is dark gray Application of IMS snow maps 3.1 Snow cover duration and drought Seasonal snow cover is an important hydrological and climate feature of mid and highlatitude areas Longer duration of continuous snow cover is most often associated with larger wintertime precipitation amounts and corresponds to larger moisture accumulation in the snow pack at the end of the cold season On the opposite, shorter snow cover duration means less moisture accumulation in the snow pack at the end of the cold season and, hence, more probable shortage of water supply for crops in spring In this study we processed all daily IMS snow cover maps over Ukraine to determine the duration of snow cover for every winter season since 1999 and average duration of snow cover for the last ten-year long period The duration of the snow cover for every grid cell was estimated by calculating the number of days the grid cell was labeled by IMS analysts as “snow covered” Maps of the yearly and multiyear-mean snow cover duration are presented in Fig Overall they demonstrate a general increase in the length of the snow season in the northern and north-eastern direction and it substantial variation from year to year Overall, the yearly duration of snow cover ranges from three to four weeks in the south of Ukraine to three to four months in the north Fig.2 Snow cover duration over Ukraine for winter seasons 1999-2000 to 2008-2009 and the mean snow cover duration for the ten-year period calculated from daily IMS snow cover maps Areas outside Ukraine are shown in white, water is black and snow cover duration is presented in shades of gray As it follows from the maps in Figure 2, the winter season of 2006-2007 was characterized by unusually short duration of snow cover Over most of Ukraine snow cover remained on the ground about twice less than normal This reduction in the snow cover duration and associated drop in the available melt water may have contributed to the shortage of water supply in spring and to severe drought in early summer 2007 According to the data of United States Department of Agriculture (USDA) the drought has affected Ukraine’s winter grain crops by noticeably reducing the yield of Ukraine’s two major winter grain crops, wheat and barley (see Fig 3) It is interesting that the decrease in the yield of winter barley in 2007 was much more pronounced than the corresponding decrease in the yield of winter wheat This difference may be explained by the fact that drought conditions were more severe in the east of Ukraine, where most barley is produced Yield, MT/HA As it is seen from Figure 2, the winter season of 2000-2001 presents another example of shorter than average snow cover duration In this year the largest reduction in the snow cover duration was observed in the south of the central part of Ukraine This anomaly, however, Barley, Yield (MT/HA) did not develop into a wideWheat, Yield (MT/HA) spread drought similar to the one 1998 2000 2002 2004 2006 2008 2010 in 2007 Lower than usual soil Year moisture content in agricultural areas in Ukraine was indeed Fig.3 Yearly yield of wheat and barley in Ukraine in 1999-2009 observed throughout most of the (Source: US Department of Agriculture) For the year 2009 winter months (USDA, 2001), projected yield values are given however above-normal early spring rainfall compensated for low winter precipitation As a result, winter-wheat and winter barley yields in 2001 were close record levels of the decade 3.2 Extent of winterkill Winterkill is one of primary reasons for the winter crop loss in Ukraine While in the winter-dormant stage winter wheat typically survives temperatures down to -17 0C to -180C (USDA, 2006) A combination of lower temperature and shallow or absent snow cover presents a potentially damaging condition that may result in the injury to the crop The most recent extensive winterkill occurred in Ukraine in December 2002 causing the country’s wheat production in 2003 to decrease by more than times from the previous year and by more than times from the annual average level In this study we have used daily IMS snow cover maps along with satellite-derived information on the land surface temperature (LST) to identify areas of possible winterkill Information on the land surface temperature was obtained from LST maps derived from observations in the infrared spectral band of Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Aqua and Terra satellites Daily daytime and nighttime LST maps from MODIS Aqua and Terra (MOD11C1 and MYD11C1 products, respectively) were acquired from NASA Distributed Active Archive Center (DAAC) Since clouds are opaque in the infrared, information on the LST is available only in clear sky conditions The projection of MOD11C1 and MYD11C1 products is latitude- longitude with a grid cell size of km, i.e the same as the projection of regridded IMS snow cover maps over Ukraine This feature of MODIS LST maps has allowed for an easy pixel-by-pixel spatial matching of the two products In this study we have used MODIS-based daily maps of the land surface temperature and corresponding maps of snow cover distribution to generate maps of the monthly minimum land surface temperature for snow-free land surface Both daytime and nighttime LST maps have been processed The minimum temperature maps for three winter seasons from 1999-2000 to 2001-2002 were derived from MODIS Terra LST products, whereas later on MODIS Aqua retrievals have been used The reason for the use of MODIS Aqua is that Aqua nighttime overpass time is later than Terra (1 AM vs 10 PM) and therefore its observations have a better chance to capture the nighttime minimum temperature Fig.4 presents maps of the minimum temperature of snow-free land surface for the month of December for nine years from 1999 to 2008 Three shades of gray correspond to surface temperatures above -17.50C, within -17.50C to -18.50C and below -18.50C The map for December 2002 clearly shows a large area of very low minimum temperatures of snow-free land surface in the central and eastern Ukraine indicating potential freeze damage There is much similarity between the spatial distribution of these areas and ground-based estimates of winter wheat damage as reported by USDA (see Fig.5) According to in situ data most affected were also the eastern and especially the central part of Ukraine where the reported loss due to the winterkill ranged from 61% to 88% Fig Maps of minimum temperature of snow-free land surface in the month of December of the years 2000 to 2008 in Ukraine Areas of potential winter crop freeze damage (minimum temperatures below -18.50C without snow cover on the ground) are shown in dark gray The analysis of maps for other years in Fig show little freeze damage to winter crops Isolated small areas of potential winterkill are most probably the result of occasional misclassification of cloud as the cloud-clear land surface Most often these misclassifications occur within a large area covered by clouds The algorithm retains observation with the minimum estimated surface temperature in every pixel of the map over a month-long period and therefore tends to accumulate these errors Fig Percent of winterkill damage of winter wheat in winter 2002-2003 based on in-situ assessment in Ukraine Source: US Department of Agriculture Discussion In Ukraine snow cover determines to a large extent the condition of winter crops in the end of winter and their development in early spring This explains the importance of information on the spatial distribution of snow cover and its seasonal changes for agricultural applications In the same time snow cover is by no means the only factor that has to be accounted for in the winter crop condition monitoring and yield forecasting Short duration of seasonal snow cover should be viewed only as an indirect evidence of reduced wintertime precipitation that may contribute to the development of drought conditions later in spring Other factors, such as soil moisture and precipitation in the preceding fall season and, most notably, the amount of liquid precipitation in early spring also affect availability of soil moisture for the crop development at the beginning of the growth season The lack of snow cover protecting crops from low wintertime temperatures and associated freeze damage is one of the primary causes for the winter crop loss in Ukraine Still this is not the only threat to the successful survival of crops through the winter season For example, in winter of 2002-2003 the freeze damage of crops in December 2002 was further exacerbated by persistent ice crusting in February and March 2003 (USDA, 2003) In situ inspection of winter grain crops in spring focuses at the assessment of the damage from all adverse factors combined but can hardly provide estimates of their individual contribution In this study we presented estimates of the crop freeze damage only in the beginning of the winter season when winterkill from freezing is most probable A more comprehensive approach to identify potential frost damage should involve the analysis of snow cover and surface temperatures expanded to the whole period of the year when freezing conditions are possible This analysis should also account for the grain development stage Winter crops are much less freeze resistant in the post-dormant stage therefore injury during these periods may occur at temperatures only slightly below freezing (Shroyer et al., 1995) Although IMS snow cover maps are one of the most reliable sources of information on the snow cover distribution, there are several limitations associated with these maps which the user has to be aware of These limitations may not be critical for the analysis of seasonal changes of hemisphere or continental snow extent, but should considered if maps are used for smaller scale applications First, due to extensive workload, IMS analysts may not be able to review and update the state of every pixel in the Northern Hemisphere map on a daily basis Specifically delays are most probable during fast largescale changes in the snow cover distribution, in particular during active snowmelt in spring In some cases changes in the snow extent since the day before may be too small for analysts to identify them Second, since analysts primarily rely on satellite imagery in the visible spectral band, clouds sometimes hamper accurate and timely reproduction of changes in the snow extent Clouds also present a serious problem for surface temperature retrievals when trying to identify areas of potential winterkill Although the coldest surface temperatures are typically associated with cloud-clear skies, cloud-caused gaps in satellite-derived maps of surface temperature may still result in misses of freeze damage cases Confusion of clouds with cloud-clear land surface by satellite image classification algorithms is another important issue In clear sky conditions the accuracy of surface temperature estimates from satellite observations in the infrared is about 20K to 30K However misinterpretation of a cloudy pixel as clear most often causes a severe underestimation of the land surface temperature and hence may lead to false freeze-damaged area identification In Figure spurious identifications of potential freeze damage in the form of isolated dark pixels or small clusters of dark pixels are clearly seen in all maps The fact that cloud misclassifications and hence, freeze damage false identifications rarely occur over large areas may help to filter at least part of these errors out by examining spatial variations of the derived land surface temperature Conclusion In this study NOAA daily interactive snow cover maps have been used to estimate the duration of seasonal snow cover over Ukraine and to determine the monthly minimum temperature of the snow-free land surface The analysis of information on the snow cover duration and on the minimum temperature of snow-free land surface for the last ten years has shown that these products indeed present an additional piece of information complimentary to available in situ agricultural reports This information may contribute to the assessment of potential crop damage due to winterkill and to the prediction of drought conditions in spring The next step in the improvement of the two products presented in the paper consists in the use of daily snow maps derived with automated snow cover mapping algorithms The most promising approach involves application of algorithms that combine satellite-based snow retrievals in the visible/infrared spectral bands and in the microwave This combination allows for generation of daily continuous (i.e., gap-free) snow cover maps at a spatial resolution of several kilometers similar to IMS maps As compared to IMS analysts, automated algorithms can detect smaller changes in the snow cover distribution and therefore have a potential to more accurately reproduce daily changes of the snow extent Some information on the snow pack properties, particularly on the snow depth can be obtained from satellite observations in the microwave This data can also help in delineating areas affected by winterkill References Derksen CA, Walker A, LeDrew E, Goodison B (2003) Combining SMMR and SMM/I data for time series analysis of central North American snow water equivalent J Hydrometeorol, 4:304-316 Foster JL, Hall DK, Eylander J, Kim EJ, Riggs GA, Tedesco M, Nghiem S, Kelly REJ, Choudhury B, Reichle R (2008) A new blended global snow product using visible, microwave and scatterometer satellite data 88th Annual Meeting of American Meteorological Society, 20-24 January 2008, New Orleans, LA, http://ams.confex.com/ams/pdfpapers/130069.pdf Frei A, Robinson DA (1999) Northern Hemisphere Snow Extent: Regional Variability 1972 to 1994 Int J Climatol, 19: 1535-1560 Hall DK, Riggs GA, Salomonson VV, DiGirolamo NE, Bayr KJ (2002) MODIS snowcover products Remote Sens Environ, 83:181–194 Helfrich SR, McNamara D, Ramsay BH, Baldwin T, Kasheta T (2007) Enhancements to, and forthcoming developments in the Interactive Multisensor Snow and Ice Mapping System (IMS), Hydrol Process, 21:1576–1586 Kongoli C, Grody NC, Ferraro RR (2004) Interpretation of AMSU microwave measurements for the retrievals of snow water equivalent and snow depth J Geophys Res-Atmos, 109 (D24), Art No D24111 Dec 29 2004 Ramsay B (1998) The interactive multisensor snow and ice mapping system Hydrol Process, 12:1537-1546 Romanov P, Tarpley D, Gutman G, Carroll TR (2003) Mapping and monitoring of the snow cover fraction over North America J Geophys Res, 108(D16), 8619, doi:10.1029/2002JD003142, 2003 Romanov P, Gutman G, Csiszar I (2000) Automated monitoring of snow cover over North America with multispectral satellite data J Appl Meteorol 39: 1866–1880 Shroyer JP, Mikesell ME, Paulsen GM (1995) Spring freeze injury to Kansas wheat Cooperative Extension Service, Kansas State University, Manhattan, http://sanangelo.tamu.edu/agronomy/pdf/spgfrze.pdf Simic A, Fernandes R, Brown R, Romanov P, Park W (2004) Validation of VEGETATION, MODIS, and GOES plus SSM/I snow-cover products over Canada based on surface snow depth observations, Hydrol Process, 18:1089-1104 USDA (2001) Russia and Ukraine: soil moisture reserves dwindle as dryness continues United States Department of Agriculture, Foreign Agricultural Service Production, Market and Trade Feature Reports-2001, February 7, 2001 http://www.fas.usda.gov/pecad2/highlights/2001/02/01feb07.htm USDA (2003) Ukraine: Winter Wheat Situation United States Department of Agriculture, Foreign Agricultural Service Production, Market and Trade Feature Reports2003, March 18, 2003 http://www.fas.usda.gov/pecad2/highlights/2003/03/ukr_march03/index.htm USDA (2006) Ukraine: Frost Damage to Winter Wheat in Eastern Region United States Department of Agriculture, Foreign Agricultural Service Commodity Intelligence Report February 10, 2006 http://www.pecad.fas.usda.gov/highlights/2006/02/ukr_09feb2006/ ... winter wheat in winter 200 2-2 003 based on in- situ assessment in Ukraine Source: US Department of Agriculture Discussion In Ukraine snow cover determines to a large extent the condition of winter... spring In this study we processed all daily IMS snow cover maps over Ukraine to determine the duration of snow cover for every winter season since 1999 and average duration of snow cover for. .. level In this study we have used daily IMS snow cover maps along with satellite- derived information on the land surface temperature (LST) to identify areas of possible winterkill Information on