Bognár Péter

Ph.D. School of Earth Sciences

Head of Ph.D. School: Prof. Dr. Márton Péter

Ph.D. program of Cartography

Head of the Ph.D. program: Prof . Dr. Klinghammer István

Project Leader: Prof. Dr. Ferencz Csaba

Department of Geophysics,

Eötvös University




Administering a modern society requires the continuous monitoring of areas of economy of vital importance with scientific accuracy. One such area is the area of combined natural and social resources. Among these food production has a key role, the basic part of monitoring this area is to know what the agricultural crops yield is. It is essential that authentic data about these yields could reach the administration as soon as possible. Almost all countries with notable agricultural production have developed their yield forecasting system mainly based on plot level observations and the producers' own yield reports. However, this system - hereafter called "traditional"- requires considerable human and financial resources.

The national level monitoring of Hungary's main crops is becoming less and less reliable with the methods based on producer reports. Reporting and assembling the yield data after the harvest takes a long time. With the spreading of private farms and constant changes in individual plot sizes the reliability of the data supplied has indeed decreased, and no improvement in this can be expected in the near future. This makes it important to development an independent yield forecasting system on a national level based on satellite data.

Presently there are several different satellite systems that can provide regular coverage of the Earth's surface. These can supply us regularly with remotely sensed data of the whole area of the country. With the help of these data, not only the spatial distribution of the crops can be determined and their status in the growing period can be monitored, but they also give an opportunity to develop yield-forecasting methods. To estimate the overall yield one has to guess both the cultivated area and the average yield. The spatial separation of the different types of crops based on satellite data can be considered a problem already solved. However, the problem of estimating yields is in an experimental stage everywhere in the world, and although much energy is spent on the experiments of global yield forecast development, at present we do not know of any operative methods based on satellite data that could tell the yield for every kind of areas, terrains and crops year independent down to even plot level with auxiliary terrestrial data supply or without it.

To get the adequate number of data in the growing season, a dense temporal sampling is necessary. According to recent experience a full satellite data coverage is necessary of the areas in question every 6 to 8 days. Taking into consideration the fact that certain images cannot be used because of the cloud cover, the satellite must return to the area at least every 3 to 4 days. Plot level observations would require a finer spatial resolution. A resolution of some 10meters would be adequate here. However, the two above requirements are not yet fulfilled by any satellite systems at present. Therefore, in practice, finer resolution data can only be used in the calibration phase, and service-like (operative) yield forecast methods can still only be based on the restricted spatial resolution data for a time.

In Hungary the Space Research Group of the Geophysics Department of ELTE with the backing of the Hungarian Space Office has been conducting studies concerning the development of yield forecasting methods for decades using low-resolution satellite images. The studies resulted in two methods of yield forecasting, one is called a "detailed" method and the other, which was developed based on the results of the first one is called the "robust" method. In these models the target is to establish a direct mathematical relationship between the satellite data and the yields. The presumption is that there is some kind of connection between the properties measured by the satellite over an area covered by a certain type of crop and the average yield of that crop in that area. Naturally the yield is affected by many factors such as the given variety of the crop, the cultivation technology, the type of soil, the weather and even the solar cycle, etc. It is important to stress that the study was not to decipher how these factors individually affect the yield, only to state a mathematical relationship between the properties a satellite can sense of a crop covered area and the yield of that crop.

The main target of my work is to show in detail the two methods mentioned above, to present the results obtained with their help and to prove that these methods - especially the "robust"one - can be a starting point of a subsequent operative yield forecast service.


Preprocessing of both low- and high-resolution satellite images (radiometric, atmospheric and geometric corrections) and land use classification with the high resolution data.

Producing the exclusion masks of urban areas, forests and water surfaces.

Assembling Greenness temporal series from AVHRR channels 1 and 2 for single plots and whole counties, with double Gauss curve fitting.

Corrections for geometric arrangement (sun-surface-sensor angles).

Establishing a linear relationship between the calculated indices and the crop yields by the least squares principle.


Two methods for estimation the yield of different crops in Hungary from satellite remote sensing data are presented. The steps of preprocessing the remote sensing data (for geometric, radiometric, atmospheric and cloud scattering correction) are described.

The first method developed for field level estimation. Reference crop fields (corn, wheat, sugar beet, sunflower and winter barley) were selected in three counties (Hajdú-Bihar, Békés, Jász-Nagykun-Szolnok) by using Landsat TM data for classification. A new vegetation index (GYURI) was deduced using a fitted double-Gaussian curve to the NOAA AVHRR data during the vegetation period. The correlation between GYURI and the field level yield data for corn for three years was R=0.87. The county-average yield data shown higher correlation (R=0.96). A significant distortion from the model give information of the possible stress of the field.

The second method presented uses only NOAA AVHRR and officially reported county-level yield data. An essential part of this method is the correction, which takes into account the regional effects of the surface geometry, essentially the effects of the relief. Besides, a new correction method was developed to eliminate the effect of the Sun-sensor-target geometry, which improved the accuracy of yield forecasting and estimation procedures. The county-level yield data and the deduced vegetation index, GYURRI were investigated for 10 different crops (wheat, corn, winter barley, spring barley, corn for sillage, rye, potatoes, green peas and alfalfa) for 8 years (1991-93 and 1996-2000). The developed robust method proved to be stable and accurate for operational use for county, region and country-level yield estimation. In the latter period (1996-2000) 39 of the estimated 50 annual yield values (10 plants and 5 years) were within the 5% limit of the official HSO (Hungarian Statistical Office) data and all of them were within the 10% error limit. It is a good result even on an international level and proves the practical value of the robust yield forecasting method. Furthermore, based on the robust process a new method was also successfully developed that gives forecasts with adequate accuracy for crop yields well before the harvest dates. The method is simple and inexpensive for application in developing countries, too.

The operational estimation requires satellite data in time. Therefore the installation of the university's own satellite receiver station, it has become possible to run a yield forecast service for several agricultural plant types.


1. Correction of the effect of Sun-sensor-target geometry

The radiance values received by the AVHRR sensor are affected by not only the atmospheric effects, but also the Sun-object-sensor geometry due to the non-Lambertian reflecting behaviour of the surface. To take into account of this effect it is necessary to know the bidirectional reflectance distribution function (BRDF) of the surface. The BRDF can be estimated by model calculations and directly from satellite measurements. I have developed a new, easily and operatively applicable correction method based on the satellite measurements, that makes it possible to compare the data received with different Sun-object-sensor geometry and also in accordance with the theoetical models. The greenness vegetation index between days of the year 140-200 and the angle enclosed by the Sun-target and sensor-target directions over the agricultural areas were used for the correction method. As a result the standard deviation around the best-fitted double-Gaussian curve has been significantly decreased, which has improved the accuracy of yield forecasting and estimation procedures using NOAA AVHRR data.

2. Correction method due to the zenith angle of the Sun

The orbits of the NOAA satellites were planned as Sun-synchronous ones. But the exact transit time of the NOAA satellites over Hungary are changing year by year (the satellites are "late") therefore the zenith angle of the Sun is different on the same days in different years when the satellite orbits over the given area. This affects considerably the accuracy of the yield estimation method, especially in the case of crops whose date of harvest are in the autumn (e.g. corn). The original greenness data set does not show this effect, but I have demonstrated that it appears on the data set produced by the above mentioned Sun-sensor-target correction method. To eliminate this effect I have developed a second correction method, and applying these two corrections the corrected greenness values are free of the BRDF effects. Thus the data of the long-life NOAA satellites can be used in long period for the yield estimation methods.

3. Application of the "detailed" yield estimation method for different crops

I have verified that the detailed yield estimation method developed by our group for corn and wheat, succesfully applicable for different crops (winter barley, sugar beet, sunflower) also. However, this method was not possible to use for the alfalfa, which has a vegetation cycle different from the corn or wheat (multiple reaping in the year). At the same time this recognition helped the development of a new yield estimation method.

4. Definition of crop-filters in the "robust" yield estimation method

I have demonstarted that the crop-filters (initial iteration parameters for the fitting of the double-Gaussian curve, the first and last day of the summation of the fitted function etc.) developed for the "detailed" yield estimation method, are applicable in the robust yield estimation method too. I have developed three types of crop-filters based on the vegetation cycles of different crops: the wheat-type for the crops with autumn sowing date, the corn-type for the crops with autumn harvest date, and the alfalfa-type. I have pointed out that all of the crops with large (Hungarian) acreage can be analyzed using either of these crop-filters (except the sunflower in the period of 1996-2000).

5. Developing and applying the final form of the relief correction

The yield estimation methods published in the literature have generally reference to smaller, and almoust in every case to flat areas. At present no such general yield estimation method exists which can handle the effect of hilly areas. However, I have succesfully developed the final form of the relief correction in the robust yield estimation method. In this correction method an index was assigned to every unit (e.g. county) which contains the effect of relief and stratum. Applying the relief correction, the robust yield estimation method was succesfully extended for the whole agricultural area of Hungary and opened the possibility to establish an operativ yield forecasting service for the whole country.

6. Applying the "robust" yield estimation method for 10 crops

Ten crops (wheat, corn, sugar beet, winter barley, spring barley, alfalfa, maize for silage, rye, potato and pea) were analyzed with the robust yield estimation method, using data from five years (1996-2000) period. The results of 50 data (10 crops and 5 years) show that the differences between the estimated and the official (Hungarian Statistical Office) yield data were less than 5 % in 39 cases, and never were more than 10 %. These results are very good in international level, and verify the practicability of the "robust" yield estimation method.

7. Development of a forecasting method

Using the robust yield estimation method it can be possible to estimate the probable yield roughly at the date of the harvest. I have improved and succesfully applied the method to forecast the yield of the crops before the harvest. To reach this goal I have extrapolated the measured data set and after that the original robust yield estimating method can be used. I have found that it is possible to forecast the yield accurately on the 150th day in the cases of wheat, winter barley and rye, and the 210th day in the cases of corn, sugar beet, spring barley, maize for silage, pea and potato, assuming that no disaster (fire, sudden drought, hailstorm etc.) would occur between the forecasting and harvest.


This work is part of a big theme of great economic importance that aims to monitor the state of natural and economically important properties of the vegetation cover quantitatively by satellite borne data. Based on the results described in the thesis it can be stated that with the demonstrated yield forecasting methods an adequately accurate estimation can be given in Hungary for the average yields of the crops having the greatest cultivated area.

With further improvement of yield forecasting methods and with making the atmospheric correction more accurate the conditions are given for running an operative yield forecasting service that is generally applicable (for all kinds of terrain, for many types of crops and also for natural vegetation) with such an exceptional accuracy which could be used not only in Hungary but also by the whole international community.

The earlier difficulties of data acquisition were eliminated by the installation of the HRPT receiving station in 2002. Since the images received here cover a vast area (the entire Europe, a part of North Africa and of the Middle-East) there is the potential to extend the "robust" method for other countries as well.


SCI publications:

Bognár P. (2003): "Correction of the effect of Sun-sensor-target geometry in NOAA AVHRR data", Int. J. Remote Sensing, vol. 24, no. 10, 2153-2166.

Ferencz Cs., Lichtenberger J., Bognár P., Molnár G., Steinbach P., Timár G. (2003): Mûholdvev állomás az ELTE Környezetfizikai Tanszékcsoportján. Geodézia és Kartográfia, 2003/9, 30-33.

Ferencz Cs., Bognár P., Lichtenberger J., Hamar D., Tarcsai Gy., Timár G., Molnár G., Pásztor Sz., Steinbach P., Székely B., Ferencz O., Ferencz-Árkos I. (2004): "Crop yield estimation by satellite remote sensing", Int. J. Remote Sensing, vol. 25, no. 20, 4113-4149.

Conference papers, annual reports:

Bognár P., Ferencz Cs. and Tarcsai Gy. (1995): "Correlated changes in sunspot numbers and in corn and wheat yields", Acta Geophys. et Meteor. (ELTE, Budapest), Tom. XI., 191-206.

Ferencz Cs., Bognár P., Ferencz-Árkos I., Hamar D., Lichtenberger J., Molnár G., Pásztor Sz., Steinbach P., Székely B., Tarcsai Gy., Timár G. (1996): "Crop yield forecasting in East Hungary using remotely sensed data and a new atmospheric correction method", Proc. URSI XXVth Gen. Ass., Lille 28 Aug. - 5 Sept. 1996, F4.-p.4.

Ferencz Cs., Hamar D., Bognár P., Tarcsai Gy., Pásztor Sz., Molnár G., Ferenczné Árkos I., Lichtenberger J., Steinbach P., Székely B., Timár G., E. Ferencz O., Erhardt Z. (1997): "Haszonnövények terméshozamának meghatározása mholdas adatok alapján", VII. Földfotó Szeminárium, MANT, MTESZ, p.18.

Ferencz Cs., Hamar D., Lichtenberger J., Bognár P., Ferencz-Árkos, Molnár G., Pásztor Sz., Steinbach P., Székely B., Tarcsai Gy., Timár G. (1998): "Yield forecasting of main crops using remote sensing data", ISPRS Commission VII. Symp. on "Resource and Environmental Monitoring", 796-804, Budapest, Hungary, Sept. 1-4, 1998.

Lichtenberger J., Ferencz Cs., Hamar D., Bognár P., E. Ferencz O., Molnár G., Pásztor Sz., Steinbach P., Székley B., Tarcsai Gy., Timár G. (1998): "Methods to elminate the atmospheric effects in remotely sensed data", ISPRS Commission VII. Symp. on "Resource and Environmental Monitoring", 787-796, Budapest, Hungary, Sept. 1-4, 1998.

Ferencz Cs., Bognár P., E. Ferencz O., Hamar D., Lichtenberger J., Molnár G., Pásztor Sz., Steinbach P., Timár G. (2003): HRPT/CHRPT és SAS mholdvev állomás az ELTE Környezetfizikai Tanszékcsoportján (poszter). rnap, ELTE TTK, Budapest, 2003. október 16.

Internal reports:

Timár G., Bognár P., Büttner Gy., Ferencz Cs., Pásztor Sz., Székely B., Tarcsai Gy. (1992): "A Bs (Gabikovo)-környéki létesítmények környezeti hatásainak vizsgálata mholdas távérzékeléssel", kutatási jelentés, ELTE Geofizikai Tanszék.

Ferencz Cs., Tarcsai Gy., Bognár P., Erhardt Z., E. Ferencz O., Ferenczné Árkos I., Hamar D., Lichtenberger J., Molnár G., Pásztor Sz., Pipás Á., Steinbach P., Székely B., Timár G. (1996): "NOAA AVHRR és Landsat adatokon alapuló megyei és országos haszonnövény hozammodell kialakítása", tanulmány. GISkard Kft., Budapest.

Poipular scientific articles:

Ferencz Cs., Lichtenberger J., Timár G., Molnár G., Pásztor Sz., Bognár P. (2003): "Parabola-ablak" Európára és környékére - új mûholdvevô állomás az ELTE Környezetfizikai Tanszékcsoportjánál. Technika, 46(9-10), pp. 10-11.

Hamar D., Bognár P., Ferencz Cs., Lichtenberger J., Molnár G., Pásztor Sz. (2003): A növényzet állapotának és várható termésátlagának meghatározása mholdas adatokból. Technika, 46(11-12), pp 33-35.


Sazhin S., Bognár P., Smith A.J., Tarcsai Gy. (1993): "Magnetospheric electron temperatures inferred from whistler dispersion measurements", Annales Geophysicae, Atmospheres, Hydrospheres and Space Sciences, 11, 619-623.

Tarcsai Gy., Bognár P., Ferencz Cs., Hamar D., Lichtenberger J. (1993): "VLF research in Hungary", Newsletter of the IAGA/URSI Joint Working Group on VLF/ELF Remote Sensing of the Ionosphere and Magnetosphere, 5, 4-5.

Ferencz Cs., Bognár P., Hamar D., Tarcsai Gy. (1993): "The whistler-mode propagation: rigorous solution of Maxwell's equations and the new whistler-model", Proc. URSI XXIVth Gen. Ass., Kyoto, 25 Aug. - 2 Sept. 1993., H6.p-05.

Ferencz Cs., Bognár P., Hamar D., Tarcsai Gy., Smith A.J. (1993): "Possible explanation of some observed features of measured whistlers by using the rigorous solution of Maxwell's equations", Proc. URSI XXIVth Gen. Ass., Kyoto, 25 Aug.- 2 Sept. 1993., H6.p-06.

Ferencz Cs., Ferencz O., Tarcsai Gy., Bognár P., Hamar D., Smith A.J. (1993): "A brief summary of the new whistler-model and its application in whistler-investigations", Proc. URSI XXIVth Gen. Ass., Kyoto, 25 Aug.- 2 Sept. 1993., H6.8.

Székely B., Stoupel E., Bognár P., Ferencz Cs., Tarcsai Gy. (1993): "Strong 11-year cycle in hospital mortality: predicted frequency becomes more accurate in the last two years", Proc. of Semmelweis Sci. Forum.

Ferencz O., Bognár P. (1995): "New method and its numerical application for whistler-mode propagation", in Advanced Computational Electromagnetics, ed. by T. Homma (Studies in Applied Electromagnetics and Mechanics 9), pp. 321-334, Elsevier, Amsterdam.

Bognár P., Ferencz Cs., Hamar D. and Tarcsai Gy. (1995): "Whistler mode propagation: The first results of model-calculations for a homogenous plasma", Acta Geophys. et Meteor. (ELTE, Budapest), Tom. XI., 53-64.

Ferencz O., Bognár P. (1995): "New computational methods with numerical approximation for modelling whistler-mode propagation", Acta Geophys. et Meteor. (ELTE, Budapest), Tom. XI., 1995, 89-114.

Ferencz Cs., Bognár P., Tarcsai Gy., Hamar D., Smith A.J. (1996): "Whistler-mode propagation: results of model calculations for an inhomogeneous plasma", J. Atmos. Terr. Phys., 58, 625-640.

Lichtenberger J., Bognár P., Ferencz Cs., Hamar D., Lefeuvre F., Smith A.J., Székely B., Tarcsai Gy. (1996): "Recent results on high resolution whistler analysis", Proc. URSI XXVth Gen. Ass., Lille 28 Aug. - 5 Sept. 1996, H5.3.