REMOTE
SENSING BASED CROP MONITORING
IN HUNGARY
|
REMOTE
SENSING BASED CROP MONITORING |
|
G. Csornai, Cs. Wirnhardt, Zs. Suba, P. Somogyi, G. Nádor, dr.
L. Martinovich,
L. Tikász, A. Kocsis, B. Tarcsai, Gy. Zelei
Phone: 36-1-252-7898, fax: 36-1-252-8282
E-mail: gabor.csornai@rsc.fomi.hu
Commission VII., Working Group 2.
ABSTRACT
The Hungarian Agricultural Remote Sensing Program and primarily its final R+D segment, the National Crop Monitoring Project (1993-96) led to a concise methodology that could further be applied operationally. First the pre-operational substantial validation results are treated in the paper. The validation was retrospective; it covered a 6 county area of the total 19 in Hungary and also diverse weather conditions in a 5-year period (within 1991-96). Both the area assessment, processing Landsat and IRS-1C data and the novel Landsat/IRS + NOAA AVHRR based crop yield forecast methodology performed well for the major crops (8) at county level. The second part deals with the overall evaluation of the first operational National Crop Monitoring Project in Hungary (1997). A novel method that combines land use information with NOAA AVHRR time series for yield prediction is also introduced.
1. BACKGROUND
Up to 1990 the crop information collection was based on the obligatory reports of some 1400 huge farms. Among those conditions the system worked fairly well. The economical and structural changes took place in Hungary, the former crop information system became gradually inadequate. The land privatisation brought dramatic changes in the holdings and parcel sizes, the number of farm owners or operators, the agricultural technology and investments. In this remarkable transition period, the need for an efficient information system became even more imperative.
The priority Hungarian Agricultural Remote Sensing Program (HARSP) was launched in 1980. Since 1980, many consecutive projects have been accomplished by FÖMI Remote Sensing Centre (FÖMI RSC) that implemented HARSP. The final objective of the program was to introduce remote sensing to the operational information system of the agriculture in Hungary. The operational system should be capable to monitor crops in the entire country, providing accurate, timely and reliable information on the area of the major crops, their development and problems (focusing to drought assessment), plus providing reliable yield forecast and final yield estimates. These data should be available at the country as well as the counties (19) levels. The main users of the information will include, primarily the Ministry of Agriculture, the grain processing and trading companies and associations, the farmers and their different organisations, associations. Beyond the technical, scientific problems to be solved, there is still a lot to do in the regulations and organisation of the system operation.
2. THE TWO MAIN PERIODS OF HARSP (HUNGARIAN AGRICULTURAL REMOTE SENSING PROGRAM)
The main results of the 1980-96 R+D program can be divided to two major periods:
At the earliest stage the necessary image processing and analysis system had to be developed in house in a rather isolated way. There were validation studies for 1-3 countries, up to 17000 km2 area. These preliminary results were good for the major crop assessment.
2.1. HARSP’93-96: toward a remote sensing based crop monitoring system
In the second period of HARSP a substantial, new R+D project (National Crop Monitoring Project, NCMP, 1993-96) was carried out with the objective of the improvement and stabilisation of both area estimation and crop development monitoring, yield forecast models to a stage that can be used routinely and later operationally for the whole country.
The main results of the methods validation in NCMP can be grouped as follows.
2.2. Crop survey, area assessment and their pre-operational validation
The method that had been developed by FÖMI RSC, used Landsat data and applied digital image analysis for the crop identification and area estimation (Csornai et al., 1983). This approach gradually expanded to 3 counties areas by 1990 (Csornai et. al., 1990). It was found that the provision of really more accurate county level data than those, that had been provided by the traditional non-remote sensing systems in Hungary, was only viable through advanced digital image analysis based crop area assessment. This approach also provides reliable crop maps, which are necessary to the crop development monitoring models.
As a result of the major final validation survey in NCMP (1993-96) it was clearly found (Csornai et. al., 1997) that the application and results of digital image analysis compares well with the data of the Central Statistical Office, Hungary (CSOH) for a 5 years, 6 counties data set (Figs.1.a.b.). The strong relationship in the Landsat TM derived (FÖMI RSC) and CSOH data for the major crops is promising to the further applications of satellite data in the inventories.
2.3. Crop monitoring and yield estimation
The most promising results of the NCMP are those related to the crop monitoring and yield forecast models. The models were developed by FÖMI RSC. They integrated NOAA AVHRR and Landsat or other high-resolution satellite data. This approach essentially combines the benefits of both data sources: the temporal resolution through NOAA AVHRR and spatial resolution by Landsat TM or other high resolution images (e.g. IRS-1C, SPOT). This approach requires fairly good classification for the performance with the high-resolution images. With the adaptation of a linear unmixing model (Puyou Lascassies et. al., 1994) to NOAA AVHRR series and Landsat TM, fairly good results were achieved for the two major crops -wheat and maize- for the same study area and period. The first results concerning the drought indication within the monitoring are good. The county wheat and maize yields predicted by the model compared favourably to the official data (Figs.2.a.b.). It was also found that the timeliness requirement can be met by the yield forecast model.
Both the crops areas and the major crops development and yields were estimated by remote sensing methods. This validation provided a firm basis for the first operative crop monitoring campaign in 1997.
3. OPERATIONAL CROP AREA ASSESSMENT
AND YIELD FORECAST IN 1997
The thorough previous validation created a firm basis to move forward an operational campaign in 1997. The crop data-reporting calendar was set by the customer, the Ministry of Agriculture .
It consisted of five dates from June 30 to October 1. The area covered directly was a characteristic subsample (6) of all the counties (19), so that 40 % of the total cropland in Hungary was directly monitored. Beyond the counties level crop area and predicted yield data these had to be expanded to the entire area of Hungary. This expansion used a subregional temporal correlation analysis plus a direct robust method (see 4.). The eight main crops monitored were winter wheat, winter and spring barley, maize, sugar beet, sunflower, alfalfa and maize to ensilage. These crops together represent the 78-82 % of the entire Hungarian cropland.
The crops area assessment was based on the multitemporal image analysis of Landsat TM and IRS-1C LISS III. data from the early May-August period, to compensate for the cloudiness in 1997. Cloud cover was some 30 % bigger than the average in the 1991-96 period. The comparison of the remote sensing results with CSOH data is obviously an indication only and the differences cannot, by any means be interpreted as the errors of the remote sensing technology. A thorough study is under way that will produce confidence values attached to the area estimates. The difference of crop areas estimates of FÖMI RSC and the Central Statistical Office, Hungary (CSOH) ranged in the 0.8-3.7 % (Fig.5.a.) for the entire cropland in Hungary. The county crop area differences occurred in the interval of 1.5-21 % depending on the crop and county. However the area weighted average difference was 4.08 %.
This partially can be explained by the main differences in definitions, that is the ownership based sampling of CSOH and the administrative, topographic boundary based total coverage of cropland by the satellite images (FÖMI RSC). The actual standard crop maps derived were also provided to MoA (Fig.3.).
The crop yield forecast was accomplished by the application of FÖMI RSC developed model which combines high-resolution satellite (Landsat TM and IRS-1C LISS III.) data and NOAA AVHRR time series. The reporting dates corresponded to those of the operative Production Forecast System of the Ministry of Agriculture . Both appeared prior to the beginning of harvest. The final official data are available after the harvest: by the end of August for wheat and barley and in December (January) for the rest. FÖMI RSC provided yield estimates for the counties and expanded to Hungary. The yield data compared favourably with CSOH values, appeared six weeks later (Fig.5.b.). The differences were less than 1 % for wheat and 4.5 % for maize average yields in Hungary. The differences at county level averages are certainly bigger. Because of the method applied, yield spatial distribution maps could also be reported (Fig.4.) for the major crops.
|
Fig.1.b. The figures compare similarly to those of wheat. The reason for somewhat weaker relationship is the practice and statistics of maize for silage. Only a part of maize is sown originally for ensilage. Many times decisions are made if needed for maize ensilage along the season. New methods suggest compensation to this effect. The missing data are the same as in Fig.1.a.
|
 |
Fig.1.a. The satellite data coverage was incomplete for these years and counties (missing year: 1994). The area estimation for winter wheat shows a strong relationship between the traditional (questionnaire) method and the remote sensing one. The data appear remarkably earlier from the remote sensing system. |
 |
Fig.2.a. The satellite data coverage was incomplete for these years and counties (missing year: 1994). The wheat yields can be predicted by remote sensing prior to the harvest. These years comprise good and extreme bad ones as well. The missing data are the same as in Fig.1.a. |
 |
Fig.2.b. The maize yields can be predicted early prior to the harvest. The sample comprises diverse years. The missing data are the same as in Fig.1.a. |
Fig.3.
Crop maps for the 6 counties in Hungary derived from multitemporal
high-resolution satellite data
(Landsat TM and IRS-1C LISS III.) from the early May-August period of 1997.
Fig.4.
Winter wheat yield forecast for the 6 counties in Hungary using our
developed
Landsat/IRS + NOAA AVHRR model.
Crop areas in Hungary, 1997 by FÖMI RSC and CSOH* |
Crop yields in Hungary, 1997 by FÖMI RSC and CSOH* |
|
 |
 |
|
| *in “Production data of the
main agricultural crops, 1997”, Jan 14, 1998, Central Statistical Office, Hungary |
 |
|
Fig. 5.a.b. Crops area assessment (Fig.5.a.) and yield forecast (Fig.5.b.) in
Hungary, 1997 |
||
Winter wheat yields |
Winter barley yields |
|
 |
Maize yields |
Sunflower yields |
|
|
Fig.6.a.-d.
In the 5 years period of 1991-96 (excluding 1994) the predicted county average yields (
· ) correlated very well with the final CSOH data. The predicted average yield to the entire country fitted even better (D ).
4. ROBUST YIELD PREDICTION BY
NOAA AVHRR SERIES
The primary yield forecast model (see 3.) performed well. There were two reasons to develop robust yield forecast model:
The pre-processed and normalised NOAA AVHRR data set was temporally filtered. The average reflectance profile and the NDVI could be decomposed in time by a thorough spectral-temporal correlation analysis. This substantial analysis showed an extremely strong relationship between the predicted county yields by this decomposition method and the CSOH data (Figs. 6.a.-d.). The county data set comprises a 5 years period in which the low and high ends of yields occurred. The model seems to be strong, independent from the year and area. Some hilly, mountainous counties or those that were covered very sparsely by the given crop had to be omitted from the analysis. Having the performance of this model by county (r2=0,85-0,96) the country level yield prediction seems to be very reliable (r2 = 0,93-0,99). These preliminary results suggest that a reliable yield prediction model can be set up.
5. CONCLUSION
Both the validation of the developed remote sensing based crop area assessment and yield forecast methods plus the first operational monitoring and crop production forecast campaign (1997) in Hungary clearly demonstrated that these methods can be efficiently applied. Substantial background and investment is certainly needed. About 300 man/year was invested by FÖMI RSC in the framework of the Hungarian Agricultural Remote Sensing Program (1980 to date). The first operational monitoring was designed very strictly by the Ministry of Agriculture , Hungary, according to its existing operational production forecast and monitoring system.
Remote sensing could be very efficiently used for precise crop area estimation and provision of crop maps. The results suggest that the necessary classification performance can be obtained in most of the cases, therefore the analysis could be cost effective. The investment to achieve this seems to be worthwhile.
The new combined AVHRR and Landsat TM or IRS-1C LISS-III. or SPOT based crop monitoring and yield prediction models and the approach performed properly and efficiently in a more counties' area application and also for the entire country. The second, the county level AVHRR based crop yield prediction model worked very well and seems to have a real potential on areas, having quite different cropping pattern.
After the first year, further assessment and gradual extension of remote sensing into the information system of the Ministry of Agriculture is under way. Together with the gradual expansion of the direct target area from 6 counties to whole country more and even earlier reporting dates are planned. This system is supposed to operate parallel to the existing dynamic system of MoA for monitoring area and crop development, plus yields of the most important crops in Hungary.
6. ACKNOWLEDGEMENT
The whole HARSP (1980-) and in particular the recent NCMP (1993-96) have been supported jointly by the National Committee for Technological Development and the Ministry of Agriculture, Hungary. Formerly, the Hungarian Academy of Sciences, since 1992 the Hungarian Space Organisation have also given both financial and scientific support to the program. The major operational crop monitoring and production forecast program from 1997, on is being supported by the Ministry of Agriculture.
SAI, EC Joint Research Centre (Ispra) generously supported a natural vegetation monitoring study by pre-processed NOAA data for 1991-95.
REFERENCES
Csornai, G., dr. Dalia, O., Gothár, Á., dr. Vámosi, J., 1983 Classification Method and Automated Result Testing Techniques for Differentiating Crop Types, Proc. Machine Processing of Remotely Sensed Data, West Lafayette, USA
Csornai, G., dr. Dalia, O., Farkasfaly, J., dr. Vámosi, J., Nádor, G., dr. Vámosi, J., 1988. Regional Vegetation Assessment Using Landsat Data and Digital Image Analysis, Proc. 5th Symp. ISSS Working Group Remote Sensing, Budapest, pp. 123-128.
Csornai, G., dr. Dalia, O., Farkasfaly, J., Nádor, G., 1990. Crop Inventory Studies Using Landsat Data on Large Area in Hungary, Applications of Remote Sensing Agriculture, Butterworths, pp. 159-165.
Puyou Lascassies P., Podaire A., Gay M.: Extracting Crop Radiometric Responses from Simulated Low and High Spatial Resolution Satellite Data Using a Linear Mixing Model: Int. J. of Remote Sensing, Vol. 15, no. 18, pp. 3767-3784, 1994.
Büttner Gy., dr. Csató É., Maucha G.: The CORINE Land Cover-Hungary Project, GIS/LIS’95 Central Europe, Budapest, Hungary, 12-16 June, 1995.
Csornai G.: Towards a satellite based national monitoring system in Hungary, Eurisy Colloquium, Budapest, Hungary, 15-16 Mai, 1997.
| © 2002, Földmérési és
Távérzékelési Intézet 1149 Budapest, Bosnyák tér 5. Tel: 222-5101 Fax: 222-5112, Postafiók: 1592 Budapest, Pf. 585. |
Az utolsó módosítás
időpontja: 2004-11-17 webmester |
