PLANT PHENOLOGY IN WESTERN CANADA - University of Calgary

PLANT PHENOLOGY IN WESTERN CANADA: TRENDS AND LINKS TO THE VIEW FROM SPACE ELISABETH G. BEAUBIEN1∗ and MRYKA HALL-BEYER2 1 Devonian Botanic Garden, University of Alberta, Edmonton, AB T6G 2E1, Canada 2 Department of Geography, University of Calgary, Calgary, AB T2N 1N4, Canada (∗ author for correspondence, e-mail: [email protected])

Abstract. One feature of climate change is the trends to earlier spring onset in many north temperate areas of the world. The timing of spring flowering and leafing of perennial plants is largely controlled by temperature accumulation; both temperature and phenological records illustrate changes in recent decades. Phenology studies date back over a century, with extensive databases existing for western Canada. Earlier spring flowering has been noted for many woody plants, with larger trends seen for species that develop at spring’s start. Implications for ecosystems of trends to earlier spring arrival include changes in plant species composition, changes in timing and distribution of pests and disease, and potentially disrupted ecological interactions. While Alberta has extensive phenology databases (for species, years, and geographic coverage) for recent decades, these data cannot provide continuous ground coverage. There is great potential for phenological data to provide ground validation for satellite imagery interpretation, especially as new remote sensors are becoming available. Phenological networks are experiencing a resurgence of interest in Canada (www.plantwatch.ca) and globally, and linking these ground-based observations with the view from space will greatly enhance our capacity to track the biotic response to climate changes. Keywords: early spring, flowering, phenology, remote sensing, satellite imagery, Western Canada

1. Introduction Spring phenology provides a simple and effective way to discover trends in the biotic response to changing weather and climate. Phenology, here defined as the ‘study of the seasonal timing of life cycle events’ (Rathcke and Lacey, 1985), can include a variety of events. Timing of bird migration, butterfly emergences, mammal hibernation, are examples from the animal world (Lechowicz, 2001). However, plants are the most common focus of phenology, as their fixed location facilitates repeated observation. Timing of budbreak, flowering, leafing, fruiting and leaf colouring are the common phenophases or growth stages recorded. Record durations of at least 5–10 years are needed to get reliable average dates to characterize event timing in an area. Longer records are needed to examine trends. Because the timing of spring flowering of perennial plants in temperate areas is largely driven by temperature accumulation (Rathcke and Lacey, 1985), trends in bloom data provide a useful measuring stick for climate warming.

Environmental Monitoring and Assessment 88: 419–429, 2003. © 2003 Kluwer Academic Publishers. Printed in the Netherlands.

420

ELISABETH G. BEAUBIEN AND MRYKA HALL-BEYER

Phenological information is important in monitoring all aspects of ecosystems (Lechowicz, 2001) and is essential to understanding the dynamics of plant communities, which of course impact animal populations as well. Much of today’s interest in monitoring stems from the impact of climate change on human society, both directly and through changes in the natural ecosystem. Once correlations are determined between the timing of both observed plants and commercially important organisms such as crops or pest insects, considerable money can be saved in sectors such as agriculture and forestry. Phenology data can support decision making in fields as diverse as medicine (pollen warnings for allergy sufferers), forensic studies, tourism, and wildlife management. These data also are also needed as input to models of future plant distribution (Chuine and Beaubien, 2001). The ideal phenology research program would involve trained and paid observers observing genetically identical plants for long periods over a wide area, as in the International Phenological Garden program in Europe (Menzel and Fabian, 1999). However, the realities of costs and institutional continuity mean that in many cases volunteer observers are recruited and trained to provide data for naturally occurring plant species. With sufficient organizational energy and funding, wide coverage of phenological data is relatively easy to obtain with public participation. Botanists or naturalists familiar with plant life histories should establish program protocols by selecting plant species and phenophases suited to public phenology surveys. By providing observers with resources including pictures of the plants and the requested phenophases, quality data can be gathered by amateurs (Beaubien, 1991a, b, 1996). Data interpretation is best done by scientists experienced with phenological data, who have themselves tracked the indicator species and seen variations in their development pattern.

2. Trends Towards Earlier Spring Development We first consider trends that have been measured in western Canada, and then in other parts of North America and Europe. Spring is coming earlier in western Canada. The area has shown significant spring warming over the last decades (Zhang et al., 2000) that is reflected in trends to earlier spring bloom times. A spring-flowering index (SFI) was calculated for Edmonton, Alberta, determined as the annual mean of first bloom dates for three spring-flowering plants: aspen poplar (Populus tremuloides), saskatoon (Amelanchier alnifolia), and chokecherry (Prunus virginiana). A comparison of the SFI for three phenology data sets that span 1936-1996 shows an 8 day trend to earlier development over the last 60 years (Beaubien and Freeland, 2000). Looking at the species individually is also useful, as the first species to bloom in spring seem to be showing the most change. The bloom time of aspen poplar trees in Edmonton, Alberta, is happening almost a month (26 days) earlier now than a century ago (Beaubien and Freeland, 2000). In this same area, saskatoon or

PLANT PHENOLOGY IN WESTERN CANADA

421

serviceberry is flowering 3 days earlier over the years 1936–2000, and chokecherry is flowering 5 days earlier over this 64-year period (E. Beaubien, unpublished data). Major climatic influences such as El Niño events are important in some years; all strong and medium El Niño events were associated with both increases in Pacific ocean temperatures and earlier bloom times in Alberta (Beaubien and Freeland, 2000). At the Pacific Agri-Food Research Station of Agriculture Canada in Summerland, British Columbia, MacIntosh full bloom apple data show a 5 day trend to earlier development from 1936 to 2000. This data set illustrates one of the challenges in merging data in phenology. It is a compilation of 4 data sets but good documentation has been found only for the most recent one (1988–2000), where full bloom was defined as 80% of buds open, and observations were made by a consistent observer (D. Neilsen, personal communication). Cayan et al. (2001) report trends to increasing spring temperatures of 1–3 ◦ C, earlier lilac bloom dates, and earlier streamflow pulse dates beginning in the 1970s for the Western USA. They found coherent fluctuations from year to year represented in both the vegetation and hydrology and dependent on regional temperature variation. Using modeled and actual lilac data, Schwartz and Reiter (2000) found a 5 to 6 day North American trend to earlier spring over the years 1959 to 1993. Trends to earlier bloom have also been reported from Wisconsin (Bradley et al., 1999). Trends to an earlier spring and to a longer growing season have been reported for many European countries (Peñuelas and Filella, 2001): for example in the Mediterranean, leaves of most deciduous plants now emerge 16 days earlier and drop off 13 days later than they did 50 years ago. Data from the International Phenological Gardens shows that over much of Europe the growing season has extended by at least 11 days over the last 40 years, and spring has advanced by 6 days (Menzel and Fabian, 1999). In Europe an early spring (February to April) warming of 1 ◦ C has been shown to cause a one week advance in the beginning of the growing season (Chmielewski and Rötzer, 2001). A lengthening of the growing season has also been detected by remote sensing studies of the northern latitudes (Myneni et al., 1997).

3. Implications of Earlier Spring Development Climate change in western Canada will likely create warmer and dryer conditions, leading to vegetation zones moving north. The response to warming will vary according to the plant species, and to the season of the year that the warming happens. Climate change results in a change in temperature, with no corresponding change in day length. Thus the normal development of plants, which are locally adapted to a certain combination of light and temperature cues, could be disrupted. Certain species with limited adaptability may be lost. Many common woody species such as poplars, willows and alders, which bloom at the start of spring,

422


may especially be affected, as increasingly warmer winters and springs result in very early flowering. Early flowers are more likely to be killed by frosts, resulting in loss of the year’s seed production. Without seeds, these early-blooming plants may not be able to shift their range north quickly enough to keep pace with changing vegetation/climate zones. As well, invasive introduced plant species and pest insects may increase their advances north (Pitelka et al., 1997). Species interactions may be disrupted if species do not change their phenological responses in concert with one another. Pollinators may no longer emerge at the correct time to pollinate plants, and herbivores such as caribou may migrate too late to take advantage of the first flush of plant growth. Evidence already suggests in Europe that some birds are suffering from changed availability of plants and insects (Peñuelas and Filella, 2001). Impacts on natural systems caused by changes in the length of the growing season include changes in food availability, nutrient budgets, competitive ability, and carbon uptake. Agriculture and forestry will experience the benefit of a longer growing season providing soil moisture is adequate. As well, these industries face changes in the timing and distribution of pests and diseases, changes in crop quality, and potentially increased frost damage. Human health will be affected by changes in the distribution of disease, as well as changes in the distribution of allergenic plants and the start and duration of allergy seasons.

4. History of Plant Phenological Records Phenology studies vary with respect to the size of the observation area, the number of observers, the duration of time observed, the type and number of species, and the selection of phenophases. They can be divided into three basic types: the snapshot study, where many observers survey phenology over a large area at one point in time; the intensive study, where one or a small number of observers survey a small area over a period of one or more growing seasons; and the extensive study, which involves a network of observers who survey a large area over a period of years (Beaubien, 1991a). Thousands of years of phenology records exist in the Orient, and one or more centuries of data exist for several countries in Europe. Phenological surveys are active in most of Europe, often based in agro-meteorological government departments. The UK and the Netherlands are currently reactivating networks of observers for their ‘Nature’s Calendar’ programs, with great success in volunteer recruitment. Germany uses rapidly reported data to issue bulletins to farmers to forecast plant diseases, irrigation needs, and fertilization timing. Their precise phenological growth models add so much to agricultural efficiency that millions of dollars are saved annually in preventing crop diseases (Deutscher Wetterdienst, ca.1995). Recorded data on phenology in North America cover shorter periods and are often discontinuous, reflecting the short history of post-Columbian settlement


423

(Beaubien and Johnson, 1994). First Peoples in British Columbia had extensive knowledge of event timing, and maximized hunting, fishing, and berry-gathering success through use of phenological indicators (T. Lantz, personal communication). In North America the largest recent phenology surveys were done on lilac and honeysuckle in the western and eastern United States of America beginning in the mid-1950s. The eastern survey had considerable Canadian involvement in the 1970s. This survey continues, co-ordinated by Dr M. Schwartz of the University of Wisconsin-Milwaukee. In Canada, the oldest extensive database is that of the Royal Society of Canada. Via the Botanical Club and later the Meteorological Service, it gathered observations on a total of 170 events, including plant bloom times, bird arrivals, and ice formation and break-up, from 1892 to 1923. Challenges in using this data include tracking changes in plant taxonomy to deduce which species the observer was actually reporting from many locations across southern Canada. The following are three examples of phenology databases for western Canada, to illustrate the range of observation, precision and geographic extent. For more details see Beaubien and Freeland (2000). Russell (1962) provided annually published data in the Canadian Plant Disease Survey, from Edmonton, Saskatoon and Winnipeg. Observations of first bloom were made for a series of native plants, and wheat development was reported. Data were gathered by paid federal technicians with minimal change in personnel over 25 years (1936 to 1961). Tagged plants were tracked over time, providing a highly reliable data set for 3 specific locations. The Alberta Wildflower Survey (Beaubien and Johnson 1994) tracked 3 bloom stages for 15 native plant species 1987–2001, with the help of about 200 volunteers annually across the province. In 2002 the species and phases have been modified and the survey renamed ‘Alberta Plantwatch’. Observer training was provided through printed program information with tips on site selection, protocols, and species identification (with colour photos and sketches). This has provided wide geographic coverage, with some inevitable annual turnover of observers. Data for a phenological event (e.g., first bloom of saskatoon, Amelanchier alnifolia), can be examined by ecoregion or by zones of abundant data. Potential sources of variability in the data include the observers (level of training, experience), the plants (variation in response to abiotic factors among and within species) and the site (macro and micro differences) (Beaubien and Johnson, 1994). Due to the above considerations, phenological data requires specific techniques to check data quality (Menzel et al., 2001, Schaber and Badeck, 2002). Plantwatch began in 1995 (Beaubien, 1996) and has enlisted volunteers in North America and internationally to track spring bloom times of eight plant species useful as key indicators for phenology. Reporting has been via the internet www. devonian/ualberta.ca/pwatch where tables and maps of data have been posted annually (see ‘archives’ at this website). In 2000–2002 this program has expanded with assistance from Environment Canada’s Ecological Monitoring and Assess-

424


ment Network Coordinating Office (EMANCO) and coordinators have been found for each of the provinces and territories. Funding from EMANCO permitted the first author to select more plant species, revise the current protocols to match international standards, and write the background text for promotional materials. The Canadian Nature Federation (CNF) is a partner for EMANCO in promotion of Nature Watch programs across Canada. In April 2002 the CNF published ‘Plantwatch: Canada in Bloom’, a booklet describing the program and 34 plant species of which 11 can be observed across much of Canada. This information is also posted at www.plantwatch.ca where reported phenology data are mapped immediately. This survey has a great potential to monitor the effects of climate change. By involving the public in tracking changes, citizens’ awareness of, interest in, and personal commitment to life-style changes (such as reducing auto emissions through use of bicycles and public transit) will be increased.

5. Remote Sensing Applications for Tracking Plant Phenology Phenology data show how areas differ in timing of biotic development, spatially and temporally. Satellite imagery can be used to create a spatially continuous map of vegetation density. Such maps have been investigated mainly as a way to provide an appropriate definition of state variables such as Net Primary Productivity, which depend on plant phenomena such as evergreen or deciduous, broad or needle leaf, and leaf-on or leaf-off among deciduous plants. Phenological observations are increasingly seen as essential contributions to verifying the accurate interpretation of satellite images on a global scale (Schwartz and Reiter, 2000). Correlating image data with ground phenological observations can provide increasingly accurate knowledge of what precisely the images are showing. In turn, once the image content is better understood for a given region, it can be used to flesh out point observations of phenology and so create a dynamic map of cyclical change in vegetation. To be used to track vegetation cycles, images must be obtained at closelyspaced time intervals over many years. Because many images are required, free or extremely low-cost access to the imagery by the researcher is also necessary. The primary data used is the Advanced Very High Resolution Radiometer (AVHRR), which has been acquiring sub-daily images since the early 1980s. This sensor, initiated mainly for weather tracking purposes, reports emission in three thermal bands, and vegetation reflectance in the red and near infrared spectral regions. This allows calculation of the Normalized Difference Vegetation Index (NDVI), which by emphasizing the vegetation-specific contrast between near infrared and red reflectance provides a value that directly indicates the density and health of green vegetation on the ground. Further details about this sensor and its products are summarized in Cracknell (1997) and Cihlar et al. (1997).


425

The AVHRR data is available in several global datasets, free of charge. File size is a major consideration, and cloud contamination of imagery is a major challenge. The most common way to deal with data volume is to degrade the nominal 1.1 km pixel size of AVHRR data to aggregate pixels of 4, 8 or 16 km. This is effective on a global scale but introduces difficulties in areas where ground cover is heterogeneous at those scales. The nominal 1 km pixel size is available worldwide for only limited time periods, while individual countries or regions have produced 1 km datasets for varying periods, for example Canada’s GeoComp project (GeoComp, 1997; GeoGratis, 2001). Cloud contamination is dealt with by a process of compositing. This defines a uniform period, such as weekly, dekadal (10-days), biweekly, or monthly. Within the period, the date yielding the maximum NDVI value is selected for each pixel. Since clouds (and snow) result in low NDVI values, only clear pixels will be selected if they are available. There is no foolproof way of making sure that the pixel was not cloud-covered for the entire compositing period, as may happen during rainy seasons or in perennially cloud-covered areas such as the tropics and the arctic. The final image for each compositing period is then a composite of clear pixels, but a different date may be selected for different pixel groups. Information is available about the dates selected for each pixel. There are concerns about radiometric fidelity of the AVHRR images: is one date’s image comparable to another’s on the primary basis of vegetation changes? NDVI values can be affected by look angle, bi-directional reflectance (the property of some leaves to reflect more light in a particular direction), atmospheric haze other than cloud, and changes in the sensor’s electronic characteristics through time. Data processing attempts to eliminate as much of this problem as possible (Eidenshink and Faundeen, 1994). For the past several years, additional image data has become available that will in future expand the capacity of AVHRR. The SPOT VEGT series of images is important, although not available free of charge, and the MODIS sensor has come online. Various products are available from these, and much improved image correction and calibration is expected. To date, the products from these have not been continuously available, and they have not yet been thoroughly investigated either for their own properties or for their potential continuity with AVHRR data. Nevertheless, they remain highly promising for future research, particularly as they have a finer spatial resolution than does the AVHRR. Until the products have been available for enough time to produce a body of peer-reviewed research, the best source of information is online (SPOTImage, 2002; MODIS, 2002). Comparing NDVI images to phenological data presents its own problems. ‘NDVI metrics’ for any location may be extracted, such as value and time of maximum or minimum greenness, dates for beginning of greenup and senescence, and total integrated ‘greenness’ over the season. These metrics, particularly greenup, are analogous to the phenological observations, but they are not identical. NDVI measures leaf area and leaf greenness of all plants within the pixel footprint, a min-

426


imum of 250×250 m for MODIS, more commonly 1×1 km or larger for AVHRR. The phenology data are composed of flowering dates for many individual species (trees, shrubs, and herbs), which in many areas do not dominate the pixel. Where leafout is measured phenologically, it is often onset of leafout, which occurs before leaves are large enough to significantly affect the NDVI value. Images show what is visible from above, thus providing information on the canopy but potentially missing the lower branches and forest floor plants. Grasslands provide particular challenges, as green shoots grow through greater or lesser accumulations of dead material. Phenological observations ideally measure events at a daily scale, and imagery at best portrays a weekly interval. Despite these potential drawbacks, if NDVI information can be extracted that is correlated in some way with phenological variables, then the imagery provides complete spatial coverage of an area that no number of observers could duplicate. Areas where vegetation development is relatively advanced or retarded can be seen, and the average rate of northward progression of greenup can be tracked (Schwartz, 1998). The relationship between phenological data and NDVI is ecosystem-dependent. It has proven effective especially for large forested regions. In these cases, observed phenology is of the species that is most visible from above and is also likely to completely cover most pixels. Where ground cover is heterogeneous and where phenology observations are limited to herbaceous or shrub species, the relationship is more problematic. Markon and Peterson (2002) have demonstrated these problems in relating AVHRR data to NPP point ground values; a similar situation exists for extrapolating point phenological observations to pixel NDVI values. 5.1. B RIEF SUMMARY OF THE AUTHORS ’ CURRENT RESEARCH RELATING AVHRR IMAGERY TO PHENOLOGY IN A LBERTA As an example of the approaches that can be taken, we will briefly summarize some ongoing research in southern Alberta, Canada. Details await publication. Our current research aims to discover which among the various greenup metrics from the NDVI time series provides data most nearly congruent with Alberta Plantwatch data for grassland and parkland ecoregions. We use data for five springflowering species (prairie crocus (Anemone patens), saskatoon (Amelanchier alnifolia), golden bean (Thermopsis rhombifolia), aspen poplar (Populus tremuloides), and choke cherry (Prunus virginiana) in 1992 and 1995, to see if the same metric works in each of these two meteorologically distinct years. We also want to see if flowering of the various species is an adequate predictor of the greenup in the surrounding area, the latter being measured by NDVI. In common usage, greenup is the time when vegetation in general appears green. Greenup onset must be defined operationally for use with NDVI images, and the variety of definitions reflects the peculiarities of individual ecosystems. Greenup may be defined as the date on which the NDVI exceeds a certain value (DeFries


427

et al., 1995). This threshold value must be determined for the local ecosystem, since, for example, grassland may green up at a threshold value lower than the year-round minimum of an evergreen forest. Tieszen et al. (1997) define greenup as the date when the NDVI value shows the greatest increase above the midwinter value (baseline) and remains above it for the greatest sustained increase. Similarly, Reed et al. (1994) define greenup for the US northern Great Plains as the beginning of greatest increase over background NDVI (in other words, the sudden upturn on a smoothed NDVI vs. time curve; smoothing may be done in a variety of ways). Correlations between NDVI metrics and phenological data were tested for individual phenological observation locations, and for means and modes of phenological data within grasslands and parklands. The best predictor of mean flowering dates within each ecoregion, for both years tested, has proved to be the date on which most of the pixels within that ecoregion surpassed its threshold NDVI value. Aggregating the data in this way allows the mean of phenological observations to represent their natural regions, and minimizes problems due to vegetation heterogeneity and scaling from a point to a pixel. For each pixel, one can record the date on which the NDVI surpasses a chosen threshold. This allows us to map the considerable spatial variability of greenup dates within the ecoregion, both because of micro and mesoclimatic differences and because of varying ground cover (such as agricultural crops). This is not possible with phenological data alone, as it is currently reported.

6. Future Needs in Canada In 1993, E. Beaubien and M. Schwartz set up a phenology study group within the International Society of Biometeorology. In the spring of 1995, E. Beaubien chaired a meeting of international phenologists hosted by the German Weather Service who met to lay out a plan for the future (Lieth, ed., 1996). Since then increasing cooperation has resulted in a number of publications highlighting trends in the response of vegetation to climate warming. Together we have developed an electronic listserv for communication, and have participated in at least eight international conferences. Major funding from the European Union was secured by A. VanVliet of the Netherlands and has resulted in many phenology workshops and conferences in Europe 2001–2003. In Canada there are many steps that need to be taken to advance plant phenology studies. A sound Plantwatch organization needs to be created, with fundraising capability to assist regions in creating and maintaining strong programs. A clear policy must be agreed upon by all partners with respect to data checking, archiving, access and publication. Frequent communication and meetings are necessary to encourage data sharing between regions of Canada, and among nations. Funding is needed for experienced scientist-phenologists to permit refining Plantwatch protocols, analysing data, and publishing results.

428


Considerable research is needed. Firstly, phenologists need to work with remote sensing scientists to better use and adapt phenology observations as validation of satellite imagery. Secondly, research is needed on the relationships between temperature and flowering or greenup of our key indicator plants, perhaps using common garden experiments to explore the role of genotype. Thirdly, we need to know the effects on phenology of other abiotic factors such as moisture stress and photoperiod, as well as biotic factors such as the physiological chilling of woody plants. Lastly, and most importantly, we must continue to research applications in fields such as agriculture, forestry, and medicine, to build partnerships to maximize benefits from these phenology data.

Acknowledgements Thanks to John Wilmshurst, Martin Lechowicz, and Mark Schwartz for comments on this paper. The phenology work has been supported by the University of Alberta Devonian Botanic Garden (administrative and facility support) since 1991. Dr Geoff Holroyd has provided major financial support since 1987. EMAN CO provided support for this paper and protocol development, and has a great interest in acquiring and expanding Plantwatch. Thanks to Natural Resources Canada: Canadian Forest Service, for assisting with two Science and Technology internship positions. Funding help was provided also by sources including Canada Trust Friends of the Environment Foundation; Shell Canada; Environment Canada: EcoAction, Biodiversity Convention Office and the Northern Ecosystem Initiative; and the Canadian Circumpolar Institute. Image analysis was performed at the Geographical Information System laboratories of the University of Calgary Department of Geography, with the assistance of Medina Hansen, GIS Technician and numerous students in advanced remote sensing classes.

References Beaubien, E. G.: 1991a, ‘Phenology of vascular plant flowering in Edmonton and across Alberta’, MSc thesis. Dept. of Botany, University of Alberta, Edmonton, Alberta. Beaubien, E. G.: 1991b, ‘Phenology, the potential for education’, in Proceedings of the second endangered species and prairie conservation workshop, Regina, Saskatchewan, January, 1989. G.L. Holroyd, G. Burns, and H.C. Smith (ed.) Provincial Museum of Alberta Natural History Occasional Paper #15. Edmonton, Alberta. pp. 117–121. Beaubien, E. G.: 1996, ‘Plantwatch, a model to initiate phenology in school classes’, Phenology and Seasonality 1, 33–35. Beaubien, E. G. and Freeland, H. J.: 2000, ‘Spring phenology trends in Alberta, Canada: links to ocean temperature’, Intern. J. Biometeorol. 44, 53–59. Beaubien, E. G. and Johnson, D. L.: 1994, ‘Flowering plant phenology and weather in Alberta, Canada’, Intern. J. Biometeorol. 38, 23–27.


429

Bradley, N. L., Leopold, A. C., Ross, J. and Huffaker, W.: 1999, ‘Phenological changes reflect climate change in Wisconsin’, Proceedings of the National Academy of Sciences, USA. 96, 9701–9704. Cayan, D. R., Kammerdiener, S. A., Dettinger, M. D., Caprio, J. M. and Peterson, D. H.: 2001, ‘Changes in the Onset of Spring in the Western United States’, Bull. Am. Meteorol. Soc. 82, 399–415. Chmielewski, F.-M. and Rötzer, T.: 2001, ‘Response of tree phenology to climate change across Europe’, Agricult. Forest Meteorol. 108, 101–112. Chuine, I. and Beaubien, E.: 2001, ‘Phenology is a major determinant of tree species range’, Ecol. Lett. 4, 500–510. Defries, R., Hansen, N., and Townsend, J.: 1995, ‘Global discrimination of land-cover types from metrics derived from AVHRR Pathfinder data’, Remote Sens. Environ. 54, 209–226. Deutscher Wetterdienst. circa 1995. Economic Benefit of the Agrometeorology Service. Eidenshink, J. R. and Faundeen, J. R.: 1994, ‘The 1-km AVHRR global land data set: first stages in implementation’, Intern. J. Remote Sens. 15/17, 3443–3463. GeoComp: 1997, http://ceocat.ccrs.nrcan.gc.ca/client_acc/guides/geocomp/geocomp.html Accessed 2002-03-11. GeoGratis: 2001, http://geogratis.cgdi.gc.ca/frames.html Lechowicz, M.J.: 2001, ‘Phenology’. Encyclopedia of Global Environmental Change, Volume 2.The Earth System: biological and ecological dimensions of global environmental change. Wiley, London. Lieth, H.: 1996, ‘The Phenology Study Group’. Phenology and Seasonality 1,1, 3–4. Markon, C. J. and Peterson, A. M.: 2002, ‘The utility of estimate net primary productivity Over Alaska using baseline AVHRR data’, Intern. J. Remote Sens. 23, 4571–4596 Menzel, A. and Fabian, P.: 1999, ‘Growing season extended in Europe’, Nature 397, 659. Menzel, A., Estrella, N. and Fabian, P.: 2001, ‘Spatial and temporal variability of the phenological seasons in Germany from 1951 to 1996’, Glob. Change Biol. 7, 657–666. MODIS: 2002, http://edcdaac.usgs.gov/modis/dataprod.html Accessed 2002-11-07. Myneni, R. B, Keeling, C. D., Tucker, C. J., Asrars. G and Nemanii, R. R.: 1997, ‘Increased plant growth in the northern high latitudes from 1981 to 1991’, Nature 386, 698–701. Peñuelas, J. and Filella, I.: 2001, ‘Responses to a Warming World’, Science 294, 793–795. Pitelka, L. F. and the Plant Migration Workshop Group.: 1997, ‘Plant Migration and Climate Change’, Am. Sci. 85, 464–473. Rathcke, B. and Lacey, E. P.: 1985, ‘Phenological patterns of terrestrial plants’, Ann. Rev. Ecol. Syst. 16, 179–214. Reed, B. C., Brown, J. F., Vanderzee, D., Loveland, T. R., Merchant, J. W. and Ohlen, D. O.: 1994, ‘Measuring phenological variability from satellite imagery’, J. Veget. Sci. 5:5, 703–714. Russell, R. C.: 1962, ‘Phenological records of the prairie flora’, Can. Plant Disease Survey 42(3), 162–166. Schaber, J. and Badeck, F-W.: 2002, ‘Evaluation of methods for the combination of phenological times series and outlier detection’, Tree Physiology 22, 973–982. Schwartz, M. D.: 1998, ‘Green-wave phenology’, Nature 394, 839–840. Schwartz, M. D. and Reiter, B. E.: 2000, ‘Changes in North American spring’, Intern. J. Climatol. 20, 929–932. SPOTImage: 2002, http://www.spot.com/home/proser/listvege/welcome.htm Accessed 2002-11-07. Tieszen, K. L., Reed, B. C., Bliss, N. S., Wylie, B. N. and Dehong, D. M.: 1997, ‘NDVI, C3 and C4 production and distribution in Great Plains grassland land cover types’, Ecol. Applic. 7(1), 59–78. Zhang, X., Vincent, L. A., Hogg, W. D. and Niitsoo, A.: 2000, ‘Temperature and precipitation trends in Canada during the 20th century’, Atmosphere-Ocean 38(3), 395–429.