Investigation of Hungarian forest health condition with special respect to climate change

a Institute of Informatics and Economics, Simonyi Karoly Faculty of Engineering, Wood Sciences and Applied Arts, University of West Hungary, Sopron, 9400, Hungary podor@inf.nyme.hu b National Food Change Safety Office, Forestry Directorate, 1023 Budapest, Frankel L u. 42-44, Hungary kolozsl@nebih.gov.hu c National Food Change Safety Office, Forestry Directorate, 1023 Budapest, Frankel L u. 42-44, Hungary soltigy@nebih.gov.hu d Institute of Informatics and Economics, Simonyi Karoly Faculty of Engineering, Wood Sciences and Applied Arts, University of West Hungary, Sopron, 9400, Hungary jereb@inf.nyme.hu ABSTRACT


INTRODUCTION
Nowadays one of the new challenges is the climate change including the protection against its negative impact, to identify the needed and possible mitigation and adaptation measures. According to the predicted climate scenarios the climate will become warmer and drier at certain regions, including Hungary (Gálos et al., 2007, Bartholy et al., 2009, Pieczka et al., 2011.
The article focuses on these issues based on the available forestry healthy data previously collected under the forest monitoring activities. In Hungary the national forest monitoring program started in 1987 with the development of the Forest Protection Network. The defoliation and the crown dieback were the main attributes, that were considered by survey and they were adapted for the characterization of healthy conditions. Since that time several other components were added to the ecosystem monitoring and the scope of the collected data was significantly extended.
In the paper two data lines of 15 years are investigated, trends of them as well as relationships between these data and the fundamental climatic components (precipitation and temperature) are examined and identified.Although the 15 years period under review is relatively short compared to the scale of the climate change, we believe that the analysis of the data collected in the forest area, and the exploration of interrelationships with other environmental data sets can contribute to better understanding of this issue.

CHARACTERIZATION AND DESCRIPTION OF DATA INPUTS
The aim of the analysis was the investigation of time series collected relating to forest health (defoliation and crown dieback) and the interrelation between these data and the data associated with meteorological components in the context of forest monitoring. The data used in this research were obtained from two main sources: (i) Forest Protection Network; (ii) meteorological data. The main characteristics of the data can be summarized as follows:

Forest Protection Information Network (level I)
From the 80-ies of the last century in all European countries -including Hungary -forest health deterioration has been observed (E.D. Schulze, 1989). As a consequence of the significant strengthening of environmental factors on the forest the development of a program observing the damages as well as the regular and systematic monitoring of forest health have got special importance.
As a part of the international cooperation the Hungarian national surveillance network joints the program (International Cooperative Programme on Assessment and Monitoring of Air Pollution Effects on Forests, briefly ICP Forests). The program was setup on the third meeting of the Executive Body for the Convention on Long-Range Transboundary Air Pollution in 1985. Using uniform methods the ICP Forests program thus primarily examines the effects of air pollution on forests in European level. In longer term the aim is to identify the key factors decisively influencing the health status of forests and to explore relationships among the factors. In Hungary the development of the forest monitoring program began in 1987, the first field recordings and the evaluation of the data started in 1988, and since that time every year complete annual sightings have been performed. In our investigation the data are taken from 1990, after the stabilization of Forest Monitoring Program and the measurement technology.

Data collection methodology
In level I wide-area health records are collected based on systematic sampling. The sampling sites are established according to a theoretical 4x4 km grid covering the entire territory of the country. The sampling sites of the grid situating in forest area are referred hereinafter to as sample points. In 2004, there were more than 1,200 sample points in the country (Fig 1). These points are geo-referred by GPS coordinates.
The examined healthy data are from 1990 to 2004, so the length of the examinable time series are 15 years. There are 12 groups of examined tree species (more than 150 tree species). The dataset contains more than 335000 records, with 36437 tree individuals. O c t o b e r 1 7 , 2 0 1 4   In accordance with the ICP Forest Manual the damages in the field observation are measured with 10 % accuracy. Defoliation of the sample trees including damages caused by known and unknown decaying agents, and furthermore crown dieback are recorded. The recordings are always collected in the same period of the year.
The sample trees are classified according to height classes using Kraft classification, and only classes 1-3 are considered in the investigation.

Meteorological data
The homogenized and interpolated monthly precipitation and temperature data series are derived from the network of the Hungarian Meteorological Service (HMS). These meteorological data covered the country approximately in a 10*10 km grid. In the paper the monthly precipitation and the average temperature data of 1961-2010 are used. Similarly to the forestry data the meteorological data are also geo-referred by GPS coordinates. The average values of the corresponding meteorological sample points were used in our investigations.

APPLIED METHODOLOGY
In each year the values of the defoliation and the crown dieback were measured in the 0-100% range with accuracy of 10%. Therefore, health of each group is characterized by a 15-year long data set. The first research direction was the analysis of temporal changes in these series. To do this, for each year the original data obtained by measurement or observation are transformed to percentage of plants showing at least x% of defoliation and crown dieback, where x = 10, 20, ..., 90. The exactly 0% and 100% defoliation and crown dieback were studied separately.
Trend tests were performed in order to decide whether the studied data sets showed statistically significant tendency in the given period. A linear trend line was fitted to each time series of 15 years and a t-test with n-2 degrees of freedom and with α = 0.05 level was applied for the significance test. The slopes of the resulting significant trend lines were examined to detect whether any upward or downward tendencies in the data set can be identified.
The other main direction of the investigation was the study of correlation among the health status and the fundamental or derived meteorological parameters as it is discussed in Chapter Meteorological data. The separate and combined effects of meteorological parameters of the given, the previous and the second previous years were analyzed on the defoliation and the crown dieback. In this analysis for each meteorological parameter as independent variable the following periods were generated:  the actual year (a)  the previous and actual year (p1a)  the last but one and last previous years and the actual year (p2p1a)  the previous year (p1)  the last but one and last previous years (p2p1)  the last but one and last two previous years (p3p2p1) It is worth mentioning that in all cases when the independent variables are taken from any previous years the derivation of certain parameters (e.g. PADI) only permits the examination of shorter series.
In the studies the average annual rate of defoliation and the crown dieback were used as dependent variables and they were generated for all specimens as well as for each group of tree species. In order to identify relations among the parameters linear correlation analysis was used, where the significance of the obtained correlation coefficients (r-values) was examined by t-tests with n-2 degrees of freedom with α = 0.05 level. This also means that the significance of the rvalues is strongly affected by the length of the data series (n) (see Table 2). The above investigations were carried out both for defoliation and crown dieback in case of all specimens and for the 12 groups of tree species as well. Considering that the length of the basic time series for both the meteorological and for the forest health parameters there are exactly 15 years between 1990 and 2004 and the significant r-values with α = 0.05 are summarized in table 2.

RESULTS
The results basically show the details of averages for all specimens and only the results of the individual groups of tree species differing significantly from the averages are discussed separately. In the paper our aim was to investigate the general trends of healthy conditions and the relationships between the healthy conditions and the basic and derived meteorological parameters. Therefore these results give a general overview of the relationships and give assistance to identify further research directions. Taking into account the slope (m) of the trend lines fitted to the data series (last but one row of Table 3) it can be stated that the defoliation level is slightly declining during this period since the slope of the regression line in case of minimum 10% defoliation is m = -0.1, but it is not significant. Statistically significant trends at level 0.05 can be observed only in the range of minimum 20-40%. The largest absolute value is obtained for minimum 20% defoliation where the value is -0.53, and the appropriate functions are y = 1097.8 -0.53 x with R 2 = 0.3308. Similarly, high slope value (m = -0.45) was found in case of at least 30% defoliation. However, the downward trend in at least 60% defoliation rate turns and it shows very slightly growing but not significant trends. Finally, the 100% defoliation rate is practically stagnant in the period under review.

Results of the trend analysis Defoliation
The results described above are summarized and displayed in Figure 3 with the regression lines fitted to the data set and with the corresponding R 2 values.

Fig 3: The trend of defoliation proportion for all species
It can be generally declared that the groups of tree species A, B, CS, EKL, ELL, GYT K and NNY follow the general trend, in the examined priority range (minimum 20-40%) and the trend turning-point (minimum 60%) as well.
The groups EF, EGYF, FF, HNY show an opposite trend direction compared to the above mentioned general trend. The upward trend in the range of minimum 20-40% defoliation usually are statistically significant and the slope values are relatively high.  Based on the derived slope (m) values the rates of minimum 10-40%, of crown dieback also show significant downward trend, and then, from the 50% level the data series are practically stable (not significant trends at level 0.05) as it is shown in Table 3. Moreover, these regression lines are statistically significant on the basis of the obtained R 2 value, and this is in nature consistent with those experienced in the context of defoliation. These results are characteristically similar to those obtained by defoliation.
The results with the corresponding R 2 values described above are summarized in Figure 4 with the regression lines fitted to the data set. O c t o b e r 1 7 , 2 0 1 4  From the examined 12 tree species group (Table 1)  In case of the crown dieback data much fewer statistically significant relationships were found (Table 7) than by defoliation (Table 6) data. The impact of the actual year is less pronounced in itself than the previous years or the current year merged with the previous year or years. Much more relationships can be observed for the precipitation indices and indices VK, PADI, P15, Pmax and Pwin show typical relationships among other parameters (but the FAI not). This table is similar to the result of defoliation, but we have here less significant relationships than in the other case.

Results of the correlation studies
The groups of GY-T, B, A tree species show similar results than the result table of all species. The CS, EKL, NNY, HNY, EgyF groups have similar correlation table, but the number of statistically significant relationships is less characteristic than in case of all species (Table 7). In the case of groups K, EF and FF the temperature indices show some relationships, but the precipitation usually not. Further and more detailed studies are needed to comprehesive interpretation of these results. O c t o b e r 1 7 , 2 0 1 4

SUMMARY
In the paper based on a quite huge forest health condition dataset (defoliation and crown dieback data) general relationships and trends of them were searched and identified. The performed investigations and the received results give general overview and help to identify further research directions and tasks. It is necessary to refine the resolution of examined forest health condition data for the more detailed interpretation of the retrieved results and for the exploration of deeper relationships. This includes the examinations of simple sample points and based on the received results we can use some kinds of grouping methods (e.g. clustering, classification) to understanding the deeper relationships.