2
Chapter 2
Economic impact of extreme weather events and the potential benefits of agrometeorological information
Vanda Cabrinha1
(1) Ministerio da Ciência, Tecnologia e Ensino Superior, Instituto de Meteorologia, I.P, Rua C – Aeroporto de Lisboa, Tel.: (+351) 21 844 7000, e-mail:informacoes@meteo.pt,
1749-077 Lisboa – Portugal Fax: (+351) 21 840 2370, URL:
2.1 Summary (Introduction)
Agrometeorology, is an interdisciplinary science requiring the knowledge of meteorology and the climatology of the agricultural systems, with the objective that “The systems” become more efficient and more sustainable. It happens, basically, on the plants and animals that are integrated in these systems, but must enclose all the factors that have a climatic component that stand out for the system to be more efficient and sustainable (Abreu, J.P, 2008).
In that sense, the climatic and meteorological information has to be treated and presented in a simple convenient form. The climatic information is static and valid for a long period of time, so it must be prepared in graphical format and supply computer tools that support strategic decisions on agriculture and industry in order to minimize the impacts that can occur due to extreme climatic events (Abreu, J.P, 2008).
Agrometeorological information and services from the National Meteorological and Hydrological Services (NMHSs) are increasingly being demanded by the farming community to cope more efficiently with climate variability and the increasing incidence of extreme meteorological events such as droughts, floods, frosts and wind erosion, which leads to serious socio-economic impacts on the agricultural sector (Sivakumar, 2007).
Some of these impacts associated to agriculture, food security and livestock are:
Due to little rainfall:
· Total/partial crop failure, in the affected areas especially;
· No support for regeneration of pasture and browse for livestock;
· Rising losses in livestock.
Due to moderate to heavy rainfall:
· Sustenance of planted food and cash crops;
· Increased yield of perennial cash crops such as coffee and tea;
· Regeneration of pasture and browse for both livestock and wildlife;
· Improved nutrition due to increased milk production.
The socio-economic impact cases studies highlighted in this report relates to anomalies of too much or too little rainfall.
2.2 Cases Studies
2.2.1 Case study 1 – Drought 2003
2.2.1.1 Europe[1]
The year of 2003, in the European Union of the 15, was characterized for a reduction of the agricultural production in real terms (Fig. 2.1), as a consequence of a drought episode that occurred in the European countries.
In the case of the vegetables production, the losses in the harvests due to drought effect were not compensated by the increase of prices. This situation was particularly severe for the four major agricultural producers, France, Italy, Germany and Spain.
The agricultural production suffered a volume decrease of 3,4%, with the vegetables production suffering a serious reduction of 6,5%, especially in the countries more affected by drought situation, as the South of Europe, France, Germany and Austria.
Figure 2.1. Evolution of the agricultural production in the European Union in 2003
The production of cereals suffered a decrease of 10,8% after an exceptionally 2002 year. The potato harvest decreased about 9,0% and the harvest of fruits suffered a new decrease (4.6%). The wine production recorded a 9,8% decreases, reaching its lowest level since 1996.
2.2.1.2 France
The agricultural harvests in 2003 were really affected by the drought situation. The agricultural production suffered a volume reduction, of, 8,6% in comparison with the year of 2002 (Table 2.1). This large decrease was mainly due to the reduction of the vegetal production in the order of 13.6%.
Table 2.1. Value of the production in 2003 and Evolution of the agricultural production in France (Volume, price and value) 2003 vs. 2002
The ice event in April followed by drought situation was terrible for the harvests. The year of 2003 was seriously affected by the drought, which resulted in a harvest reduction of cereals and fruits, and the consequent increase of prices.
This drought was responsible for the lost production of the forage crops cultures and the cereals. The consequence was lower incomes, which leaded to a reduction of the sown area (4.0%). In the case of the common wheat (Triticum aestivum L.) the harvest had a reduction of about 18,0% in relation to the average of the 5 previous years.
These extreme climatic events had serious consequences for the wine production that decreased 13.5%, particularly in Champagne region that had a 26,3% loss. It was the fourth consecutive loss in production and the weakest since 1991. The productions of fruits, potato, beetroot and vegetables had suffered a reduction of 10%, 8.8%, 6.2% and 3.1%, respectively, after an abundant harvest in 2002.
As a consequence of this situation the exploration subsidies reached the 2,6 thousand million euros in 2003 being in 2002 of 1,9 thousand million euros, recording an increase of the subsidies in the order of 34.9% in the year of 2003 when compared with 2002 (Table 2.2).
Table 2.2. Value of financial compensation for agricultural exploration in 2003 and comparison with 2002
Thus in 2003, 662 million Euros were used as compensation from those natural disasters (ice, drought) with impacts in the agricultural sector.
2.2.2 Case study 2 – Drought and Floods 2005
2.2.2.1 Europe[2]
The year of 2005, in the European Union of the 25, was characterized by a reduction of the agricultural production when compared with the previous year, with exceptional harvests. However during 2005 the prices remained steady or even decreased slightly. This situation is more evident in the four major European producers: France, Italy, Germany and Spain (Figure 2.2). The agricultural production decreased, in volume, about 2,9%. The vegetal production had lost of 5,3% after a particularly abundant year of 2004.
Figure 2.2. Evolution of the agricultural production in European Union in 2005
A large part of the South of Europe, in particular the Iberian Peninsula suffered the effect of drought, while certain regions of Central Europe were affected by floods episodes.
The harvest of cereals decreased about 11,1%. Spain was particularly disastrous, with a lost of 42,0%. The potato harvest also decreased (8.8%), as well as the fruits (2.0%) and wine that recorded a 10,1% lost after 2004 that have reached its highest value in the last 15 years.
Low moisture content in soil means a highly increased risk of forest fires, such as those that raged on the Iberian Peninsula. Forest fires were a particular problem in Portugal, claiming the lives of 18 people and destroying 300,000 hectares (750,000 acres) of forest. In Spain, fires also raged last year, which, in combination with the lack of water, wreaked havoc on Spain's 2005/06 durum wheat crop, which was down 65 percent from the previous year's production level. Durum wheat is grown primarily in the southern provinces, where drought has been especially severe. Similarly, in Portugal, where durum averages 70 percent of its total wheat output, the durum crop suffered a reduction of almost 60 percent.
2.2.2.2 Portugal [3]
Agricultural year 2004/05 in Portugal was initially favourable (Fig. 2.3) with respect to precipitation amounts, because October was an extremely rainy month, except in the Southern region where it was dry to normal. Although, next months were classified as dry to extremely dry, which caused a drought period with high intensity.
As a consequence agricultural year 2004/05 was characterized by a situation of severe and extreme drought, which affected the whole Mainland territory during the almost whole agricultural year.
|[pic] |
Figure 2.3. Left: Precipitation (agricultural year 2004/2005), Right: Temperature (agricultural year 2004/2005) (source: INE)
Figure 2.4 shows the monthly evolution of PDSI index regarding the drought area affected in Mainland Portugal.
| [pic] |
Figure 2.4. Area (%) affected by drought (Sep 2004 - Oct 2005) (source: IM)
The analysis of the figures and the table reveals an aggravation of the drought conditions in winter months, with attenuation in March, due to precipitation occurrence in the Northern and inner regions of the territory. In summer (June, July and August), the least rainy season, which contributes with about 6% for the annual precipitation, the drought situation worsened. In September and due to the precipitation occurred in the first fifteen days, in particular in the Northern and Centre regions there was attenuation in these regions. At the end September 30, 2005 the whole territory continued in a drought period with severe to extreme intensity, only in October there was a decrease of drought intensity.
This situation caused serious damages in agriculture. The severe effect of drought had serious consequences in cereal campaign, which was the worse of the last two decades. The worst pasturing conditions and the scarce forage crops reserves had compelled most of the productive units to the extraordinary consumption of industrial rations and to the acquisition of straws outside the national market at very high prices
Some of the most important findings for year 2005, comparing with 2004, show:
In production terms
Cereals production: the worst of the last two decades
Industrial tomato: the national processing quota was reached for the second year
Pear: production drops 30%
Olive oil and wine: superior quality
Bovine meat: decrease in production for adult animals
Milk: increase in production
In economical terms
Decrease in output price index
Decrease in input price index
Decrease in Gross Value Added at current prices on Agriculture
Decrease in Agricultural Income
Cereal Production
Autumn/Winter Cereals (Fig. 2.5):
The cereal production was the worst of the last decades, with a general drop, particularly significant in the production of straw and grain. With exception of the common wheat (Triticum aestivum L.), all the cereals recorded decreases of production, compared to the previous harvest. Due to low production and to the bad quality of grain, many cereal fields had been haying and/or shepherded.
|[pic] |
Figure 2.5. Left: Autumn/Winter Cereals area, Right: Autumn/Winter Cereals production (source: INE)
Besides the severe drought, in Portugal the change of support schemes for farmers was changed in the scope of the new Common Agricultural Policy (CAP), with the closure of payments for the arable crops production, which modified the cereal structure of the country. The area of durum wheat (Triticum durum L.) that, in recent years, recorded strong expansion due to complementary aid, presented an accented drop, compensated, in part, by the increment of the area of soft wheat (241%), barley (116%) and triticale (72%).
Spring/Summer Cereals and Maize:
The Spring/Summer sowing happen with delay and great uncertainty on the part of the agriculturists whom, due to lack of moisture and water scarcity for irrigation, opted to reducing or not effecting the areas habitually sown.
The area of dry land maize decreased 16%, face the 2004, placing itself in the 110 a thousand hectares (Fig. 2.6). For the maize in irrigated land, the breaking was of 20%, justified by the drought situation, and by the introduction of the Unique Payment Regime that, when guaranteeing an income for exploration, had as a consequence the delay of sowing cultures most demanding in investment terms.
|[pic] |
Figure 2.6. Left: Maize area, Right: Maize production (source: INE)
Maize (Zea mays L.) with a lower production record, comparatively with the previous harvests, showed an accented decrease (- 35%). This break was due not only to the decrease of the sown areas and to the transference of grain for maize silage, but also to the reduction of the unitary incomes, as a consequence of the high temperatures and absence of moisture.
The surface sown with rice (Fig. 2.7) followed the same trend of the maize, with a breaking around 14%, with respect to the previous year. This decrease was more accented in the irrigation regions of the South. The production lost reached the 19%.
|[pic] |
Figure 2.7. Left: Rice area, Right: Rice production (source: INE)
Potatoe (Fig. 2.8):
The production of the potato decreased about 25%, with respect to the previous campaign. For the dry land potato, the adverse conditions at the beginning and continuation of the productive cycle, lead to consequences in the formation of tubercles. Also the high temperatures in the time of the harvest had as consequence a water stress situation and the “burns of the plant”. This situation originated tubercles with presented minors bores, recorded decreases of production of 40% and 25% in the regimen of dry land and irrigated land, respectively.
|[pic] |
Figure 2.8. Left: Potatoe area, Right: Potatoe production (source: INE)
Olive oil:
The olive oil production (Fig. 2.9) went up to around the 325 a thousand hectoliters, what it represents a 35% lost, face to the previous harvest. It contributed for this break the decrease of the production due to dry and hot weather that conditioned the wadding of the fruits and originated the premature fall of the olive and the low industrial income of the olive.
Figure 2.9. Olive Oil production (source: INE)
Vegetal Production:
Table 2.3 shows the vegetal production in the most important cultures from 2003 to 2005:
Table 2.3. Principals cultures production in Portugal (Source: INE)
|[pic] |
Prices and Income of the Agricultural Activity:
In 2005, comparison with the year of 2004, shows for the index of prices of the agricultural products a -3% variation, as a consequence of the observed decreases, as much in the index of prices of the vegetal products (- 4.1%), as in the index of prices of the animals and animal products (- 1.3%).
|[pic] |
Figure 2.10. Index of prices of agriculture products (source: INE)
In 2005, the value of the production of the agricultural sector, at the market prices, lowered 7.8%, face to 2004, as a consequence of the decrease of the vegetal production value (- 15.6%) (Figure 2.10). In the vegetal production, reference to the cereals and wine, with evolutions of -55,1% and -21,9%, respectively.
The expenses in current consumptions in agriculture had remained globally steady (the intermediate consume decreased 0.6%, in value). However, the breaking in the value was conditional for the volume (- 4.3%), because the prices increased 3.9%. The main reasons for this behavior had been the climatic scenarios and the instability in the petroliferous market.
The drought situation reflected a lesser consumption of seeds and seasonings products (Figure 2.11). The lesser agricultural activity also induced a reduction of the fuels volume. However, the significant increase of prices determined the value of the total consumption of Energy (+15.2%).
|[pic] |
Figure 2.11. Structure of the production of the agricultural sector, at base prices (source: INE)
The paid subsidies had increased 8.4%, having an anticipation of payments for 2006, due to drought situation.
The Remunerations had increased 4.4%, in nominal terms. This evolution is explained by the increase of the agricultural workforce volume wage-earning, because the total workforce volume had decreased. The incomes to pay had lowered 4.8%, given the lesser-cultivated area, due to the adversities of the agricultural year. The interests to pay had increased 6.0%, although the interest taxes had decreased. This increase was due to granted credit, essentially through credit facilities granted realized due to drought.
2.2.2.3 France [4]
The agricultural production decreased in volume about 3,3%, comparatively with the previous year (Table 2.4). This breaking was due to the reduction of the vegetal production of 6,4%.
The drought situation affected the incomes gotten with the maize culture harvested in the autumn.
The potato production decreased 7.1%. The wine production also suffered with a lost of 8,8%, comparatively with the good harvest occurred in the previous year
Table 2.4. Value of the production in 2005 and evolution of the agricultural production
(volume, price e value) 2005 vs. 2004
The harvest of cereals decreased about 8,8% but it was similar to the average value of the 5 preceding years. This breaking was consequence of the reduction of the cereal surface and the reduction of the incomes received in consequence of the drought situation. For example, the harvest of soft wheat (Tricum aestivum L.) had a lost of 7,0% when compared with the 2004 harvest.
In value terms the cereals production had a reduction of 15,8% in comparison with the year of 2004 (Figure 2.12).
Figure 2.12. Variation of the value of the agricultural production in comparison with the price base of 2004
The production of maize in grain suffered a reduction from 19,0% in 2005, with part of the planted surface designated to suppress the necessities of the forage crops production, which had a serious lost as a consequence of drought occurred in this year.
2.2.3. Case study 3 – Flood episode - October 2006
2.2.3.1 Portugal[5]
Agricultural year 2006/07 was characterized for an autumn of great meteorological instability, with continuous precipitations, sometimes intense, followed with strong winds and thunderstorms. Some soil had reached the saturation, occurring intense outflow for the water lines.
The extremely rainy autumn, with uninterrupted precipitations until the first week of December, had saturated the agricultural soil, especially in the fertile valley zones, and compelled to the suspension of the sowing of autumn/winter. These meteorological conditions made it difficult to achieve the accomplishment of the last harvests of the maize and rice, and conditioned the end of harvesting of wine grapes, and even in some cases, caused the interruption of the works of preparation of the soil for the new agricultural year.
Cereal Production - Autumn/Winter Cereals:
Autumn rains conditioned the catch-crops and reduced the cereal surface, specially in the hard wheat (- 57%) and soft wheat (- 47%), following by the triticale with a 17% decrease and oats of 14%; the areas sown with barley and rye recorded less expressive breakings, about 8% and 5%, respectively (Fig. 2.13).
|[pic] |
Figure 2.13. Left: Autumn/Winter cereals area, Right: Autumn/Winter Cereals production (source: INE)
On the other hand, the prolonged saturation of the soil compelled to the widening of the sowing period and conditioned the nitrogenated seasoning and herbicides application, promoting a very heterogeneous development of cereal fields. This situation lead to the losses of production that would come to be aggravated by the raised temperatures and for the dry winds of March 2007, which originated advances in the phonological state of the cultures. The precipitation, hail and strong winds occurred in May 2007aggravated even more the situation giving origin to the revival of some weeds.
In conclusion cereal production 2006/07 was affected by serious losses of production, relatively to the previous year, and even relatively to the average of the last 5 years, as a consequence of decreases in the surfaces and the unitary incomes.
2.2.4 Case study 4 – Drought 2007
2.2.4.1 Moldova[6]
In 2007, the Republic of Moldova was facing great difficulties caused by an extraordinary drought that affect nearly all the country. According to international estimates of the impact this phenomenon had on ecology and agriculture of catastrophic proportions.
For the entire period of meteorological observations conducted on the territory of the country, similar phenomena were observed only twice – at the end of 19th century and in the period of 1946-1947.
The extreme climate conditions leading to an unusual situation and a profound crisis in the agricultural and food sector of the Republic of Moldova, demonstrated once again that agriculture is the most vulnerable and risky sector of the national economy; the yield and productivity of agricultural crops in 2007 registered the lowest rate within the last decade, prompting a major collapse in the sector.
During 2007, the amount of losses in the agricultural sector increased daily, as a result of the maintenance of high temperatures, thus causing a diminishing yield cultivated crops (sun-flower, corn, and sugar beet), vegetables and fruits.
Other difficulties were consequence of the process of preparation (ploughing) of agricultural land for the autumn sowing, including land too dry to sow, a lack of financial recourse for farmers as a result of the damages suffered, and a lack of seeding material.
As a consequence, the drought had a major impact on the financial situation of farmers, causing partial loss of the seeding fund, fodder and the genetic fund of animals. This decreased the amount of raw material available to the food industry and thus, budgetary revenues for 2007-2008.
2.2.4.2 Romania[7]
Thirty-four of Romania's 42 counties suffered from severe drought in 2007. After helping hundreds of families recover from previous year's floods, several ministries have issued a request to all NGOs to contribute and help in any way they could.
Iasi County was one of the most affected in the country, with 46 of its 98 communes confronted with lack of water supplies, dried pastures and extreme heath. Over 2,000 hectares of crops were beyond hope. Some 1,000 families find themselves unable to feed their animals. As a direct consequence, they were forced to sell them for prices two or three times less than their real value.
Experts say that 2007 summer was more arid than 1946. In March 1947, more than 90 percent of the population in the south of Romania, a former northeastern region, was suffering from hunger, following the destruction of 1946's crops. Temperatures over 55 degrees Celsius were registered for days in a row. In 2007 the situation was even more critical, with the level of humidity in the soil being two times lower because of the lack of rain (according to the Romanian Agrometeorology Laboratory), and that the lack of water in the soil would be hard to compensate even in the case of heavy rains.
Analysis of underground water supplies showed that, from the beginning of June, the levels had decreased 5-20 meters. Over 32,000 wells in Iasi County were dry, 550 kilometers of river course dried up and 35 small rivers had vanished. Pastures in 30 villages wilted and people started slaughtering their cows and poultry because they could not feed them anymore and because they were scared of famine.
At a national level, the drought has destroyed over 19,000 hectares of spring crops and over 10,000 hectares were 50 percent compromised. Of 19,000 hectares of fall crops, over 8,500 were destroyed.
2.3 Conclusions
Unfortunately, an increased frequency of severe droughts in southern and Western Europe in the years to come seems to be a reality, according to several European and American climate studies. While a warmer and wetter climate is predicted in Northern Europe, particularly in Scandinavia, a dryer climate is forecast all year in Southern Europe.
Heat waves will be more frequent, dry periods will be longer, and there will be more intense precipitation events. Temperature rise and changing precipitation patterns are expected to exacerbate the already acute water shortage problem in southern and southeastern regions. Changes in frequency and intensity of droughts and floods are projected for the years to come, which are likely to cause significant environmental, financial and human loss throughout Europe.
Better information and warnings on weather, better detection/monitoring of natural hazards and improved climate projections, together with sustainable resources management, can mean the difference between life and death for all parties involved in European agriculture. Today’s sophisticated Earth observation systems are vital on several time scales: Real-time, for monitoring natural disasters as they unfold; short term, for predicting severe weather events by providing input to numerical weather prediction models; and even long term, to monitor the climate and as input in climate models for climate change predictions.
2.4 References
Abreu, J.P, 2008. A Agrometeorologia no Presente e no Futuro. Jornadas Técnicas “A Importância da Meteorologia na Agricultura”. Beja, Portugal.
INE, 2006. Boletim Mensal da Agricultura Pescas e Agro-Indústria - Novembro 2006. Lisboa, Portugal.
Sivakumar, M. 2006. Dissemination and communication of agrometeorological information—global perspectives. Meteorological Applications, vol 13. Cambridge University Press. Great Britain.
INE, 2008. Estatísticas Agrícolas 2007. Rev. Agricultura Florestas e Pescas, Instituto Nacional de Estatística, Lisboa
INE, 2006. Estatísticas Agrícolas 2005. Rev. Agricultura Florestas e Pescas, Instituto Nacional de Estatística, Lisboa
IPCC, 2007.Climate Change 2007: Synthesis Report. Fourth Assessment Report – FAR. November. 2007
Internet Sites:
World Vision International:
Chapter 3
PRELIMINARY TITLE:
Review and recommend applications of seasonal to interannual climate forecasts to agriculture in Europe (TOR c) and To assess the feasibility of using numerical weather products in operational applications of agrometeorology (TOR d)
(Applications of Numerical Weather Products and Seasonal Forecasts in agrometeorology)
Pierluigi Calanca1, Federica Rossi2, Teodoro Georgiadis2, Andreja Sušnik3 and Gregor Gregoric3
(1) Agroscope Reckenholz-Tänikon, Research Station ART, Reckenholzstr. 191, 8046 Zürich, Switzerland; pierluigi.calanca@art.admin.ch
(2) Inst. Biometeorologia, Consiglio Nazionale delle Ricerche, via P. Gobetti 101, 40129 Bologna, Italy; f.rossi@r.it and t.georgiadis@r.it
(3) Environmental Agency of the Republic of Slovenia, Vojkova 1/b, 1000 Ljubljana, Slovenia; andreja.susnik@gov.si and gregor.gregoric@gov.si
3.1 Introduction
The use of numerical weather products in agricultural decision problems has a long history (see e.g. Wilks, 1997, for a review of case studies conducted after 1960) and it is nowadays recognized that short- to medium-range weather forecasts and climate outlooks out to the season represent a key service for agricultural production (Obasi, 2000; Sivakumar et al, 2000; Stigter et al., 2000). More than this, information recovered from weather forecasts and climate outlooks could be vital for the whole food production chain, including the insurance industry, food processors and distributors, as well as the consumers (Dutton, 2002).
For many agronomic applications (e.g. frost protection or irrigation, fertilization or harvest scheduling) the decision process takes place on a time scale of the order of days, which is also the lead-time for which there exists a significant skill of weather forecasts. This may explain why agrometeorological products currently available from national meteorological and hydrological services (NMHSs), such as the traditional agrometeorological bulletins, are limited to short-term predictions.
But the reviews by Wilks (1997) and Meinke and Stone (2005) as well as reports published by the WMO (e.g. Harrison et al., 2007; Das et al., 2008) indicate that a variety of agricultural decision problems would actually profit from forecasts on time scales extending beyond the weekly, the monthly and even the seasonal (Tab. 3.1). Examples of weather-sensitive decision problems operating on long time scales are the choice of crops, land management (land allocation), the management of fertilizers, and the management of products (e.g forage preservation). In many cases the necessity of long-range forecasts can be motivated from an agronomic, environmental and even economic perspective (e.g. Wilks, 1997).
At the European level, monthly to seasonal forecasts are now routinely issued by the European Centre for Medium-Range Weather Forecasts (ECMWF). The forecasts are distributed to the NMHSs of the member states, and are more and more used as a supplement to the traditional, deterministic forecasts. Ensemble weather predictions out to the season, though, are not yet considered as part of operational forecasting systems aiming at specific agrometeorological products. Two reasons may explain this situation. First of all, the fact that for Europe the skill of weather predictions at lead time exceeding roughly two weeks is modest (Buizza et al., 1999; Rodwell and Doblas-Reyes, 2006, Weigel et al., 2008). Second, the fact that the availability of tools for linking monthly forecasts and climate outlooks to application models, such as crop growth simulation models, is still limited (Doblas-Reyes et al., 2006).
Table 3.1. Agricultural decisions at a range of temporal and spatial scales that could benefit from targeted climate forecasts. Source: Meinke and Stone (2005)
|Example of decision types |Frequency (years) |
|Logistics (e.g. scheduling of planting/harvest operations) |Intra-seasonal (< 0.2) |
|Tactical crop management (e.g., fertilizer/pesticide use) |Intra-seasonal (0.2 ( 0.5) |
|Crop type (e.g., wheat or chickpeas) or herd management |Seasonal (0.5 ( 1.0) |
|Crop sequence (e.g., long or short fallows) or stocking rates |Inter-annual (0.5 ( 2.0) |
|Crop rotations (e.g., winter or summer crops) |Annual/bi-annual (1 ( 2) |
|Crop industry (e.g., grain or cotton; native or improved pasture) |Decadal (~10) |
|Agricultural industry (e.g., crops or pastures) |Inter-decadal (10 ( 20) |
|Landuse (e.g., agriculture or natural systems) |Multi-decadal (> 20) |
|Landuse and adaptation of current systems |Climate change |
The question of the post-processing of ensemble prediction is of course of outmost importance. Yet it is only but one of the key activities required for establishing forecast systems that support agrometeorological decision problems. As schematically depicted in Fig. 1, other aspects of relevance concern the verification of weather forecasts and climate outlooks (specifically with respect to variables other than temperature and precipiation), the development and calibration of application models, and the post-processing and verification of agrometeorological forecasts.
In this scheme, NMHSs play a key role as providers of numerical weather products, but the participation of other partners, such as research institutions, agrometeorological services as well as the stakeholders is crucial to arrive at fully operational system. The necessity for a close collaboration among these difference actors was clearly one of the main outcomes of a project for the development of a web based climate forecast information system for agricultural risk management in the southeastern U.S.A. (Fraisse et al., 2006).
The purpose of the present report is to examine some of the elements included in Fig. 3.1 and required for the application of weather forecasts and climate outlooks to agrometeorology. The report has not the pretension of being comprehensive and focuses on applications relevant for crop production; utility and exploitation of numerical weather products for the livestock industry as well as extension services (public administration, insurances, food processors and distributors, and the like) are not envisaged. The report is primarily based on the assessment of products and information available through the internet as well as in peer-reviewed journals.
The setup of the report is as follows. Currently available agrometeorological products are briefly reviewed in Section 2, while numerical weather products are presented in Section 3. The question of post-processing is treated in Section 4, and a few types of application models are presented in Section 5. Section 6 is dedicated to seasonal prediction by statistical methods. The applications discussed do not rely on the availability of weather forecasts and climate outlooks; rather, they exploit knowledge of the current conditions for estimating agronomic outcomes later in the season. While this is not strictly the subject of the terms of reference, it nevertheless provides an overview of alternative methods of forecasting that, in Europe, have been somewhat neglected during the recent decade. The report closes with some conclusions and an outlook in Sections 7 and 8.
[pic]
Figure 3.1. A possible setup of actors, activities, systems and products required for the application of numerical weather products to agrometeorological decision problems.
3.2 Currently available types of agrometeorological forecasts
As mentioned in the Introduction, there are a number of operational products available from the NMHSs and providing agrometeorological information. Examples of this type of products are the agrometeorological bulletins released by the Austrian Weather Services (ZAMG)[8], the German Weather Services (DWD)[9] (see also Fig. 1), the Servizio Idro Meteo of the Agenzia Regionale Prevenzione e Ambiente dell'Emilia Romagna (ARPA-SIM)[10], the Swiss Federal Office of Meteorology and Climatology (MeteoSwiss)[11], the French Weather services (Météo France)[12] or the U.K. Met Office[13].
These bulletins are either available on the internet or supplied on a subscription basis. The content is usually in a format suitable for the end-users, but the amount of information varies considerably from product to product. In most cases, 5- to 7-days forecasts are given for mean, minimum and maximum temperature, rainfall probability and rainfall amounts, solar radiation, potential and actual evapotranspiration (reference evapotramspiration), the climatological water budget, soil water status, soil temperature, growing degree days and phenological stages, harvesting dates. In addition, weekly and monthly retrospectives with focus the most important agro-meteorological elements are also available from various NMHSs[14]. They can be used by the stakeholders to better interpret the forecasts.
To the same category of short-term forecasts also belong operational products developed for the management of pests and diseases, for example PLASMO (Plasmopora Simulation Model, Rosa et al., 1993)[15], PHYTOPRE (a comprehensive information and decision support system for late blight in potatoes, Forrer et al., 1993; Steenblock et al., 2002)[16], or SOPRA (an operational model to forecast the appearance of apple aphid, apple sawfly apple moth, Samietz et al., 2007)[17]. Interestingly, recommendations for field operations are indeed issued based on the simulated development of pest populations, but the forecasting system does actually not make use of weather forecasts at all. In SOPRA, for instance, phenological stages for a variety of major insect pests in fruit orchards are calculated for a time span of two weeks beyond the issuing date using a combination of observed weather data up to the current date and climatological data for the projections (Samietz et al., 2007).
In addition to the products summarized so far, there is huge amount of meteorological data available through the internet and of potential usefulness in agrometeorological applications. A comprehensive list of sources was compiled by Montserrat (2001) as a contribution to the activities of COST Action 718 (Meteorological Applications for Agriculture). The report is available on the Action’s web server[18], and the reader is referred to it for further details.
3.3 Numerical weather products
3.3.1 Types for forecasts
Forecasts products are traditionally classified in the following groups (see also Das et al., 2007):
• Nowcasts, i.e. very short-range (1–3h) forecasts with a spatial resolution of the order of 1-10 km. Nowcasting goes back to the pioneering work of Ligda (1953), who showed that useful forecasts could be made based on the persistence and movement of radar echoes. Nowadays, nowcasting is essentially the identification of small-scale features based on the application of RADAR and satellite monitoring, sometimes combined with short-time forecasts and neural networks.
Predictive capabilities can be expected to grow rapidly over the next decade, despite the fact that forecasting convective cells and sever storms still represent a formidable challenge (Smith and Austin, 2000; Kryvobok, 2007). With respect to agrometeorology, valuable new services and strategies should be expected in the future, including better warnings of heavy precipitation, unhealthy atmospheric conditions, and for pollution application such as event-dependent emissions restrictions. The short time scales (1 to 2 hours) of these products requires, however, taking rapid decision.
• Very short-range forecasts (0–12 hour) of smaller-scale, short-lived, often intense weather phenomena such as tornadoes, hail storms, and flash floods. The difficulty in forecasting small-scale systems is due to insufficient computational ability, inadequate observational capabilities, and limited understanding of the physical processes that are taking place during these events. Overall, areas where these systems are likely to form can often be predicted up to 3 days in advance but more specific forecasts rapidly become unreliable with increasing lead time. Forecasts of small-scale features have improved though in regions where weather-related phenomena are generated or modulated by the topography, land-sea contrast, and land-use gradients.
Very short-range forecasts are suited for the development of warming systems. Observing and detecting when conditions are favourable for the development of severe convection and then monitoring each storm's evolution remain key tasks of operational very short-range forecasts, and the interpretation of radar and satellite imagery and local spotters by the forecasters play a critical role.
• Short-range forecasts (12–72 hour) of the evolution and movement of large- and medium-sized weather systems. Accuracy and reliability of the forecasts decreases rapidly with decreasing spatial scales of the key weather systems and with increasing lead time; nevertheless, the ability to predict the evolution of major, larger-scale weather systems have continuously improved over the last decades. Forecasts of precipitation amounts 36-60-hours ahead are now more accurate than corresponding 12-36-hour predictions were during the late 1970s.
Details of precipitation patterns are often tied to smaller-scale structures such as fronts, thunderstorm outflow boundaries, and mesoscale convective systems that are still difficult to manage for the current generation of numerical models. The introduction of high-resolution models with more advanced physics and the advances in observing systems give reason to believe that further improvement in forecasts of precipitation is likely.
• Medium-range forecasts (3–7 days) of synoptic scale systems. Medium-range forecasts have shown significant improvement in the last two decades. For instance, three-day forecasts of major low pressure systems that determine the general evolution of the weather are more skilful today than 36-hour forecasts were 15 years ago. Also, in the late 1970s day-5 forecasts of precipitation were no more accurate than climatology. Since then, skill of day-5 forecasts has more than doubled, with predictions now being as skilful as day-3 forecasts were a decade ago.
Temperature forecasts have also improved and now show considerable skill on day 3, with the skill decreasing rapidly beyond this lead time. However, there is reason to believe skilful day-7 forecasts will be possible in the future given the steady improvements in computer models, observational approaches, and forecast strategies.
• Extended-range forecasts (2 weeks). The predictability of the day-to-day weather beyond day 7 is usually small. Operationally, forecasts at these time ranges have taken the form of 6-10-day mean temperature and occurrence of precipitation departures from normal. The accuracy of these 5-day mean temperature and precipitation forecasts has more than doubled since the 1970s. The accuracy of precipitation forecasts is less than that for temperature, even though the skill of both has increased at about the same rate. Advances in observing systems, computer models, and statistical techniques are the reasons to believe that skilful forecasts of the mean rainfall conditions out to week 2 will be feasible in the near future.
• Probabilistic seasonal forecasts (1–6 months). Due to the chaotic nature of the atmosphere it is not possible to deterministically predict day-to-day weather changes in detail on a monthly to seasonal scale. However, more or less reliable forecasts can be issued for the likelihood of weather patterns. At the basis of seasonal forecasting is the observation that variations of sea-surface temperature are relatively slow and can therefore be reasonably well predicted up to several months ahead in some parts of the world. Accordingly, dynamically coupled climate models represent the core of seasonal prediction systems.
The strongest relationship between sea surface temperatures and seasonal weather trends are found in tropical regions. Strong signals are associated with the El Niño Southern Oscillation (ENSO), which can disrupt the global pattern of normal weather including large changes in seasonal rainfall patterns (droughts in some regions and floods in others). Weaker links between sea surface temperature and average seasonal conditions are found in other parts of the globe. For regions under the influence of the El Niño Southern Oscillation (ENSO), recent advances in understanding the climate system have allowed successful forecasts of temperature and precipitation at lead times up to the season. This has promoted the application of seasonal predictions to various fields.
Concerning Europe, the influence of ENSO on the atmospheric circulation regime has still to be completely understood (Brönnimann, 2007). Nevertheless, Pavan et al. (2003) found that ENSO can explain up to 30 % of the weather variability observed in the North Atlantic/European sector.
3.3.2 Forecasting systems
Deterministic forecasts to the medium range and probabilistic predictions out to the season are operationally carried out through Europe. To large extent, the systems employed are developed and maintained in the framework of international centres or coordinated activities. This holds true for the following systems: the prediction systems by ECMWF; the Unified Model, i.e. the suite of atmospheric and oceanic numerical modelling software developed and used at the U.K. Met Office[19]; ALADIN[20], the system developed in collaboration among the NMHSs of Algeria, Austria, Belgium, Bulgaria, Croatia, Czech Rep., France, Hungary, Morocco, Poland, Portugal, Romania, Slovakia, Slovenia, Tunisia and Turkey; the prediction system of the Consortium for Small-Scale Modeling (COSMO), with partners from Germany, Switzerland, Italy, Greece, Poland and Romania; the HIRLAM prediction system[21], with partners from Denmark, Estonia, Finland, Iceland, Ireland, The Netherlands, Norway, Spain, and Sweden.
Prediction systems cover spatial domains that extend from the global to the regional scale. Correspondingly, a full palette of models is employed for operational forecasting, including global as well as regional models and high-resolution, limited-area models (LAMs). To obtain a consistent set of predictions, it has nowadays become customary to apply nesting techniques, whereby forecasts from the low-resolution models are used to infer boundary conditions for the high-resolution simulations (Fig. 3.2). Further details about model nesting and operational schedules can be found e.g. in the information provided by the U.K. Met Office[22] and the Consortium for Small-Scale Modeling (COSMO)[23].
Due to the fact that knowledge of the state of the world ocean is necessary to predict atmospheric variability on time scales of months, coupled ocean-atmosphere models and specific ocean analysis systems are used to issue ensemble forecasts on lead times ranging from 1 to 6 months. At ECMWF, for instance, the operational system used for monthly and seasonal forecasts includes: (i) an ocean model, developed at the Max Plank Institute for Meteorology in Hamburg (HOPE), run at horizontal resolution of 1.4° (~155 km in the N-S direction) in the extra-tropics; and (ii) an atmosphere model run at horizontal resolution of 1.125° (~125 km in the N-S direction) for the monthly forecasts and of 1.875° (~210 km in the N-S direction) for the seasonal forecasts[24].
[pic]
Figure 3.2. The three nested numerical weather prediction models of the COSMO system: the global model operated by the European Centre for Medium Range Weather Forecast , and two versions of the COSMO model, the first running at 7 km, the second at 2 km spatial resolution. Source: COSMO[25]
To account for the fact that (i) time integration of the dynamic equation governing the atmosphere is sensitive to the specification of the initial conditions and (ii) in simulations of the coupled ocean-atmosphere system many atmospheric states can potentially be consistent with a given set of boundary conditions, monthly forecasts and climate outlooks are carried out as ensemble predictions. The number of members varies depending on the system. At ECMWF, for instance, monthly forecasts[26] consist of 51 individual members, whereas climate outlooks comprise 41 individual members[27]. In summarizing the information provided by the ensemble, results from the individual model simulations are usually aggregated to produce probability distributions for time-mean departures from the climatic norm, and the probability distributions are further stratified in terms of probability terciles.
Despite the advances in the field of seasonal forecasting in particular during the last decade, work conducted in the framework of DEMETER (Development of a European Multimodel Ensemble System for Seasonal to Interannual Climate Prediction)[28] has shown that ensemble predictions issued from a single forecasting system still suffer from weaknesses introduced by uncertainties in the model equations and numerical inaccuracies related to their integration. To improve this situation, so-called multi-model ensemble predictions have been tested first in PROVOST (Prediction of Climate Variations on Seasonal to Interannual Time Scales) (Doblas-Reyes et al., 2000) and later in DEMETER (Palmer et al., 2004). Results from DEMETER clearly indicated that multi-model forecast systems do indeed provide on a higher skill than achievable using a single-model ensemble system, at least on the average (Doblas-Reyes et al., 2005 and 2006; Hagedorn et al., 2005).
Seasonal forecasting requires large installed computing power and is therefore not appropriate to be performed at the national level. Instead seasonal forecasting products are provided by so called global producing centres (GPCs) and are distributed to the NMHSs under different conditions and formats. Recommendations for the standardization of products available to the NMHSs were formulated by a task team on provision of seasonal to inter-annual prediction in RA VI. The report of the task team is available on the internet[29]. Meanwhile the list of GPCs consists of 9 officially designated WMO GPCs and 4 other centres that provide global seasonal forecasting products; the list containing web links is available on the CLIPS web page.
3.3.3 Skill of weather forecasts and climate outlooks
Thanks to enhanced capabilities of the forecasting systems in terms of observational data and data assimilation, numerical algorithm and physical parameterizations, as well as spatial resolution over the last decades (Kalnay, 2003; Lynch, 2006) considerable progress has been achieved in terms of skill and lead-time of deterministic forecasts (Simmons and Hollingworth, 2001). As seen in Fig. 3.3, operational global models run in a deterministic mode are now able to produce useful forecasts of the large-scale weather patterns at a lead-time of roughly 8 days.
Less satisfactory is the performance of monthly and seasonal forecasting systems with respect to the extra-tropics, in general, and Europe, in particular. Rodwell and Doblas-Reyes (2006) found that probabilistic forecast skill for temperature exists over the first month. On the other hand the predictability limit of precipitation is only in the order of several days (Buizza et al., 1999). For precipitation, one possible reason for the rapid decrease in skill with increasing lead time is the relatively low spatial resolution of the prediction systems (Mullen and Buizza, 2002).
As noted by Weigel et al. (2008), while processes with pronounced variability on the intra-seasonal time scale, specifically the Madden-Julian oscillation (Madden and Julian 1971) and stratosphere-troposphere interactions (e.g. Baldwin et al. 2003), do have the potential to act as sources of predictability in the North Atlantic/European sector, the key aspects of their influence on the atmospheric circulation patterns is not yet captured by operational global circulation models (e.g. Vitart et al. 2007; WCRP 2008). In their assessment of the performance of the ECMWF monthly ensemble forecasting system, Weigel et al. (2008) found that even in relation to near-surface temperature extratropical continental predictability essentially vanishes after 18 days of integration.
[pic]
Figure 3.3. Evolution of the skill of the operational ECMWF prediction system for Europe, 1980 to 2008. The skill is expressed in terms of anomaly correlation scores (correlation between forecast anomaly and verifying analysis anomaly). Source: ECMWF[30]
3.3.4. Standard products from high-resolution weather forecasts
Agrometeorological applications of deterministic forecasts should in principle be in the position to take advantage of the massive enhancement in spatial resolution of atmospheric models that has taken place over the last few years. At ECMWF, for instance, the increment in the resolution of the global model has been from T106L19 in 1985 (100 km horizontal resolution in the mid latitudes, and 19 vertical levels) to T799L91 (approximately 25 km horizontal resolution and 91 vertical levels) at present[31]. Similar advances have been achieved with respect to regional models and LAMs. Both COSMO as well as the U.K. Met Office, for instance, are now operating LAMs at a horizontal resolution of 1 to 3 km, making it possible to realistically depict topographic features that are crucial for obtaining a realistic depiction of the local weather in complex terrain.
[pic]
Figure 3.4. The benefit of higher resolution in complex topography. Simulation of 30 meter above ground wind field in Canton de Vaud, with COSMO-7 (right plot) and COSMO-2 (left plot). Source: COSMO[32]
To illustrate the possibilities offered by the new generation of operational LAMs, Fig. 3.4 illustrates the remarkable progress achieved with the COSMO model in simulating the near-surface wind field in the Alpine region.
As a result of this evolution, short-term forecasts are now available for a full palette of meteorological elements. This is illustrated by the catalogue of products offered by COSMO[33]. Operational products include among others:
• surface and screen level temperature;
• dew point and relative humidity at screen level;
• total precipitation (rain and snow) for 1, 3 or 6 hours, respectively, and sum of total precipitation (rain and snow) for 3, 6, 12 or 24 hours, respectively;
• fraction of snow in total precipitation;
• wind direction and speed and wind gust at 10 m above surface;
• surface pressure and 3 hours pressure tendency;
• cloud cover (total, high, mid-level, low), and global radiation (hourly averages) and 3 hour sums of evaporation;
• vertical profiles of temperature, dew point and wind at the locations of SYNOP radiosoundings, and height of the zero degree line.
In addition, specialized products are available that include among others:
• water content of the soil layers;
• net radiation at (shortwave/longwave) and alberdo of the surface;
• photosynthetically active radiation at the surface;
• sensible and latent heat fluxes at the surface;
• stratiform and convective rain, as well stratiform and convective snowfall;
• surface runoff and runoff within soil.
It is clear that the availability of numerical weather products at a spatial resolution of the order of 1 km considerably reduces the needs for post-processing. In the next section we therefore consider the question of spatial downscaling and time disaggregation exclusively from the point of view of probabilistic forecasts (monthly to seasonal scale).
3.4 Post-processing of probabilistic forecasts
In relation to monthly forecasts and climate outlooks, the necessity to post process operational products arises for at least three reasons (see e.g. Feddersen and Andersen, 2005; Pavan et al., 2005; Doblas-Reyes et al., 2006; Hansen et al., 2006): (i) the relatively low spatial resolution of the prediction systems (of the order of 100 km at the best in the mid-latitudes); (ii) the need to correct for inaccuracies/errors in the predicted fields; and (iii) the need to recover hourly or daily values from time-aggregated statistics.
3.4.1 Statistical downscaling
In practice, the first two issues are often addressed in conjunction with the help of statistical/empirical spatial downscaling. Different approaches are accessible (e.g. Wilby and Wigley, 1997), including analogue techniques, multiple linear regression (MLR), singular value decomposition analysis (SVDA), canonical correlation analysis (CCA), empirical orthogonal functions (EOFs) and, in recent years, non-linear statistical methods such as artificial neural networks (Huth, 2002; Huth et al., 2008). In all cases, the approach essentially consists in mapping patterns of large-scale predictors onto realizations of local variables or predictands.
To obtain satisfactory results, the design of the method is as crucial as it is the choice of predictors (Feddersen et al., 1999; Huth, 2002). In the past, a considerable amount of research has been conducted to find suitable predictors for surface or screen level temperature and precipitation, but little effort has been produced to find appropriate predictors for other variables of interest in agrometeorology, such as humidity variables or radiation. With respect to the latter, only a few studies can be cited: Enke and Spekat, (1997) in relation to sunshine duration, Kaas and Frich (1995) and Enke and Spekat (1997) for cloudiness, Karl et al. (1990) for cloud ceiling height, and Enke and Spekat (1997), Weichert and Bürger (1998) and more systematically Huth et al. (2005) concerning humidity variables.
Recently, investigations of statistical approaches to the generation of high-resolution atmospheric fields have been extended to the downscaling of extreme events. Hundecha and Bárdossy (2008) tested two different downscaling models in terms of their ability to reproduce observed indices of daily extreme temperature and precipitation. In their analysis, variables derived from reanalysis data were used as predictors. Performance of the models in capturing precipitation indices generally was found to vary between seasons and indices. For both models, the skill was highest in winter and lowest in summer, suggesting that statistical downscaling can more reliably be done during seasons when the local climate is determined by large scale circulation than local convective processes. Moreover, it was found that both models tend to underestimate the inter-annual variability of temperature and precipitation indices in all seasons.
The possibility to statistically downscale extreme precipitation events in complex topography was also examined by Busuioc et al. (2008) with respect to the winter season. Probably the most interesting outcome of this study was that statistical downscaling is not stable in time, as eventually the model setup depends on the time window selected for calibration. The reason is that the relationships between predictors and predictands are affected by decadal changes in the atmospheric circulation patterns. Regular recalibration of the models is therefore required to ensure an appropriate performance.
3.4.2 Stochastic weather generators
In climate change studies the disaggregation of time-mean anomalies is usually accomplished with the help of stochastic weather generators, and it is natural to think that these tools can also be used in combination with probabilistic monthly or seasonal forecasts (Wilks, 2002).
Most of the weather generators currently in use in Europe can be traced back to the models proposed by Richardson (1981), on the one hand, and Rackso et al. (1991), on the other hand. Examples are Met&Roll (Dubrosky, 1995 and 1996), for the former type, and LARS-WG[34] (Semenov et al., 1998), for the latter. A good overview of the procedures used in the two types of generators is given in Tab. 1 of Semenov et al. (1998), and to recap the main differences between the two types of generators are as follows:
• in WGEN-type generators (Richardson et al., 1984), the succession of wet and dry periods is simulated using transition probabilities of a 2-state Markov chain, while in LARS-WG the length of dry and wet periods is obtained from a fitted, semi-empirical distribution;
• in WGEN-type generators precipitation intensity is modelled as a 2-parameter gamma distribution, whereas a semi-empirical distribution is used in LARS-WG;
• WGEN-type generators adopt a normal distribution to model both temperature as well as radiation, but in LARS-WG the latter is again computed on the basis of a semi-empirical distribution;
• Finally, in WGEN-type generators, temperature is not conditioned on precipitation, but it is in LARS-WG.
Weather generators were originally designed to provide synthetic series of weather data at a single site, but the possibility to use the models to generate spatial information has been considered in recent years. Semenov and Brooks (1999), for instance, discuss an interpolation method for LARS-WG output that combines the local interpolation of the weather generator parameters with the use of globally interpolated monthly mean statistics, applying thin plate smoothing splines with elevation as an independent variable for the interpolation of mean monthly rainfall and temperature.
In Wilks (1998) the WGEN-type chain-dependent-process stochastic model of daily precipitation is extended to simultaneous simulation at multiple locations by driving a collection of individual models with serially independent but spatially correlated random numbers. Next, Wilks (1999a) discusses the use of a high-dimensional autoregressive model for the spatial simulation of time series of temperature and solar radiation. Finally, Wilks (2008) describes an approach to defining weather generators at locations for which no real data exist through parameter interpolation with weighted local regressions. As pointed out by Wilks (2008) this method can capture nonlinear parameter variations in space and allows objective and automatic selection of both the regression predictor and the size of each local neighbourhood of influence.
A further aspect in relation to the use of stochastic weather generators is addressed by Wilks (1999b), who investigated the capacity of different formulations for the occurrence and amount of precipitation to represent inter-annual variability in precipitation, extreme precipitation amounts and the persistence of wet or dry conditions. His results suggest that in respect to variability and extremes the conventional formulation of WGEN-type models, consisting of first-order Markov dependence for precipitation occurrence and Gamma-distributed precipitation amounts does not always provide an adequate description of the observed characteristics.
Problems in the representation of extreme events do also appear in applications of LARS-WG (Semenov, 2008). While means of yearly maxima for daily precipitation and 10 and 20 years return values appear to be reproduced accurately by this stochastic generator, the performance of LARS-WG in reproducing means of yearly maxima for daily maximum temperature is less satisfactory.
So far, most of the research concerning the application of statistical post-processing has been done in the framework of climate change studies (Wilby and Wigley, 1997). Experiences with the downscaling of probabilistic, long-range forecasts are more limited (e.g. Misra et al., 2003; Hansen et al., 2006). In Feddersen and Andersen (2005), for instance, the downscaling from seasonal ensemble predictions to daily precipitation time series for individual stations was performed in three steps: (i) a spatial downscaling of ensemble mean seasonal means from dynamical model output to station level by means of patterns derived from a singular value decomposition analysis of model output and observations; (ii) application of the downscaling transformation to the model output ensemble and subsequent calibration of the downscaled ensemble; (iii) a stochastic generation of daily precipitation conditioned on predictions of the probability of a wet day in the season and daily persistence. In the majority of the examples, the downscaling was found to provide more skilful predictions than the raw dynamical model output.
Pavan et al. (2005), on the other hand, tested a combination of multiple linear regression obtained using the method of the best linear unbiased estimate, standard statistical downscaling technique and empirical selection of the predictands, to downscale winter DEMETER forecasts over a limited area. The results indicate that predictions obtained with this approach have a much higher detail than the DEMETER direct model output predictions and, in parts of the domain, they are characterized by substantially significant skill.
A systematic investigation of stochastic weather generation conditional on seasonal forecasts was also conducted by Wilks (2002). His analysis relies on the results of Briggs and Wilks (1996), who proposed that sub-seasonal statistics consistent with a given probabilistic seasonal forecast could be estimated by re-sampling the observed climate record for a location according to the probabilities in the forecast. It is shown by Wilks (2002) that only a subset of the parameters characterising a WGEN-type weather generator (proportion of wet days, precipitation mean parameters on wet days, and daily temperature means and standard deviations) depends appreciably on the seasonal temperature and precipitation outcomes. These results support therefore the conclusion, that stochastic simulation of multiple daily weather series conditional on seasonal forecasts could be implemented without further ado as part of an operational post-processing.
3.4.3 Bias correction of precipitation forecasts
It was already remarked, that the correct simulation of precipitation events and amounts remain one of the big challenges in weather forecasting. Systematic errors in predicted precipitation can have serious consequences for the application of seasonal forecasts to agrometeorological decision problems and require therefore an apposite treatment. In particular, it is known that even the current generation of atmospheric models systematically overestimates the frequency of rainfall events in the extra-tropics, underestimating at the same time rainfall intensity (e.g. Frei et al., 2003; Ines and Hansen, 2006; Schmidli et al., 2006).
As opposed to temperature, for which bias in the model output can typically be corrected with a seasonally varying additive shift, the correction of forecasted precipitation is more problematic. Simple multiplicative shift do in general a good job of correcting monthly and seasonal rainfall totals (Ines and Hansen, 2006) but does not remove the substantial bias in frequency and intensity. For this reason, Ines and Hansen (2006) considered a procedure that calibrates both the frequency and intensity distribution of daily model outputs. In a first step, rainfall frequency is corrected by fitting a threshold value to truncate the empirical distribution of the simulated daily rainfall, such that the mean frequency above the threshold matches the observed mean rainfall frequency. This done, bias in the intensity distribution is eliminated by mapping the cumulative distribution function of daily rainfall amounts above the calibrated threshold onto the observed intensity distribution.
A two-step correction of precipitation bias was also discussed by Schmidli et al. (2006). Starting from the same frequency correction as in Ines and Hansen (2006), their method proceeds by scaling the precipitation intensity on wet days with the ratio of observed and simulated mean wet-day intensities, the latter being conditional on exceeding the calibration threshold.
Although both procedures are able to constrain simulated precipitation to have the same climatological wet-day frequency and intensity as the observations, in general they are not able to improve the skill of the forecasts. The reason for this situation is that both procedures do not address the autocorrelation structure of rainfall time series, which is important e.g. for understanding the persistence of wet or dry spells (Ines and Hansen, 2008). In fact, it can be shown (Bolius et al., 2009) that both procedures introduce a conditional bias that negatively affects skill scores. Explicitly this can be illustrated with respect to a skill score measure based on the mean-square error using the decomposition proposed by Murphy (1988).
3.5 Application models for agrometeorological forecasts
During the last decades, the use of simulation models to support agrometeorological decision making has increased considerably. Statistical or mechanistic models exist for addressing a range of questions. Besides those already discussed in Section 2, viz. models for the management of pests and diseases, one can find models for forecasting frost occurrence (e.g. Katz et al., 1982; Stewart et al., 1984; Ghielmi and Eccel, 2006), models for the management of water resources (see e.g. the review of irrigation models provided by the COST Action 718[35]), or even models for the control of storage facilities (e.g. Keesman et al., 2003). As it is not possible to provide a complete overview, in the following we restrict our attention to two classes of models; crop growth simulation models, on the one hand, and models for predicting the moisture content of grains in cereals and hay, on the other hand.
3.5.1 Crop growth simulation models and yield forecasts
A summary of models with potential for the operational assessment of crop status and yield forecasts has recently been compiled by Alexandrov (2002). Various sources of information have been consulted during the preparation of that report, resulting in a collection of more than 50 entries. Tables 2 to 6 in Alexandrov (2002) provide a good overview of the models and model versions, the references and the countries for which the application of crop growth models is documented.
Crop growth models can provide timely information for a number of farm operations. Moreover, they can be used to produce yield forecasts at the regional and national scale. Worth mentioning in this context are the crop yield predictions prepared by the Agriculture Unit of the Joint Research Centre (JRC) of the European Commission[36] to provide the Directorate General for Agriculture with real-time yield estimates during the growing season for the European Union member states. The forecasting system is built on the Crop Growth Monitoring System (CGMS) developed by Alterra and which makes use of the agrometeorological model WOFOST (WOrld FOod Studies) (Supit et al., 1994). At JRC, crop yield forecasting is carried out in the framework of the so-called AGRI4CAST action (ex MARS-STAT), whose web page[37] provide all necessary information.
A peculiar aspect of AGRI4CAST is that yield forecasts are actually issued without accessing weather forecasts and climate outlooks (Fabio Micale, pers. comm.), that is to say using observed weather data are used up to the current date and climatological fields for the remaining of the growing season. However, the possibility to combine seasonal forecasts and the CGMS is currently being tested and has been examined in the framework of the DEMETER project.
In DEMETER, the general approach was to link WOFOST to each member of the ensemble weather forecasts to obtain full probability distributions of yield. As the spatial resolution of the involved atmospheric models was inadequate to represent weather variations at the local to regional scale (see Table 1 in Palmer et al., 2004), statistical downscaling and stochastic weather generators were applied to obtain daily input fields on a representative spatial grid (Feddersen and Andersen, 2005; Marletto et al., 2005 and 2007).
Using this method, Cantelaube and Terres (2005) examined the reliability of yield forecasts for wheat. Results of the simulations driven by the DEMETER downscaled forecasts were compared to both yield statistics as well as the results of the JRC operational system. Two main conclusions can be drawn from this exercise. First, it was clearly shown that yields forecasts issued in February using the DEMETER data set were more accurate than the operational forecasts issued in July. This means a significant potential for extending the lead time of yields forecasts. Second, it was found that reliable predictions could be obtained only for central and northern Europe, while results for Mediterranean countries such as Italy, Spain and Portugal were less positive. As pointed out by Cantelaube and Terres (2005) this may be due to the fact that the DEMETER forecasts were not able to correctly simulate one specific drought event, namely the one affecting Spain and Portugal in 1995/1996.
While in Cantelaube and Terres (2005) the focus was on large-scale yield predictions, Marletto et al. (2005) tested the feasibility of using seasonal weather forecasts for yield predictions at the local to regional scale. The utility of weather forecasts at various lead times, viz. 90, 60 ad 30 days, was examined. Crop yield simulations driven with observed weather date up to the current date and climatological data for the remaining of the growing season were used as a benchmark. For lead times of 90 and 60 days, the adoption of multi-model, downscaled ensemble forecasts from DEMETER did not improve skill and reliability of the yield predictions. Improvements were obtained, on the other hand, for simulations at 30 days lead time. However, it is likely that this result simply reflects the fact that a large percentage of the variability of yields is already explained by the variability in above-ground biomass one month before harvest (Lawless and Semenov, 2005; Marletto et al., 2007).
The analysis in Marletto et al. (2005) was extended by Marletto et al. (2007) by comparing simulations of winter wheat yield with actual field data. For these simulations, WOFOST was driven with observed weather data up to 60 days before harvest and downscaled DEMETER forecasts thereafter. With a few exceptions, yields forecasts were able to capture the year-to-year variability of observed yields over a ten year period (1977-1987). Based on this outcome, Marletto et al. (2007) concluded that properly downscaled seasonal forecasts for surface air temperature and precipitation can provide a reliable input for crop predictions even in regions characterized by complex topography.
3.5.2 Crop growth simulation models and forecasts of nitrogen requirements
Many if not all of the mechanistic crop growth models reviewed by Alexandrov (2002) are more than just mathematical models for predicting yield. As they incorporate equations for describing the fluxes of water and nutrients in the soil-plant-atmosphere continuum, they can be use to simulate the effects of management and ultimately to optimize the use of resources as affected by the environmental conditions.
Managing the application of nitrogen fertilizers is a difficult task, as a balance must be achieved between covering the crop needs while at the same time minimizing the environmental impacts. Because nitrogen has been a relatively cheap input and the direct consequences of applying too little are yield and income losses, farmers have tended to push fertilization beyond the physiological needs of the crops. In view of various environmental regulations (climate and water) that became effective during the last decades, the option to over apply nitrogen should be discarded, making place to a more conscious use of fertilizers.
Running crop growth model in conjunction with weather forecasts offers means for predicting the nitrogen needs by the crops and the nitrogen availability in the soil, providing the necessary input for optimizing fertilization. As an example of an attempt to quantify the potential of medium-term weather forecasts to improve the accuracy and benefit of nitrogen fertilizers recommendations we briefly discuss in the following the work of Dailey et al. (2006).
In a series of computer simulations with SUNDIAL (Smith et al., 1996), Dailey et al. (2006) tested the effects of prior knowledge of weather following the date of nitrogen fertilizer application on the nitrogen use efficiency of arable crops in the U.K. Changes in nitrogen losses to the environment and plant uptake due to forecast quality were calculated, and yield and gross profit changes were estimated from nitrogen uptake for the arable industry in England and Wales.
They found that the benefits of using seasonal forecasts were small with respect to the environmental impacts. However, reliable weather forecasts with lead times ranging from a few weeks to 6 months were shown to systematically and significantly improve crop nitrogen uptake and reduce the risk of under-application of nitrogen. Concretely, Dailey et al. (2006) estimated that a perfect forecast out to 3 weeks could increase farm profits at the national scale in the order of £20M annually, with benefits in the order of £100M annually for a perfect 6-month forecast.
With this in mind, Dailey et al. (2006) argued that although the benefits of medium-term weather forecasts to crop nitrogen offtake do not appear to be large, the distribution of crop responses is typically skewed towards the risk of yield loss from nitrogen scarcity. Assuming that farmers in England and Wales would strictly adhere to a nitrogen prediction system to minimise losses, Dailey et al. (2006) calculated that the benefits of even a limited weather forecast (3-week forecast, 50% reliability) could be worth £10M per year at the national scale, increasing to about £10M per year for extended forecasts (27 weeks, 90% reliability).
3.5.3 Models for harvest scheduling
A class of models that have not yet received much attention by the RA VI Working Group on Agricultural Meteorology are models for predicting drying time for harvested hay and cereals, which is essential information for harvest scheduling. Quantitative predictions of the moisture content of grains are of paramount importance in agronomy. Because of the risk of mould formation, too high a moisture content makes the products unsuitable for storage. But too low a moisture content is also undesirable because it has negative impacts on the profit, as less moisture is sold.
Rephrasing, we could say that too early a harvest has negative impacts on grain quality. However, delayed harvest in persisting cool wet weather conditions can also result in reduced grain quality (Hayward, 1987). In some cases, therefore, early harvest is beneficial even though there is a risk of incurring in extra costs for artificial drying. In this situation, medium-range forecasts with lead times up to 15 days could help assessing the occurrence/persistence of favorable/unfavorable conditions and defining the best overall harvest strategy, while short-term forecasts, including forecasts of the moisture content of the grains, would provide means for scheduling specific field operations.
There are two main types of models for calculating the moisture content of cereals and hay: empirical or semi-empirical models and mechanistic models. Models based on regression equations are for instance those developed by Voigt (1955), Boyce (1965), Crampin and Dalton (1971), Smith et al. (1981) or Philips and O’Callaghan (1974) for cereals, and Dyer and Brown (1977) and Gupta et al. (1989) for hay. Appropriately calibrated, statistical models are able to simulate the evolution of the moisture content after harvest (including night-time rewetting caused by dew formation) as observed in field studies (Fig. 3.5).
[pic]
Figure 3.5. Comparison of simulated and observed moisture content (% wet basis) of conditioned and unconditioned ladino (Trifolium repens L.) hay at Guelph, Ontario, in 1964 (Dyer and Brown, 1977).
The success of predictions with statistical models has prompted Dyer and Baier (1981) to test the possibility to use weather forecasts to improve hay-making reliability. Since hay (or cereals) drying takes place only over a few days, only short-term forecasts were considered. Dyer and Baier (1981) found that scheduling hay cuts could indeed benefits from the use of forecasts, but also argued that for two-, three- or four-day forecasts of dry weather to be more valuable than chance, the reliability of the forecasts must exceed a critical level typically in the range 30 to 50%.
Contrasting empirical models are physical models such as those developed by van Elderen and van Hoven (1973) and more systematically by Atzema (1992, 1993a, 1993b, 1993c) for hay and main cereals. In these models, predictions of the moisture content are realized in four steps; (i) by calculating the drying potential via the Penman-Monteith equation (Monteith, 1965); (ii) by evaluating the equilibrium moisture content via Henderson (1952) equation; (iii) by estimating the amounts of free and absorbed water; and, (iv) by computing rewetting through dew formation and rain. In all models, hourly values of air temperature, global radiation, dew-point temperature, wind speed are required on input.
Atzema (1992, 1993a, 1993b, 1993c) showed that these models can capture variations in the moisture content observed in field trials to within a few percent, and on this background Atzema (1995 and 1998) tested the feasibility of predicting grain moisture content with knowledge of the meteorological conditions a few days in advance. He found that predictive skill exist on day 1 and 2, but rapidly decline after day 3. In forecast moisture content of cut grass, skill loss was mostly caused by inaccurate forecasts of precipitation, air temperature and global radiation. For wheat and barley, dew-point temperature and wind speed also played a role.
It is likely that owing to the improvements in forecast skill over the recent decade (see Section 3.3) the results of Atzema (1995 and 1998) are overly pessimistic. Even so, his analysis indicates that this type of models is very sensitive to he inputs, which calls for extensive verification of the model results in operational applications.
3.6 Seasonal prediction by statistical methods
3.6.1 Wheat grain quality in the U.K.
The reflections on the skill of seasonal forecasts (Section 3.3.3) and the experiences gained in DEMETER and by the JRC in the framework of AGRI4CAST (Section 5.4) appear to indicate that the application of seasonal forecasts to agrometeorological decision problems is, for the time being, precluded by the unreliability of probabilistic weather predictions out to the month or the season. There are, however, application problems for which seasonal predictability can be realized without having to resort to seasonal weather forecasts, but exploiting lag-relationships between meteorological predictors and agronomic predictands need to be established.
As an illustration we discuss here the question of wheat grain quality in the U.K. Grain quality is usually quantified in terms of crude protein concentration, Hagberg falling number and specific weight (Gooding and Davies, 1997). The three measures have to individually exceed some given threshold for wheat to be accepted for milling, which is the condition for obtaining a price premium. As noted by Smith and Gooding (1999) predictions of grain quality at harvest time would therefore be of considerable value to grain buyers and farmers, who could use this information for optimizing late-season agronomy. In fact, in the U.K. a deficit in crude protein concentrations is usually associated with early nitrogen losses induced by increased rainfall during late winter and spring. Notwithstanding, quality can usually be improved with nitrogen applications at later growth stages.
Based on data collected between 1975 and 1995, Smith and Gooding (1999) found that (Fig. 3.6): (a) Grain protein concentration is negatively associated with late winter/spring rainfall, but positively correlated with summer temperature. About 70% of the inter-annual variability observed in national mean crude protein concentration at harvest can be explained by winter/spring rainfall patterns, and are therefore predictable as soon as in May; (b) Hagberg falling number is positively associated with summer temperature. It is only in August that 70% of the inter-annual variability can be explained by meteorology, indicating little predictability at lead times of more than a few weeks; (c) specific weight is negatively associated with previous autumn rainfall patterns, and positively associated with winter temperature. About 50 % of the inter-annual variability is predictable from knowledge of these two parameters, suggesting a potential lead time of the order of the season.
[pic]
Figure 3.6. The effect of using successive weekly meteorological data on the percentage variance accounted for in annual variations of national grain quality. Smith and Gooding (1999).
These findings and the observation of a the quasi-cyclical variations characterizing the grain quality of U.K. wheat, which are reflected in variations of premium prices, prompted Kettlewell et al. (1999) to examine whether variations in grain quality could be explained as a response to a NAO forcing. They found that in the U.K. two of the quality attributes, namely Hagberg falling number and specific weight, are indeed positively correlated to the winter NAO index. They also noticed that the influence of the NAO on cereal quality extends to other parts of Europe.
In two following studies, Kettlewell et al. (2003) and Atkinson et al. (2005) showed that the NAO influence on wheat grain quality stems from a negative/positive association between winter NAO and summer rainfall/radiation, that is to say from a memory effect in the climate system. Specifically, they tested the hypothesis that a high winter NAO index could lead to sunnier and drier weather during grain growth and ripening, giving better grain quality. Prediction of two key summer climate variables (cumulative sunshine between anthesis and the end of grain-filling, and unconditional probability of a wet day between the end of grain filling and harvest) from knowledge of the NAO status in January accounted for 70% of the variance of the relationship between the January NAO and specific weight.
Based on these results, it can thus be argued that knowledge of the NAO status in winter is sufficient to accurately predict wheat grain quality at harvest during the following season, at least in the U.K; whether this also applies to grain quality in other areas of Europe or to other agronomic problems, remains to be verified. But for sure, there is a considerable interest for better understanding the NAO and answering the question of its predictability.
3.6.2 The North Atlantic Oscillation and its predictability
The North Atlantic Oscillation (NAO) is the dominant pattern of atmospheric circulation variability in the North Atlantic region, explaining most of the European winter lower to mid-tropospheric flow variability on a wide range of timescales from months to decades (van Loon and Rogers, 1978; Hurrell, 1995). Recent investigations provide evidence that the NAO is at least partially responsible for variations in storm tracks across the North Atlantic (Rogers, 1997), atmospheric blocking variability (Scherrer et al., 2006), European summer climate (Sutton and Hodson, 2005), including summer heat waves (Cassou et al., 2005).
The NAO is known to have significant effects on marine and terrestrial ecosystems. Ottersen et al. (2001) found for instance that the ecological responses to the NAO encompass changes in timing of reproduction, population dynamics, abundance, spatial distribution and inter-specific relationships, and proposed a classification of the NAO effects in direct, indirect and integrated. These findings were also confirmed by Blenckner and Hillebrand (2002).
It is clear that not all of the responses of natural ecosystems highlighted by the Ottersen et al. (2001) and Blenckner and Hillebrand (2002) do also appear in relation to agricultural ecosystems. But one can assume that phenology and crop growth reflect to some extent the variability in regional climate as constrained by the NAO. This is attested by the results discussed in the previous section.
This said, for the operational application of numerical weather products the question arises as to whether the NAO itself is predictable, as it is the case for the ENSO. If this were the case, predictability could be even extended beyond the season to the annual scale.
Progress in understanding and predicting the NAO was recently reviewed by Bojariu and Gimeno (2003). They found that predictive potential exists in relation to Atlantic sea-surface temperature anomalies preceding specific phases of the NAO by up to 6 months, atmospheric temperatures anomalies in the previous November, Eurasian snow cover and sea-ice extent over Arctic. Moreover, they noticed that the use of simulations based on ensemble prediction to estimate potential predictability shows the possibility of capturing the upward trend of the NAO and suggests that multiannual to multi-decadal variations in the NAO are more predictable than inter-annual fluctuations.
This latter conclusion was motivated by the findings of Rodwell et al. (1999), Hoerling et al. (2001) and Rodwell and Folland (2002). Common to these studies is the conclusion that the NAO primarily responds to variations in the state of the ocean. Differences emerge, though, concerning which portion of the ocean appears to be determinant. Rodwell et al. (1999) and Rodwell and Folland (2002) identified the North Atlantic sector as the key area, whereas Hoerling et al. (2001) came to the conclusion that the NAO is controlled by the thermal structure of the tropical Ocean. Summing up, many processes remain to be understood before the NAO prediction problem can be treated at the same level as the ENSO prediction problem.
3.7 Discussion and conclusions
In this report we have examined a number of aspects of relevance in relation to the operational application of numerical weather products to agrometeorological decision problems.
From the point of view of the forecasts, we noted that with respect to the North Atlantic-European sector the skill of deterministic forecasts out to 10 days has significantly improved over the last decade. Moreover, the tremendous increase in computer power now allows issuing limited-area forecasts out to 2 to 3 days at a spatial resolution that comes close to the characteristic scale of many agrometeorological applications. This significantly reduces the necessity of post-processing for this particular type of forecasts. Still in relation to high-resolution forecasts from LAM simulations, we further noticed that the catalogue of products routinely available through the NMHSs or forecasting consortia now spans a much wider spectrum of model outputs than available only a few years ago.
Next we discussed the performance of ensemble prediction systems and argued that the skill of monthly and seasonal forecasts is low concerning precipitation and remains unsatisfactory beyond week 2 in general. Additionally, the spatial resolution of these systems is still far from being adequate. Post-processing, in particular spatial downscaling and temporal disaggregation are therefore essential to provide meteorological information at the scales required by agrometeorological applications. While a vast amount of research has been conducted in terms of downscaling in the framework of climate change studies, far little has been done so far to provide the users of monthly to seasonal forecasts with specific tools. Specific developments in dedicated projects or research programs should therefore be pursued with high priority.
With potential difficulties in mind, we observed that possibility to use seasonal forecasts in agrometeorological applications has nevertheless been investigated in the framework of DEMETER with respect to crop yield predictions. Cantelaube and Terres (2005), Marletto et al. (2005) and Marletto et al. (2007) found that there is indeed a significant potential for the use of ensemble seasonal forecasts for instance for yield at the national level, despite the fact that skill and reliability of the forecasts significantly vary across Europe.
Results from these and more systematic investigations by Lawless and Semenov (2005) appear to indicate that benefits from using ensemble seasonal predictions over climatology exist out to 60 days, which represent a significant extension of lead-time of useful predictions as compared to the weather forecasts themselves. Yet, variation in lead-time of predictions for selected crop characteristics between locations and between lead-times among different crop characteristics at a single location are large (Fig. 3.7), and conclusions drawn from case studies can not be generalized.
In reviewing the literature in relation to the application of seasonal forecasts, we faced contrasting attitudes toward the significance and benefits of downscaling. While the findings of Pavan et al. (2005) seem to indicate that downscaling has beneficial effects, Semenov and Doblas-Reyes (2007) argued that downscaling of ensemble weather forecasts does not necessarily improve yield predictions, in spite of being effective in removing systematic errors. From this point of view, we again come to the conclusion that efforts are needed to advance the methodologies for post-processing and to systematically validate agrometeorological forecasts at different lead-times.
[pic] [pic]
Figure 3.7. Left: Cumulative distribution functions of lead-time for Sirius predictions estimated over 30 years of observed weather data (1960–1989) at Rothamsted, UK {Predicted variables shown are: final leaf number (FinLN), anthesis date (AnthD), maturity date (MatD), final above ground biomass (Biomass) and final grain yield (Yield)}. Right: Cumulative distribution functions of lead-time for simulated yield across sites in Europe and New Zealand. Lawless and Semenov (2005).
The modest success obtained so far in traditional application of seasonal forecasts does not preclude the possibility that other approaches may be more promising. Statistical predictions, in particular, appear to have a potential that currently is probably not fully appreciated. A successful realization of seasonal forecasting based on a statistical model was discussed in Section 6, where we reported that in the U.K. wheat grain quality at harvest can reliably be forecasted months in advance from knowledge of the NAO phase during the preceding winter. Assuming that one day operational predictions of the NAO become reality, this would pave the way for extending agronomic/agrometeorological predictions beyond the season.
In the same spirit, it is to hope that possibilities to exploit knowledge of the ENSO and/or ENSO forecasts will be explored more systematically in the near future. It is true that we are not yet in the position to provide a complete theory explaining the influence of ENSO on the atmospheric circulation in the Atlantic/European sector (Brönnimann, 2007), and this may lead to the premature conclusion that current capabilities in forecasting ENSO with coupled atmospheric-oceanic models have little significance for practical applications. But, again, statistical approaches may provide alternatives that should be more closely investigated in the coming years. First steps in this direction were already taken more than a decade ago by Fraedrich (1994) and Stone et al. (1996).
3.8 Outlook
Efforts should be undertaken in the coming years to foster the development of monthly to seasonal forecasts suitable for agricultural decision-making problems. In many regions of the world, notably the extratropics, a better understanding of the climate system as it affects agriculture is needed (McCrea et al., 2005; McIntosh et al., 2007). This calls for targeted research programs supported by universities, weather services and agrometeorological agencies (e.g. Fraisse et al., 2006).
Advancing the frontiers of knowledge should not be the only criterion for future research. As a result of a workshop on “Climate Change/Variability and Medium- to Long-Range Predictions for Agriculture” organized in 2005 by the WMO and the Queensland (Australia) Department of Primary Industries and Fisheries, Garbrecht et al. (2005) concluded that “there appears to be an inappropriate focus in the science community on developing complex, integrated forecast impact and decision support software intended for end-user operation rather than on providing choices, options and recommendations for problems that require solving”.
Garbrecht et al. (2005) also noted that "the major issue underlying the lack of successful forecast applications … is that climate forecasts and impact prediction systems offered by the science community often do not align with practical and application-specific decision needs of the intended user community". They further remarked that "neither farmers nor policy makers have easy access to relevant decision information beyond that offered by general climate forecasts, mainly because application-specific impact predictions on user-relevant decision variables are not available or cannot readily be inferred from the climate forecasts".
Promoting new products that make explicit use of seasonal forecasts could be a goal for the activities of the World Climate Applications and Services Programme (WCASP) and the Climate Information and Prediction Services (CLIPS)[38]. As stated by the WMO, CLIPS “deals with the implementation of climate services around the globe. It exists to take advantage of current data bases, increasing climate knowledge and improving prediction capabilities to limit the negative impacts of climate variability and to enhance planning activities based on the developing capacity of climate science”. Experiences from the first years of CLIPS have suggested that the provision of effective training within National Meteorological and Hydrological Services is the key issue to improve climate information and prediction services as well as to benefit appropriately from related developments.
A key activity promoted by CLIPS is provision of consensus-based climate outlook products, particularly through support to the Regional Climate Outlook Fora (RCOF). The RCOF have already become established, very useful mechanisms for the CLIPS community to coordinate national efforts in the development and application of CLIPS. There have been more than 50 RCOFs established all over the world since 1997, mainly in Africa, Asia and Latin America. Some RCOF products are available through CLIPS web page. Yet products and outlooks for Europe are not available through this mechanism.
3.9 References
Alexandrov V (2002) Summarizing crop growth simulation models in Europe with potential for operational assessment of crop status and yield prognosis. In: Dunkel Z, Alexandrov V, Braslavska O, Friesland H, Gat Z, Guerreiro R, Kleschenko A, Özalp Y, Report of the RA VI Working Group on Agricultural Meteorology, World Meteorological Organisation, CAgM Report No. 8, WMO/TD No. 1113, Geneva.
Atkinson M.D., Kettlewell P.S., Hollins P.D., Stephenson D.B. and Hardwick N.V. (2005) Summer climate mediates UK wheat quality response to North Atlantic Oscillation. Agric. Forest Meteor. 130, 27–37.
Atzema, A.J. (1992) A model for the drying of grass with realtime weather data. Journal of Agricultural Engineering Research 53, 231–247
Atzema, A.J. (1993a) A model for the prediction of the moisture content of cereals at harvesting time with realtime weather data. Journal of Agricultural Engineering Research 54, 245–256.
Atzema, A.J. (1993b) Moisture content of cereals at harvesting time by comparing microclimate values and standard weather data. Netherlands Journal of Agricultural Science 41, 167–178
Atzema, A.J. (1993c) The effect of the weather on the drying rate of cut diploid and tetraploid perennial ryegrass (Lolium perenne L.) and diploid hybrid ryegrass (Lolium perenne × L. multiflorum). Grass and Forage Science 48(4), 362–368
Atzema, A.J. (1995) Determination of the quality of forecast moisture content of cut grass using forecast weather elements. Journal of Agricultural Engineering Research 61, 1–9
Atzema, A.J. (1998) Determination of the Quality of Forecast Moisture Content of Cereals at Harvesting Time using Forecast Weather Elements. Journal of Agricultural Engineering Research 71(2), 119–125(7)
Baldwin M.P., Stephenson D.B., Thompson D.W.J., Dunkerton T.J., Charlton A.J., and O’Neill A. (2003) Stratospheric memory and skill of extended range weather forecasts. Science 301, 636–640.
Blenckner T. and Hillebrand H. (2002) North Atlantic Oscillation signatures in aquatic and terrestrial ecosystems a meta analysis. Glob. Change Biol. 8, 203–212.
Bojariu R. and Gimeno L. (2003) Predicatbility and numerical modelling of the North Atlantic Oscillation. Earth-Science Review 63: 145–168.
Bolius D., Calanca P., Weigel A.P. and Liniger M.A. (2009) Does a frequency correction of ensemble precipitation forecasts enhance forecast skill? Meteorol. Appl., in prep.
Boyce D.S. (1965) Grain moisture and temperature changes with position and time during through drying. J. Agric. Engng. Res. 10, 333–341.
Briggs W.M. and Wilks D.S. (1996) Extension of the climate prediction center long-lead temperature and precipitation outlooks to general weather statistics. J. Climate 9, 3496–3504.
Brönnimann S. (2007) Impact of El Niño–Southern Oscillation on European Climate. Rev. Geophys. 45, RG3003, doi:10.1029/2006RG000199.
Buizza R., Hollingsworth A., Lalaurette F., Ghelli A. (1999) Probabilistic Predictions of Precipitation Using the ECMWF Ensemble Prediction System. Weather and Forecasting 14, 168–189
Busuioc A., Tomozeiu R. and Cacciamani C. (2008) Statistical downscaling model based on canonical correlation analysis for winter extreme precipitation events in the Emilia-Romagna region. Int. J. Climatol. 28, 449–464.
Cassou C., Terray L. and Phillips A.S. (2005) Tropical Atlantic Influence on European heat waves. J. Climate 18, 2805–2811.
Crampin D.J. and Dalton G.E. (1971) The determination of the moisture content of standing grain from weather records. J. Agric. Engng. Res. 16, 88–91.
Das H.P., Doblas–Reyes F.J., Garcia A., Hansen J., Mariani L., Nain A., Ramesh K., Rathore L.S. and Venkataraman R. (2008) Weather and Climate Forecasts for Agriculture. Chapter 4 of the Guide to Agricultural Meteorological Practices (GAMP), Draft 3rd Edition (WMO–No.134).
Dailey .G., Smith J.U. and Whitmore A.P. (2006) How far might medium-term weather forecasts improve nitrogen fertiliser use and benefit arable farming in the England and Wales? Agric. Ecosystems Environm. 117, 22–28
Doblas–Reyes F.J., Déqué M. and Piedelievre J.–P. (2000) Multi–model spread and probabilistic seasonal forecasts in PROVOST. Q. J. Roy. Meteorol. Soc. 126, 2069–2087.
Doblas–Reyes F.J., Hagedorn R. and Palmer T.N. (2005) The rational behind the success of multi–model ensembles in seasonal forecasting – II. Calibration and combination. Tellus 57A, 234–252.
Doblas–Reyes F.J., Hagedorn R. and Palmer T.N. (2006) Developments in dynamical seasonal forecasting relevant to agricultural management. Clim. Res. 33: 19–26.
Dutton J.A. (2002) Opportunities and priorities in a new era for weather and climate services. Bull. Am. Meteorol. Soc. 9, 1303–1311.
Dyer J.A. and Brown D.M. (1977) A climatic simulator for field–hay drying. Agric. Meteorol. 18, 37–48.
Dyer J.A. and Baier W. (1981) The use of weather forecasts to improve hay–making reliability. Agric. Meteorol. 25, 27–34.
Enke W., Spekat A. (1997) Downscaling climate model outputs into local and regional weather elements by classification and regression. Climate Research 8, 195–207.
Feddersen H. and Andersen U. (2005) A method for statistical downscaling of seasonal ensemble predictions. Tellus 57A, 398–408.
Feddersen H., Navarra A., Ward M.N. (1999) Reduction of model systematic error by statistical correction for dynamical and seasonal predictions. J. Climate 14, 1974–1989.
Forrer H.R., Gujer H.U. and Fried P.M. (1993) PhytoPRE – A comprehensive information and decision support system for late blight in potatoes. Workshop on Computer–Based DSS on Crop Protection, Parma, Italy. Contribution published in SR–Report, Danish Inst. Plant and Soil Sci. 7, 173–181.
Fraedrich K. (1994) An ENSO impact on Europe? Tellus 46A, 541–552.
Fraisse C.W., Breuer N.E., Zierden D., Bellow J.G., Paz J., Cabrera V.E., Garcia y Garcia A., Ingram K.T., Hatch U., Hoogenboom G., Jones J.W. and O’Brien J.J. (2006) AgClimate: A climate forecast information system for agricultural risk management in the southeastern USA. Computers and Electronics in Agriculture 53, 13–27.
Frei C., Christensen J.H., Deque M., Jacob D., Jones R.G. and Vidale P.L. (2003) Daily precipitation statistics in regional climate models: evaluation and intercomparison for the European Alps. Journal of Geophysical Research 108(D3), 4124, doi:10.1029/2002JD002287
Garbrecht J., Meinke H., Sivakumar M.V.K., Motha R.P., and Salinger M.J. (2005) Seasonal Climate Forecasts amd Adoption by Agricultura. EOS, Transactions of the American Geophysical Union, 86, 227
Ghielmi L. and Eccel E. (2006) Descriptive models and artificial neural networks for spring frost prediction in an agricultural mountain area. Computers and Electronics in Agriculture 54, 101–114.
Gooding M.J. and Davies W.P. (1997) Wheat Production and Utilization: Systems, Quality and the Environment. CAB International, Wallingford, UK.
Gupta M.L., MacMillam R.H., McMahon T.A. and Bennett D.W. (1989) A simulation model to predict the drying time for pasture hay. Grass Forage Science 44, 1–10.
Hagedorn R., Doblas–Reyes F.J. and Palmer T.N. (2005) The rational behind the success of multi–model ensembles in seasonal forecasting – I. Basic concept. Tellus 57A, 219–233.
Hansen J.W., Challinor A., Ines A., Wheeler T., Moron V. (2006) Translating climate forecasts into agricultural terms: advances and challenges. Clim Res. 33, 27–41.
Harrison M., Kanga A., Magrin G.O., Hugo G., Tarakidzwa I., Mullen C. and Meinke H. (2007) Use of seasonal forecasts and climate prediction in operational agriculture. CAgM Report No. 102, WMO/TD No. 1344, 91 pp.
Hayward C.F. (1987) The effect of date of harvest on the grain quality of winter wheat at Terrington E.H.F. 1982–85. Aspects of Applied Biology 15, Cereal Quality, 181–187.
Hoerling M.P., Hurrell J.W. and Xu T. (2001) Tropical origins for recent North Atlantic climate change. Science 292, 90–92.
Hundecha Y. and Bárdossy A. (2008) Statistical downscaling of extremes of daily precipitation and temperature and construction of their future scenarios. Int. J. Climatol. 28, 589–610.
Hurrell J.W. (1995) Decadal trends in the North Atlantic Oscillation regional temperatures and precipitation. Science 269, 676–679.
Huth R. (2002) Statistical downscaling of daily temperature in Central Europe. J. Climate 15, 1731–1742.
Huth R. (2005) Downscaling of humidity variables: a search for suitable predictors and predictands. Daily Temperature in Central Europe. Int. J. Climatol. 25, 243–250.
Huth R., Kliegrová S. and Metelká L. (2008) Non-linearity in statistical downscaling: does it bring an improvement for daily temperature in Europe? Int. J. Climatol. 28, 465–477.
Ines A.V.M. and Hansen J.W. (2006) Bias correction of daily GCM rainfall for crop simulation studies. Agric. Forest Meteorol. 138, 44–53.
Kaas E. and Frich P. 1995. Diurnal temperature range and cloud cover in the Nordic countries: observed trends and estimates for the future. Atmospheric Research 37, 211–228.
Kalnay E. (2003) Atmospheric Modeling, Data Assimilation and Predictability. Cambridge University Press, Cambridge, 341 pp.
Karl T.R., Wang W.C., Schlesinger M.E., Knight R.W. and Portman D. (1990) A method of relating general circulation model simulated climate to the observed local climate. Part I: seasonal statistics. Journal of Climate 3, 1053–1079.
Katz R.W., Murphy A.H. and Winkler R.L. (1982) Assessing the value of frost forecasts to orchardists: a dynamic decision-making approach. J. Appl. Meteorol. 21, 518–531.
Keesman K.J., Peters D. and Lukasse L.J.S. (2003) Optimal climate control of a storage facility using local weather forecasts. Control Engineering Practice 11, 505–516.
Kettlewell P.S., Sothern R.B. and Koukkari W.L. (1999) U.K. wheat quality and economic value are dependent on the North Atlantic Oscillation. J. Cereal Sci. 29, 205–209.
Kettlewell P.S., Stephenson D.B., Atkinson M.D. and Hollins P.D. (2003) Summer rainfall and wheat grain quality: relationships with the North Atlantic Oscillation. Weather 58, 155–163
Kryvobok O. (2007) Using of meteosat HRVIS channel for the improvement of the detection of rapid developing thunderstorm clouds. 4th European Conference on Severe Storms, 10–14 Sept 2007, Trieste, Italy.
Lawless C. and Semenov M.A. (2005) Assessing lead-time for predicting wheat growth using a crop simulation model. Agric Forest Meteorology 135, 302–313.
Ligda, M.G., 1953. The horizontal motion of small precipitation areas as observed by radar. MIT Tech. Report 21, Dept. of Metereology, Cambridge, MA, p. 60.
Lynch P. (2006) The Emergence of Numerical Weather Prediction: Richardson's Dream. Cambridge University Press, Cambridge, 279 pp.
Madden R.A. and Julian P.R. (1971) Detection of a 40–50 day oscillation in the zonal wind in the tropical Pacific. J. Atmos. Sci. 28, 702–708.
Marletto V., Zinoni F., Criscuolo L., Fontana G., Marchesi S., Morgillo A., Van Soetendael M.R.M., Ceotto E. and Andersen U. (2005) Evaluation of downscaled DEMETER multi-model ensemble seasonal hindcasts in a northern Italy location by means of a model of wheat growth and soil water balance. Tellus Ser A 57, 488–497.
Marletto V., Ventura F., Fontana G. and Tomei F. (2007) Wheat growth simulation and yield prediction with seasonal forecasts and a numerical model. Agric. For. Meteorol. 147, 71–79.
McCrea R., Dalgleish L. and Coventry W. (2005) Encouraging the use of seasonal climate forecasts by farmers. Int.. J. Climatol. 25, 1127–1137.
McIntosh P.C., Pook M.J., Risbey J.S., Lisson S.N., Rebbeck M. (2007) Season climate forecasts for agriculture: Towards better understanding and value. Field Cop Res. 104, 130–138.
Meinke H. and Stone R.C. (2005) Seasonal and inter–annual climate forecasting: the new tool for increasing preparedness to climate variability and change in agricultural planning and operations. Climatic Change 70, 221–253.
Misra V., Dirmeyer P.A. and Kirtman B.P. (2003) Dynamic downscaling of seasonal simulations over South America. J. Climate 16, 103–117
Montserrat L.M. (2001) Meteorology data available on the internet for agrometeorology and general data application. Report COST Action 718, 77 pp. Available under: –cost.bo.r.it/fileadmin/cost718/repository/montserrat.pdf, accessed 14/11/2008
Monteith J.L. (1965) Evaporation and environment. Proceedings Symposium of the Society for Experimental Biology 19, 205–234.
Mullen S.L. and Buizza R. (2002) The Impact of Horizontal Resolution and Ensemble Size on Probabilistic Forecasts of Precipitation by the ECMWF Ensemble Prediction System. Weather and Forecasting 17, 173–191.
Murphy A.H. (1988) Skill scores based on the mean square error and their relationships to the correlation coefficient. Mon. Wea. Rev. 116, 2417–2424.
Obasi G.O.P. (2000) Preface. Agric. For. Meteor. 103, 3–4
Ottersen G., Planque B., Belgrano A., Post E., Reid P.C. and Stenseth N.C. (2001) Ecological effects of the North Atlantic Oscillation. Oecologia 128, 1–14.
Palmer T.N., Alessandri A., Andersen U., Cantelaube P. and 20others (2004) Development of a European multi–model ensemble system for seasonal to inter–annual prediction. Bull Am Meteorol Soc 85, 853–872
Pavan V., Marchesi S., Morgillo A., Cacciamani C and Doblas–Reyes F.J. (2005) Downscaling of DEMETER winter seasonal hindcasts over Northern Italy. Tellus 57A, 424–434
Philips P.R. and O’Callaghan J.R. (1974) Cereal harvesting – a mathematical model. J. Agric. Engng. Res. 19, 415–433.
Racsko P., Szeidl L. and Semenov M.A. (1991) A serial approach to local stochastic weather models. Ecol. Modelling 57, 27–41.
Richardson C.W. (1981) Stochastic simulation of daily precipitation, temperature, and solar radiation. Wat. Resour. Res. 17, 182–190
Richardson C.W., Wright D.A. (1984) WGEN: A model for generating daily weather variables. U.S. Department of Agriculture, Agricultural Research Service, ARS-8. USDA, Washington, DC
Rodwell M.J., Rowell D.P. and Folland C.K. (1999) Oceanic forcing of the wintertime North Atlantic Oscillation and European climate. Nature 398, 320–323.
Rodwell M.J. and Folland C.K. (2002) Atlantic air–sea interaction and seasonal predictability. Q. J. R. Meteorol. Soc. 128, 1413–1443.
Rodwell M.J. and Doblas–Reyes F.J. (2006) Medium–Range, monthly, and seasonal prediction for Europe and the use of forecast information. J. Clim. 19, 6025–6046.
Rogers J.C. (1997) North Atlantic storm track variability and its association to the North Atlantic Oscillation and climate variability of Northern Europe. J. Climate 10, 1635–1647.
Rosa M., Genesio R., Gozzini B., Maracchi G. and Orlandini S. (1993). PLASMO: a computer program for grapevine downy mildew development forecasting. Computers and Electronics in Agriculture 9, 205–215.
Samietz J., Graf B., Höhn H., Schaub L. and Höpli H.U. (2007) Phenology modelling of major insect pests in fruit orchards from biological basics to decision support: the forecasting tool SOPRA. OEPP/EPPO Bulletin 37, 255–260.
Scherrer S.C., Croci-Maspoli M., Schwierz C. and Appenzeller C. (2006) Two-dimensional indices of atmospheric blocking and their statistical relationship with winter climate patterns in the Euro-Atlantic region. Int. J. Climatol. 26, 233–249.
Schmidli J., Frei C. and Vidale P.-L. (2006) Downscaling from GCM precipitation: a benchmark for dynamical and statistical downscaling methods. Int. J. Climatol. 26, 679–689.
Semenov M.A. (2008) Simulation of weather extreme events by a stochastic weather generator. Clim. Res. 35, 203–212
Semenov M.A. and Doblas-Reyes F.J. (2007) Utility of dynamic seasonal weather forecasts in predicting crop yield. Clim. Res. 34, 71–81.
Semenov M.A., Brooks R.J., Barrow E.M. and Richardson C.W. (1998) Comparison of the WGEN and LARS-WG stochastic weather generators for diverse climates. Clim. Res. 10, 95–107.
Semenov M.A. and Brooks R.J. (1999) Spatial interpolation of the LARS-WG stochastic weather generator in Great Britain. Clim. Res. 11, 137–148.
Sivakumar M.V.K., Stigter C.J. and Rijks D. (2000) Foreword. Agric. For. Meteor. 103, 1–2
Smith E.A., Bailey P.H., and Ingram G.W. (1981) Prediction of the field moisture content of mature barley and wheat by commonly used drying equations. J. Agric. Engng. Res. 26, 171–178.
Smith G.P. and Gooding M.J. (1999) Models of wheat grain quality considering climate, cultivar and nitrogen effects. Agric. Forest Meteor. 94, 159-170
Smith J.U., Bradbury N.J. and Addiscott T.M. (1996) SUNDIAL: a PC–based system for simulating nitrogen dynamics in arable land. Agron. J. 88, 38–43.
Smith K.T. and Austin G.L (2000) Nowcasting precipitation–a proposal for a way forward. J. Hydrology 239, 34–45
Steenblock T., Forrer H.R. and Fried P.M. (2002) phytopre.ch: the internet based decision support system (DSS) PhytoPRE+2000 to control late blight on potatoes in Switzerland. In: Proceedings of the sixth Workshop of a European network for development of an integrated control strategy of potato late blight, C.E. Westerdijk and H. Schepers (eds), Edinburgh, Scotland, 26–30. September 2001, PPO–Special Report No. 8, 183–192.
Stewart T.R., Katz R.W. and Murphy A.H. (1984) Value of weather information: a descriptive study of the fruit-frost problem. Bull. Am. Meteorol. Soc. 65, 126–137.
Stigter C.J., Sivakumar M.V.K. and Rijks D.A. (2000) Agrometeorology in the 21st century: workshop summary and recommendations on needs and perspectives. Agric. For. Meteor. 103, 209–227.
Stone R.C., Hammer G.L. and Marcussen T. (1996) Prediction of global rainfall probabilities using phases of the Southern Oscillation Index. Nature 384, 252–255.
Supit I., Hooijer A.A: and van Diepen C.A. (1994) System Description of the WOFOST 6.0 Crop Simulation Model implemented in CGMs. European Commission, Joint Research Centre, Luxembourg, 144 pp.
Sutton R.T. and Hodson D.L.R. (2005) Atlantic Ocean forcing of North American and European summer climate. Science 309, 115–118.
van Elderen E. and van Hoven S.P.J.H. (1973) Moisture content of wheat in the harvesting period. J. Agric. Engng. Res. 18, 71–93.
van Loon H. and Rogers J. (1978) The seesaw in winter temperatures between Greenland and Northern Europe, Part I: general description. Monthly Weather Review 106, 296–310.
Vitart F., Woolnough S., Balmaseda M.A. and Tompkins A.M. (2007) Monthly forecast of the Madden-Julian Oscillation using a coupled GCM. Mon. Wea.Rev. 135, 2700–2715.
Voigt V. (1955) Der Kornfeuchtigkeitsverlauf auf dem Halm stehenden Getreides unter dem Einfluss der Witterung und Folgerungen für den Mähdrusch. PhD Thesis, Hohenheim, 60 pp.
WCRP (2008) WCRP position paper on seasonal prediction. Report from the First WCRP Seasonal Prediction Workshop (Barcelona, Spain, 4-7 June 2007). ICPO publication no. 127.
Weichert A, and Bürger G. (1998) Linear versus nonlinear techniques in downscaling. Climate Research 10: 83–93.
Weigel A.P., Baggenstos D., Liniger M.A., V. F., Appenzeller C. (2008) Probabilistic verification of monthly temperature forecasts. Mon. Wea. Rev., in press.
Wilby R.L. and Wigley T.M.L. (1997) Downscaling general circulation model output: a review of methods and limitations. Progress in Physical Geography 21, 530–548.
Wilks D.S. (1997) Forecast value: prescriptive decision studies. In: Katz R.W. and Murphy A.H. (Eds.) Economic Value of Weather and Climate Forecasts. Cambridge University Press, Cambridge, 222 pp.
Wilks D.S. (1998) Multisite generalization of a daily stochastic precipitation generation model. J. Hydrol. 210, 178–191.
Wilks D.S. (1999a) Simultaneous stochastic simulation of daily precipitation, temperature and solar radiation at multiple sites in complex terrain. Agric. Forest Meteorol. 96, 85–101.
Wilks D.S. (1999b) Interannual variability and extreme-value characteristics of several stochastic daily precipitation models. Agric. Forest Meteorol. 93, 153–169.
Wilks D.S. (2002) Realizations of daily weather in forecast seasonal climate. J. Hydrometeorol. 3, 195–207.
Wilks D.S. (2008) High-resolution spatial interpolation of weather generator parameters using local weighted regressions. Agric. Forest Meteorol. 148, 111–120.
-----------------------
[1] Source:
[2] Source:
[3] Sources: IM, Meteorological Institute of Portugal and
INE – Instituto Nacional de Estatística “National Statistic Institute” of Portugal
[4] Source:
[5] Sources: INE – Instituto Nacional de Estatística, “National Statistic Institute” of Portugal
and IM, Meteorological Institute of Portugal
[6] Source:
[7] sources:
World Vision International
[8] , accessed 17/09/2008
[9] , accessed 17/09/2008
[10] , accessed 17/09/2008
[11] , accessed 17/09/2008
[12] , accessed 17/09/2008
[13] , accessed 17/09/2008
[14] , accessed 27/10/2008
[15] accessed 17/09/2008
[16] accessed 27/10/2008
[17] , accessed 27/10/2008
[18] , accessed 27/10/2008
[19] , accessed 27/10/2008
[20] , accessed 27/10/2008
[21] , accessed 27/10/2008
[22] , accessed 17/09/2008
[23] , accessed 17/09/2008
[24] , accessed 27/10/2008
[25] , accessed 17/09/2008
[26] , accessed 02/11/2008
[27] , accessed 02/11/2008
[28] , accessed 02/11/2008
[29] , accessed 27/10/2008
[30] , accessed 17/09/2008
[31] , accessed 17/09/2008
[32] , accessed 10/11/2008
[33] , accessed 27/10/2008
[34] , accessed 02/11/2008
[35] , accessed 27/10/2008
[36] , accessed 14/10/2008.
[37] , accessed 10/11/2008.
[38] , accessed 14/10/2008
................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.