Parameterization of the Corine Land Cover dataset for the calculation of the energy balance on a high spatial resolution

(1)

Parameterization of the Corine Land Cover dataset for the

calculation of the energy balance on a high spatial resolution

Bachelor Thesis by Daniël Kooij, student Earth Sciences University of Amsterdam, July 2016

Supervisors:

(2)

Abstract

In this study, a high resolution land cover dataset is produced that contains a variety of parameter values necessary to accurately calculate the sensible heat flux on a 100m spatial resolution and 1 hour temporal resolution. This high spatial resolution land cover dataset is produced using the parameterized Global Ecosystem map, developed by the US Geological Survey. This land cover map, which has a 1km spatial resolution is translated to a higher resolution land cover map; which is the Corine Land Cover, developed by the European Environmental Agency. Two different translation methods are conducted. The first translation is based on the nomenclature of the classes of both maps and the second translation is based on the percentage agreement of the classes of both maps when they are compared to each other. Subsequently, the newly produced and parameterized land cover maps. As an result, parameter fields are produced for six parameters, namely; volumetric wilting point, leaf area index, vegetation fraction, maximum water storage capacity and basal crop factor. The accuracy of the produced parameter fields and seasonal variation is analyzed. Whereas the results point out that the Corine Land Cover map lacks the ability to accurately predict parameter fields for areas that encompass multiple climates.

1. Introduction

This research is part of a research project that aims to develop a high resolution sensible heat flux model for Europe, which is able to estimate the sensible heat flux on a 100 by 100 meter spatial resolution and 1 hour temporal resolution. In order to model the sensible heat flux, knowledge about the different types of land cover is required for describing the interactions of the energy fluxes between a land cover and the atmosphere (Bonan et al. 2002). A land cover is any physical groundcover which is located at the Earth's surface. This includes different types of vegetation, anthropogenic structures and even open waters and unvegetated bare areas, all of which have their own distinct characteristics. The characteristics of a land cover can be split into a variety of parameters, which are integrated into the calculations of the diverse energy fluxes. In order to perform these calculations over a certain area, information on the spatial differentiation of different types of land cover is necessary. Such maps are produced using a classification method, whereby the observed physical cover is classified to a certain type of land cover. One of the most widely used global land cover classification is the Global Land Cover Characteristics Database (GLCCD), developed by the U.S. Geological Survey (USGS; U.S. Geological Survey 2001). It has 1 km spatial resolution and contains parameter values for different types of land cover (Loveland et al. 2000). However, the 1 km spatial resolution of this dataset lacks the ability to predict the sensible heat flux, which often behaves on an even higher resolution (Bonan et al. 2002). Therefore, another land cover map, the Corine Land Cover (CLC), developed by the European Environment Agency (EEA; European Environment Agency, 2012) is used. This land cover map has a 100m spatial resolution, which is more suitable in order to calculate the sensible heat flux. Yet, the Corine Land Cover dataset lack parameter values for its land cover classes which the Global Land Cover Characteristics Database does have. Therefore, the classes of the Global Land Cover Characteristics Database are translated to the Corine Land Cover database, so that the parameter values of the GLCCD can be assigned to the CLC classes. This translation is done in two ways, one that is based on the description of the classes of both the databases and another that is based on the percentage agreement between the different

(3)

classes from both land cover maps. In addition, the parameters provided by the USGS do not contain values that vary over time. Therefore, known parameter values and meteorological data is used to describe and model the seasonal variation of these parameters, in other words, the seasonal phenology.

The aim of this study is threefold. Firstly, it tries to accurately translate the Global Ecosystem classes to the Corine Land Cover classes. Secondly, the translated CLC map is parameterized and parameter fields are produced that vary spatially, but often also temporally. Finally, it aims to validate the conducted translation and parameterization using existing parameter fields and knowledge from academic literature and researches.

2. Methods and Materials

2.1.1 Land cover maps

For this study, two different land cover datasets were used. The first land cover dataset, which is parameterized in this research, is the CORINE land cover (CLC) retrieved from the European Environment Agency (EEA; European Environment Agency, 2012). This dataset is using a 3-level nomenclature and consist out of 44 classes in total. The used CLC map has a grid of 100m spatial resolution and the minimum mapping unit is 25 ha. The dataset is updated every 6 years and the used version is the latest update which dates from 2012. The dataset is tested on its thematic accuracy, which is the accuracy of a land cover being correctly classified by the used classification method. Through ground surveys and the use of remote sensing imagery, researchers found that the average thematic accuracy of the preceding CLC map is > 75% (European Environmental Agency, 2007).

The second land cover map is the Global Ecosystems map, which classes are translated to the CLC map, is part of the Global Land Cover Characteristics Dataset (GLCCD) from the United States Geological Survey (USGS; U.S. Geological Survey 2001). This map is produced on the basis of the classification provided by Olson (1994a, 1994b). Olson made this classification using Advanced Very High Resolution Radiometer (AVHRR) data of the year 1992 to 1993 (Eidenshink & Faundeen, 1994). The map has a 1 km spatial resolution and consist out of a total of 98 classes. This dataset is also validated by ground surveys, whereby the thematic accuracy of GE is estimated at 60 to 80% (Scepan, 1999).

2.1.2 Land Surface Parameters & Meteorological Data

In order to calculate the different fluxes of the surface energy balance, including the sensible heat flux, it is necessary to know certain parameter values for different types of land cover. These parameters of interest are: volumetric wilting point (Fpwp), leaf area index (LAI), vegetation fraction

(Cv), maximum water storage capacity (Smax) and basal crop factor (Kcb). The volumetric wilting point

is the volumetric water content at which the plants, typical for that type of land cover, are wilting if the water content is lower than this threshold. The leaf area index is the one-sided green leaf surface area per unit ground surface area and the vegetation fraction is the percentage of ground surface area effectively covered by the vegetation canopy. The maximum water storage capacity is the maximum amount of water that can be stored within the canopy expressed in mm. The basal crop factor is the evaporative power of a type of vegetation compared to the evaporative power of a reference , usually grass or alfalfa.

(4)

The Global Land Cover Characteristics Dataset, in addition to classification, provides some of these parameters values in a land surface parameter (LSP) database for the classes listed in Global Ecosystems. Claussen (1994) was the first to develop this LSP dataset for the classes within GE, since then, this dataset is subsequently adjusted and updated by Hagemann (1999, 2002). The parameters that are retrieved from the latest LSP dataset include the Fpwp and the LAI and Cv for both the

dormant and the growing season. The LSP dataset for the GE classes are implemented in multiple climate models, such as HIRHAM, REMO and ECHAM (Christensen et al., 2001; Gao et al., 2015; Roeckner, 1996). The accuracy of the parameterization of Claussen and Hagemann on the GE classes have been tested using empirical data on the spatial distribution of LAI maps derived from spectral radiance data retrieved from satellite imagery. The accuracy of the LAI was found to be around 85% up to 90% (Defries & Los, 1999). Therefore the updated LSP dataset based on GE is considered to be an adequate benchmark for the parameterization of LAI to be produced in this study.

Subsequently, with the use of meteorological data together with the parameters retrieved from the LSP dataset, the maximum water storage capacity and crop factors are calculated. The meteorological data is retrieved from ERA-Interim, managed by the European Centre for Medium-Range Weather Forecasts (ECMWF) which is a reanalysis of measured weather data. The data retrieved in particular are the maximum, dew and normal temperatures and the wind speed components. The data has a spatial resolution with a detail up to 10 kilometer and the temperatures as well as the wind speed components are estimated at two meters above ground level.

2.2 Translation Processes

In order to acquire the parameter values for the different classes of the CLC dataset, the parameter values of the GE classes are assigned to the CLC classes. This translation was done in two ways. Firstly a translation based on the nomenclature of both classifications and secondly a translation based on a spectral mixture analysis by allocating the parameter values of the GE classes to the CLC classes based on their agreement in percentages. The spectral mixture analysis is a widely used method in remote sensing (Heinz & Chang, 2001). However a spectral mixture analysis tries to determine the percentage of land cover involved in each grid cell, not the other way around. But the concept can be used to estimate parameter values, assuming that a CLC class is represented by more than one GE class.

The translation based on nomenclature is conducted directly on the 3rd level of the CLC classes. For both datasets, the descriptions of the different classes were retrieved from literature in order to compare them to each other. For a GE class to be translated to a CLC class, the descriptions of both classes must be in compliance and should not contradict or exclude each other. When the descriptions were too vague or missing, the GE class was translated to the class that resembled the name of that class and the pictures provided by the USGS and EEA (Anderson et al. 1976; Olson, 1994a, 1994b; Olson et al. 1983; European Environmental Agency, 2000). Furthermore, as two or more GE classes fit the description of one CLC class, the one that most closely matches the class description is selected. However, if not one class of GE meet the description of a CLC class, the class is represented with the class that had the highest percentage agreement for the Netherlands in the comparison part.

Since some of the translations described above are prone to a certain degree of subjectivism, some CLC classes might be falsely assigned to the parameters of their corresponding GE classes. Therefore, a second translation is carried out, based on the percentage agreement between multiple GE classes

(5)

and one CLC class. A CLC class can thereby be represented by multiple GE classes by allocating the parameter values of the corresponding GE classes to the amount of percentage agreement with that CLC class. Since this method resembles that of the concept of 'fuzzy logic', I call this the fuzzy translation.

In order to carry out this translation, the GE map was resized to a spatial resolution of 100 by 100 meter, so that the grid cells of both land cover maps could be compared with each other. Four regions were picked to compare both land cover maps, namely, the Netherlands, Spain, South-Scandinavia and the Alps. Subsequently, the GE classes have to fulfill the following conditions in order to be selected for the translation.

- One Corine land cover class can be represented by a maximum of three Global Ecosystem classes. - The minimum amount of agreement between a CLC class and a GE class has to be 10% to be included in the top three GE classes that represent the CLC class.

- The amount of grid cells one CLC class consist of must be more than 5000 in order to give a reliable representation of that class.

The translation was mainly done using the data from the Netherlands, but when insufficient amount of grid cells are available for a particular class, the class is translated using data from Spain, South-Scandinavia and the Alps. In so doing, the region with the relative highest amount of grid cells for that class is selected for the translation. The produced class allocation table and additional information for this translation is provided in appendix 2 section 2 and 3.

The comparison is also performed on a 1km basis, to see if there were any major changes between a comparison on the lower spatial resolution, the GE map, instead of on the higher spatial resolution, the CLC map. The difference of agreement between the CLC and GE classes for the different resolutions within the Netherlands is on average 3%, with a maximum of 12% difference for the agreement between any GE and CLC class.

2.3 Parameter calculations

Some of the land surface parameters have a seasonal component, which means they differ gradually through time and space based on the location-specific climatic conditions and seasonality through the year. Specifically, these temporal and spatial varying parameters are the leaf area index, the maximum water storage capacity and the crop factor. In order to catch the seasonality of these parameters, the aforementioned meteorological data from ERA-Interim is used and interpolated to fit the higher resolution produced parameterized CLC map.

Table 1. Example of Global Ecosystem class that is translated to a Corine land cover class based on their matching descriptions.

CLC class Description GE class Description

Coniferous Forest

24 Coniferous trees represent more than 75 % of the formation. Three heights under normal climatic conditions are higher than 5 m

Conifer Forest 27 Coniferous trees represent more than 75% of the formation. More than 60% of the vegetation should consist out of trees.

(6)

As can be expected, the leaf area index (LAI) increases during spring and decreases during autumn, especially in areas with an high amount of deciduous vegetation. In order to capture the seasonal variation of the leaf area index, a formula proposed by Hagemann et al. (1999) is used, where the LAI fluctuates between the maximum and minimum LAI of a certain land cover retrieved from the land surface parameter (LSP) dataset. With this formula the LAI at week i is defined by

Eq. (1)

Here LAI_max is the maximum LAI for a given land cover, analogous to the LAI in the growing season, LAIg. Conversely, LAImin is the minimum LAI for the dormant season, analogous to LAId. In addition, Fi

is the growth factor of week i, which is a coefficient between 0 and 1, multiplied by the difference between LAI_d and LAI_gto determine the LAI at a given time. Based on Hagemann et al. (1999) this growth factor can be deduced from spectral radiance or temperature data. For higher latitudes, the growth factor can be computed as a product from a relative difference in maximum, minimum and current temperature, such that

Eq. (2)

Where Tmax and Tmin are the average site-specific hottest and coldest month of the past 30 years,

respectively. Ti is the 30-day average temperature of week i. Hagemann assumes that the maximum LAI is present for T_max > 298K, so that T_max is always less or similar to 298K. The same holds for T_min, where the minimum vegetation is present for Tmin < 0. In order to avoid unreal growth curves, the

growth factor of the previous week cannot be higher than the next week before the midst of the year is reached, and vica versa for the second half of a year.

Using the produced seasonal LAI values, it is possible to calculate the basal crop factor and the maximum water storage capacities. For the crop factor, a formula from the FAO is used (Allen et al. 1998), where the crop factor for natural vegetation is defined by

Eq. (3)

In which LAI and Kcbmid are respectively the leaf area index and the crop factor at week i. The Kcb

has the subscript 'mid' so it refers to the mid-season. In order to avoid any confusion, these calculated Kcbmid values are applicable to any time of the year due to the use of the seasonally

varying LAI in the formula. Subsequently, Kcmin is the minimum crop coefficient for a dry soil with no

ground cover and Kcb_full is the maximum crop coefficient for a mid-season fully developed land cover. Kcbfullis not given for every type of vegetation, but calculated based on another equation

from Allen (1998), whereby

Eq. (4)

U2 is the 30-year average wind speed during the mid-season, composed of the wind components from ERA-Interim. RHmin is the 30-year average mean minimum daily relative humidity during the

(7)

Kcbh is a proxy for Kcbmid based on vegetation height for fully developed vegetation. The Kcbh can be

approximated as

Eq. (5)

Where h is the mean maximum vegetation height for a specific land cover. Allen et al. (1998) proposes that kcbh is limited to 1.2 for vegetation with a mean maximum height larger than 2

meters.

Finally, the maximum water storage capacity of a certain land cover, or S_max as it is called, can also be determined using the produced LAI values. Smax is heavily related to the amount of leaves present

for a type of vegetation, Since the intercepted water is stored on top of the leaves. S_max can therefore be calculated based on a linear relationship with the LAI, as shown in Eq (6).

Eq. (6)

Where Smax is the storage capacity in millimeters and LAI is the leaf area index of week i. For types of

land cover without any vegetation the standard storage capacity is 1 mm. The formula above (eq6) is being used for all the different types of land cover. Due to a lack of time, the calculation of the maximum water storage capacity is being generalized. However, for many types of land cover, especially those with a high amount of vegetation cover, the calculation of Smax does not differ much

from the average equation which is eq. 6 (Aston, 1979; Goméz, 2001).

2.4 Data analysis and validation

In order to determine the accuracy of the translation of the GE classes to the CLC classes a

comparison is conducted for the three parameters retrieved from Hagemann (2002) between the GE map and both the translated CLC maps. For the analysis of the accuracy of the spatial differentiation the parameter leaf area index during the growing season is investigated. Using the parameterized description- and fuzzy based translated maps and the original GE map, the difference of LAI across Europe is mapped on a 1km scale. The other parameters are omitted in this analysis, as the LAI is the only parameter of the LSP dataset from which the accuracy is tested on the GE map and known to be higher than 85% (Defries & Los, 1999). In addition, most of the calculated parameters are originally partially derived from the LAI.

(8)

3 Results

3.1 Validation of the translation processes

The following section describes how well the translated CLC maps corresponds with the original GE map. Using the description based translation, 16,4% of the total area hold the same class for both the translated CLC map and the GE map. The GE map may contain 98 classes, however, only five classes from GE, conifer boreal forest, crops and towns, cool crops and towns, forest and field, conifer boreal forest and grass crops represent 60% of the total land area. Moreover, 60% of the land area of the CLC map is also represented by only five classes, namely: Non-irrigated arable land, broad-leaved forest, coniferous forest, mixed forest and pastures. For the description based translation (DBT), only two of above named CLC classes are represented by one out of the top five most prevalent GE classes. In addition, in the DBT map just 23 GE classes represent the CLC classes. Remarkably, for the fuzzy based translation (FBT) only 19 GE classes represent the CLC classes, despite the fact that the fuzzy translation request almost thrice the number of representative classes. However, due to the combination of multiple GE classes representing one CLC class, this translation does contain more unique parameter values for each land cover. In addition, using the fuzzy logic method, the sum of the agreement of the top three GE classes for every CLC class is 77% on average (Appendix 2.2 column 8).

The distribution of land area over two types of land cover, forest and urban areas, are compared between the acquired CLC and GE map and the newly produced translated fuzzy and description based maps. From table 2 it can be derived that due to a lack of spatial resolution, the GE map indentifies only a quarter of urban areas of what the CLC map observes. While the description based translation has the same amount of urban area as the CLC map, the fuzzy translation with 1,16% urban area is more equal to the GE map. Furthermore, while the CLC, GE and DBT maps have around the same amount of forest area, the FBT map has zero recognized forests, as the agreement between the CLC forest classes and GE forest classes is lower than the agreement of the top three classes that represent a CLC forest class. The forest classes of the FBT are now mainly recognized as intermediate classes between open areas and forests (> 75% covered with trees). For instance, the GE class 'forest and field', which contains between 40 - 60% trees, now represents an average 54% of the forest classes, but which is strictly speaking not seen as a forest class according to the GE classification.

Table 2. Percentage of occurrence of different types of land cover. A class is regarded forest when > 75% of the vegetation cover consists of trees. The fuzzy based translation is also included in this table, by multiplying the translated area of the classes with the percentage of urban or forest agreement for a class.

CLC

GE

Description

Fuzzy

Urban % 3.86 % 0.79 % 3.84 % 1.16 %

Urban total 198.803 km2 40.628 km2 197.489 km2 59.514 km2

Forest % 31.32 % 28.82 % 33.26 % 0 %

(9)

Table 3. Averages differences in parameter fields of the GE map and the CLC maps for both translation methods. The positive and negative means only contain the positive and negative differences, respectively. All the means and standard deviations are expressed in percentages.

The difference in parameter values between the GE map and parameterized CLC map for both translations are presented in table 3a and 3b for the three parameters retrieved from Hagemann (2002). Variation in LAI between the original GE map and the DBT map is 14,6 %, which points out that on average, the LAI is overestimated. Yet, it is interesting to see if there are also negative LAI values, as the positive and negative LAI values cancel themselves out when taking an average over the whole dataset. It turns out that the average of the positive LAI differences is 33%, more than twice as high than the normal mean. The average of the negative LAI differences is -18%, which is less negative than the positive is positive. The standard deviation for all the means of the three parameters are quite high, even ranging between 20-30%, indicating that the difference in LAI, C_v and F_pwp between two cells may vary considerably. For the vegetation fraction, the average difference is for both the FBT and the DBT almost the same. The mean negative and positive means are also quite similar. It should be noted that the average vegetation fraction for the FBT is now positive, whereas the average LAI for the FTB was negative. Changes in volumetric wilting point are less extreme, ranging between a positive mean of 6,7% and negative mean of - 5,7%. Also the standard deviation is much less, as could be expected, since the values of Fpwp in the LSP dataset

differ between a minimum and maximum of 0,38 to 0,55, thus a maximum difference of 45 % could be achieved.

Figure 2, LAI difference between the parameterized CLC map and the GE map during the growing season for the translated CLC map based on descriptions (left) and based on the fuzzy logic (right). The difference is expressed in LAI, whereas a negative difference means an underestimation of the LAI and a positive difference an overestimation of the LAI compared to the LAI of the GE map (CLC - GE).

a) Description based translation

LAI growing Cv growing Fpwp

Mean ± SD 14.6 ± 21.9 % 7.6 ± 26.0 % 0.9 ± 7.2 %

Positive mean ± SD 33.3 ± 22.8 % 18.2 ± 29.9 % 6.7 ± 7.0 % Negative mean ± SD - 18.6 ± 2.7 % -10.5 ± 3.7 % -5.7 ± 4.1 % b) Fuzzy based translation

LAI growing Cv growing Fpwp

Mean ± SD -17.9 ± 26.8 % 7.9 ± 23.3 % 0.6 ± 6.0 %

Positive mean ± SD 8.8 ± 28.8 % 16.2 ± 25.4 % 5.4 ± 6.3 % Negative mean ± SD - 26.7 ± 1.7 % -8.2 ± 3.3 % -4.8 ± 2.8 %

(10)

3.2 Accuracy of the parameter fields

The results from the previous paragraph point out that the different parameters vary for the different translation methods. However, it does not determine the spatial differentiation of the analyzed parameters throughout Europe to see whether these differences are clustered in certain areas or spread gradually across the land surface. Figure 2 visualizes this spatial difference in LAI, showing the description based translation (DBT) on the left side and the fuzzy translation (FBT) on the right side. For the DBT the overall LAI is overestimated (table 3a), but this overestimation is strongly centered in the Nordic countries, while in the south, the LAI tends to be more negative. This trend is shown in figure 3a, where the average LAI difference (blue line) is approaching the average of the positive LAI differences (yellow line) with higher latitudes. For Scandinavian coniferous forest, the LAI difference can reach up to 4, which is pretty significant, given that the highest LAI that occurs in the GE LAI map is 9,7.

For the FBT map, the difference in LAI is negative almost everywhere. In figure 3c and 3d, the average difference in LAI is closely linked to the average of the negative LAI differences, meaning that the LAI at higher as well as at lower latitudes and longitudes is underestimated. Interestingly, the LAI for countries with a maritime climate located at the middle latitudes, the LAI almost show no deviation. This is also shown in figure 3c, where the average difference in LAI approaches zero between 50 and 60 degrees latitude. Since the fuzzy logic translation is performed on the Netherlands, which allocate the classes based on the percentage of agreement, the LAI of the FBT comply well with the LAI values of the GE map in these regions.

Figure 3. The average difference in LAI is plotted against the latitude or longitude. The upper plots visualize the average LAI (blue line) for the description based translation over the latitude (a) and longitude (b). While the plots underneath visualize the average LAI for the fuzzy based translation also over the latitude (c) and longitude (d). The red lines are the averages taken only over the negative differences, while the yellow line are the averages of only the positive differences. The red and blue line indicate if the difference in LAI at a certain lat- or longitude is truly zero, or because the negative and positive values cancel themselves out over the distance

(11)

4. Discussion

The result show that parameterized and translated CLC maps contain different deviations for the variety of parameters when compared to the parameterized GE map. The Olson classification which produces the GE map is based on plant phenology, as it uses radiometer data to differentiate the diverse types of land cover (Olson et al. 1983). Because vegetation grows better under weather conditions that suits their needs, the plant phenology is largely dependent on the average weather conditions for a specific location. Thus the climate plays an important role in the location of the appearance of some ecosystems in the GE map. On the contrary, the CLC map is based on photo interpretation using an hierarchical decision tree (European Environment Agency, 2000). The emphasis of this classification is not on the basis of plant phenology. As a result, vegetation types that contain the same observable characteristics are assigned to the same land cover, whether or not they are located in different climate regions. Now the description based translation is conducted on the Netherlands, where the climatic conditions are very suitable for plant development. As a consequence, land cover types in specific climates such as boreal coniferous forest are now indicated as 'normal' coniferous forest, resulting in an overestimation of the LAI in these regions due to the negligence of the harsh environment, which usually results in lesser plant growth. The same accounts for land cover types with an average low amount of LAI in areas which have a very suitable environment, resulting in the spatial differentiation of the LAI as visualized for the Alps in figure (4). For using the CLC map for climate modeling purposes, I urge for an improved translation or classification of the CLC map in order to catch the climatological aspect.

I expected that the fuzzy based translation would be more climate-related as the CLC classes are now represented by the average parameter values of the three matching GE classes. The benefit of the fuzzy translation, is that heterogeneous classes such as mixed coniferous and deciduous forests are represented by the average actual composition of both the individual coniferous and deciduous parameters. The drawback is that many other classes consisting out of one type of land cover are suddenly assigned to a composition of parameter values of multiple unrelated GE classes. Which on its turn causes an overall equalization of the different parameters values. As is visualized in figure 2, the result is an underestimation of the LAI for types of land cover that have an high amount of LAI, such as trees. Due to the lower resolution of the GE map, it does not recognize types of vegetation that constitute less than 50% of the land cover for any of the 1 km2 grid cells (Loveland et al. 2000). This cartographic generalization causes that for CLC land covers consisting out of a numerous amount of small areas have a high agreement with arbitrary GE classes. It is likely that GE classes that are prevalent throughout Europe, such as grassland, return often in the top three GE classes to represent a CLC class. In order to improve this classification method, I would suggest to perform this translation on specific locations where a large area is assigned to the same CLC class and where the lower resolution GE map also recognizes that type of land cover due to its size. This way, the CLC class is

Figure 4. When the Alps are zoomed in for the description based translation map, a clear distinction can be seen in the LAI difference between northern and the southern part of the Alps. Because the southern part of the Alps lies within the rain shadow, vegetation growth is retained. While on the other hand, the northern part has more favorable weather conditions to support vegetation growth. The resulting difference in vegetation development is not accounted for in the classification of the CLC map.

(12)

matches with types of land cover that are also truly identified by the GE classification and not matched with other land covers that are the result of an overlap due to a difference in spatial resolution. Subsequently, a part of the difference in parameter fields for both translation methods is likely caused by a change in land cover over the years. The time difference between the production of the GE map and the production of the CLC map is more than a decade. However, this is only causing a minor deviation, as, for instance, urban areas did not increase fourfold in the past ten years (table 2).

As mentioned at the results, a few CLC classes cover a large part of the different analyzed areas in Europe. In the Netherlands, around 26% of the land area is classified as ‘non-irrigated arable land’. Yet, agricultural areas itself consist out of numerous different types of agricultural land covers, determined by the crop that is cultivated. crop factors for different crops can vary widely (Allen et al. 1998), which would influence the sensible heat flux due to transpiration differences significantly, but which is now neglected through the classification of these diverse areas under one class and thus one crop factor. This is not an unusual problem, the classification of agricultural areas remains difficult, according to Loveland (2000); ''due to the high temporal and spatial diversity, combined with the

availability of limited agricultural references’’. The crop factors were calculated using the LAI, for

natural vegetation this makes sense, as natural vegetation does not have a harvest or sowing period. However, also cropland was calculated in the same way, due to its earlier mentioned spatial diversity and uncertainty of present crop species. Yet, this may overestimate the crop factors before and after the sowing and harvest period, respectively. I recommend to allocate more effort in the parameterization of the prevalent classes, especially those who are have an heterogeneous nature, in order to produce accurate parameter fields for use in climate modeling.

Finally, the added seasonality seems to have about the right growth curve through the year. However, the growth factor formula is a simple equation, purely based on the difference between the current and climatological temperatures. Because the seasonality of all the parameters is based on the growth factor through the use of the LAI, the temporal variety of the parameters are heavily influenced by the temperatures. It would be interesting to look to more advanced equations for temperature based vegetation growth, such as the growing degree days (McMaster & Wilhelm, 1997). Moreover, a major part of the parameterization of the land cover maps could be replaced by directly retrieving the LAI values using high spatial resolution NDVI data. Though this might be time-consuming and costly, based on the requested amount of maps through the year.

5. conclusion

The classification systems of the CLC and GE database are both based on a different source of data. Thus the classes of both maps show little correspondence with each other. The spatial differentiation of the parameters for both translation methods varies widely. Overall, the description based translation overestimates the LAI in Europe, whereby the LAI is underestimated in lower latitudes, gradually changing to an overestimation of the LAI in the northern latitudes. The fuzzy based translation underestimates the LAI for both the lower and higher latitudes and longitudes. Yet, the FBT seems to be fairly accurate around the mid-latitudes, showing almost no difference in the LAI compared to the GE map. Both translation methods show potential for the use in high resolution climate modeling. However, due to the vast extent of Europe, the influence of the climate can be observed. This influence is not accounted for in the CLC map, which lacks the ability to accurately predict parameter fields for areas that encompass multiple climatic regions. More research is needed to account for the influence of the climate and overall improvement of translating the GE map to a land cover map on a higher resolution over the extent of Europe.

(13)

References

Allen, R. G., Pereira, L. S., Raes, D. & Smith, M. (1998). Crop evapotranspiration: Guidelines for computing crop water requirements. FAO Irrigation and Drainage Paper 56, 300 pp.

Anderson, J. R., Hardy, E. E., Roach, J. T. & Witmer, R. E. (1976).A Land Use And Land Cover Classification System For Use With Remote Sensor Data. Geological Survey Professional Paper 964.U.S. Geological Survey, Washington. Available at: http://pubs.usgs.gov/pp/0964/report.pdf.

Aston, A.R., (1979) Rainfall interception by eight small trees. Journal of Hydrology, 42, pp. 383–396. Bonan, G. B., Levis, S., Kergoat, L. & Oleson, K. W. (2002). Landscapes as patches of plant functional types: An integrating concept for climate and ecosystem models. Global Biochemical Cycles. 15 (2), GLOBAL BIOGEOCHEMICAL CYCLES

Carson, T. N. & Ripley, D. A. (1997). On the Relation between NDVI, Fractional Vegetation Cover, and Leaf Area Index. Remote Sensing of Environment. 62, 241-252.

Christensen, J.H., Christensen, O.B., Schulz, J-P., Hagemann, S. & Botzet, M. (2001). High resolution physiographic data set for HIRHAM 4: An application to 50 km horizontal resolution domain covering Europe. DMI, Technical Report 01-15 4

Christensen, J.H., Christensen, O.B., Lopez, P., van Meijgaard, E., & Botzet, M. (1996). The HIRHAM 4 regional atmospheric climate model. DMI Scientific Report 96-4, Copenhagen, Denmark.

Claussen, M., Lohmann, U., Roeckner E. & Schulzweida, U. (1994). A global dataset of land surface parameters. Max-Planck-Institute for Meteorology, Report 135, Hamburg, Germany

Eidenshink, J.C. & Faundeen, J.L. (1994). The 1 km AVHRR global land data set-first stages in implementation: International Journal of Remote Sensing, 15 (17), 3443-3462.

European Environment Agency (2000). CORINE land cover technical guide – Addendum 2000. Technical report No 40, Copenhagen, available at: file:///C:/Users/Eigenaar/Downloads/tech40add%20(1).pdf. European Environment Agency (2007). CLC2006 technical guidelines. Technical report No 17,

Copenhagen, available at:

http://www.eea.europa.eu/publications/technical_report_2007_17/download

European Environment Agency (2012). CLC2012 addendum to CLC2006 technical guidelines. EEA Technical Report by the European Topic Centre Spatial Information and Analysis, Copenhagen, available at:

http://gamta.lt/files/Addendum_finaldraft.pdf.

Gao Y., Weiher S., Markkanen T., Pietikäinen J.-P., Gregow H., Henttonen H.M., Jacob D. & Laaksonen, A. (2015). Implementation of the CORINE land use classification in the regional climate model REMO. Boreal Env. Res. 20: 261–282.

Goméz, J. A., Giráldez, J. V. & Fereres, E. (2001). Rainfall interception by olive trees in relation to leaf area. Agricultural Water Management. 49, 65–76.

Hagemann, S., Botzet, M., Dümenil, L. & Machenhauer, B. (1999). Derivation of global GCM boundary conditions from 1 km land use satellite data. Max-Planck-Institute for Meteorology, Report 289, Hamburg, Germany.

Hagemann, S. (2002). An Improved Land Surface Parameter Dataset for Global and Regional Climate Models. MPI Report 336; MPI: Hamburg, Germany, Volume 336.

(14)

Heinz, D. C. & Chang, C. L. (2001). Fully constrained least squares linear spectral mixture analysis method for material quantification in hyperspectral imagery. IEEE Trans. on Geoscience and Remote Sensing. 39, 529 - 545.

Loveland T.R., Reed B.C., Brown J.F., Ohlen D.O., Zhu Z., Yang L. & Merchant J.W. (2000). Development of a global land cover characteristics database and IGBP DISCover from 1 km AVHRR data. Int. J. Remote Sensing. 21,1303–1330.

McMaster, G. S. & Wilhelm, W. W. (1997). Growing degree-days: one equation, two interpretations. Agric. For. Meteorol., 87 (4), 291–300.

Olson, J.S., Watts J. A. & Allison, L. J. (1983). Carbon in live vegetation of major world ecosystems. ORNL-5862, Oak Ridge National Laboratory, Oak Ridge

Olson, J.S., (1994a). Global ecosystem framework-definitions: USGS EROS Data Center Internal Report, Sioux Falls, SD, 37 p.

Olson, J.S., (1994b). Global ecosystem framework-translation strategy: USGS EROS Data Center Internal Report, Sioux Falls, SD, 39 p.

Defries, R .S . & Los, S .O. (1999). Implications of Land-Cover Misclassification for Parameter Estimates in Global Land-Surface Models : An Example from the Simple Biosphere Model (SiB2): Photogrammetric Engineering and Remote Sensing, v. 65, no . 9, p. 1083-1088.

Roeckner, E., Arpe, K., Bengtsson, L., Christoph, M. Claussen, L. Dümenil, L., Esch, M., Giorgetta, M., Schlese, U. & Schulzweid, U. (1996). The atmospheric general circulation model ECHAM-4: model description and simulation of present-day climate. Max-Planck-Institute for Meteorology, Report 218, Hamburg, Germany

Scepan, J . (1999). Thematic Validation of High-Resolution Global Land-Cover Data Sets. Photogrammetric Engineering and Remote Sensing, 65 (9),1051-1060.

The thematic accuracy of Corine land cover 2000; Assessment using LUCAS (land use/cover area frame statistical survey). http://land.copernicus.eu/user-corner/technical-library/technical_report_7_2006.pdf U.S. Geological Survey (2001). Global land cover characteristics, data base version 2.0. U.S. Geological Survey. Available at: https://lta.cr.usgs.gov/GLCC.

(15)

Appendixes

Appendix 1.1

Global ecosystem legend

Indice Name Indice Name

1 Urban 50 Sand Desert

2 Low Sparse Grassland 51 Semi Desert Shrubs

3 Coniferous Forest 52 Semi Desert Sage

4 Deciduous Conifer Forest 53 Barren Tundra

5 Deciduous Broadleaf Forest 54 Cool Southern Hemisphere Mixed Forests

6 Evergreen Broadleaf Forests 55 Cool Fields and Woods

7 Tall Grasses and Shrubs 56 Forest and Field

8 Bare Desert 57 Cool Forest and Field

9 Upland Tundra 58 Fields and Woody Savanna

10 Irrigated Grassland 59 Succulent and Thorn Scrub

11 Semi Desert 60 Small Leaf Mixed Woods

12 Glacier Ice 61 Deciduous and Mixed Boreal Forest

13 Wooded Wet Swamp 62 Narrow Conifers

14 Inland Water 63 Wooded Tundra

15 Sea Water 64 Heath Scrub

16 Shrub Evergreen 65 Coastal Wetland, NW

17 Shrub Deciduous 66 Coastal Wetland, NE

18 Mixed Forest and Field 67 Coastal Wetland, SE

19 Evergreen Forest and Fields 68 Coastal Wetland, SW

20 Cool Rain Forest 69 Polar and Alpine Desert

21 Conifer Boreal Forest 70 Glacier Rock

22 Cool Conifer Forest 71 Salt Playas

23 Cool Mixed Forest 72 Mangrove

24 Mixed Forest 73 Water and Island Fringe

25 Cool Broadleaf Forest 74 Land, Water, and Shore (see Note 1) 26 Deciduous Broadleaf Forest 75 Land and Water, Rivers (see Note 1)

27 Conifer Forest 76 Crop and Water Mixtures

28 Montane Tropical Forests 77 Southern Hemisphere Conifers

29 Seasonal Tropical Forest 78 Southern Hemisphere Mixed Forest

30 Cool Crops and Towns 79 Wet Sclerophylic Forest

31 Crops and Town 80 Coastline Fringe

32 Dry Tropical Woods 81 Beaches and Dunes

33 Tropical Rainforest 82 Sparse Dunes and Ridges

34 Tropical Degraded Forest 83 Bare Coastal Dunes

35 Corn and Beans Cropland 84 Residual Dunes and Beaches

36 Rice Paddy and Field 85 Compound Coastlines

37 Hot Irrigated Cropland 86 Rocky Cliffs and Slopes

38 Cool Irrigated Cropland 87 Sandy Grassland and Shrubs

39 Cold Irrigated Cropland 88 Bamboo

40 Cool Grasses and Shrubs 89 Moist Eucalyptus

41 Hot and Mild Grasses and Shrubs 90 Rain Green Tropical Forest

(16)

43 Savanna (Woods) 92 Broadleaf Crops

44 Mire, Bog, Fen 93 Grass Crops

45 Marsh Wetland 94 Crops, Grass, Shrubs

46 Mediterranean Scrub 95 Evergreen Tree Crop

47 Dry Woody Scrub 96 Deciduous Tree Crop

48 Dry Evergreen Woods 99 Interrupted Areas (Goodes

Homolosine Projection)

49 Volcanic Rock 100 Missing Data

Appendix 1.2

Corine Land Cover legend

First-level

Second-level nomenclature Third-level nomenclature Nr.

Ar

tifi

cia

l s

ur

face

s

Urban fabric Continuous urban fabric 1

Urban fabric Discontinuous urban fabric 2

Industrial, commercial and transport units Industrial or commercial units 3

Industrial, commercial and transport units

Road and rail networks and

associated land 4

Industrial, commercial and transport units Port areas 5

Industrial, commercial and transport units Airports 6

Mine, dump and construction sites Mineral extraction sites 7

Mine, dump and construction sites Dump sites 8

Mine, dump and construction sites Construction sites 9

Artificial, non-agricultural vegetated areas Green urban areas 10 Artificial, non-agricultural vegetated areas Sport and leisure facilities 11

Ag

ricu

lt

ur

al

a

rea

s

Arable land Non-irrigated arable land 12

Arable land Permanently irrigated land 13

Arable land Rice fields 14

Permanent crops Vineyards 15

Permanent crops Fruit trees and berry plantations 16

Permanent crops Olive groves 17

Pastures Pastures 18

Heterogeneous agricultural areas

Annual crops associated with

permanent crops 19

Heterogeneous agricultural areas Complex cultivation patterns 20

Heterogeneous agricultural areas

Land principally occupied by agriculture, with significant

areas of natural vegetation 21

Heterogeneous agricultural areas Agro-forestry areas 22

Fo

res

t

an

d s

emi

na

tur

al

a

rea

s

Forests Broad-leaved forest 23

Forests Coniferous forest 24

Forests Mixed forest 25

Scrub and/or herbaceous vegetation associations Natural grasslands 26 Scrub and/or herbaceous vegetation associations Moors and heathland 27 Scrub and/or herbaceous vegetation associations Sclerophyllous vegetation 28 Scrub and/or herbaceous vegetation associations Transitional woodland-shrub 29

(17)

Open spaces with little or no vegetation Beaches, dunes, sands 30

Open spaces with little or no vegetation Bare rocks 31

Open spaces with little or no vegetation Sparsely vegetated areas 32

Open spaces with little or no vegetation Burnt areas 33

Open spaces with little or no vegetation Glaciers and perpetual snow 34

Wet

lan

ds

Inland wetlands _{Inland wetlands} Inland marshes _{Peat bogs} 35 ₃₆

Maritime wetlands Salt marshes 37

Maritime wetlands Salines 38

Maritime wetlands Intertidal flats 39

Wat

er

bod

ies

Inland waters Water courses 40

Inland waters Water bodies 41

Marine waters Coastal lagoons 42

Marine waters Estuaries 43

Marine waters Sea and ocean 44

Appendix 2.1

Description based translation of Global Ecosystems to Corine Land Cover

CLC nr. CLC name CLC description GE nr. GE name GE description 1 Continuous urban fabric

Buildings, roads and artificially surfaced areas cover more than 80 % of the total surface

1 Urban Comprised mainly of construction, lawn with or without parks with plantings, weeds vegetation and a cold to hot climate.

2 Discontinuou s urban fabric

Buildings, roads and artificially surfaced areas cover between 50 to 80 % of the total surface

3 Industrial or commercial units

Artificially surfaced areas (cement, asphalt, tarmacadam or stabilized e.g. beaten earth) without vegetation occupy most of the area

4 Road and rail networks and associated land

Motorways and railways, including associated installations (stations, platforms, embankments). Minimum width for inclusion: 100m

5 Port areas Infrastructure of port areas, including quays, dockyards and marinas.

6 Airports Airport installations: runways, buildings and associating lands.

2a Low Sparse Grassland

Comprised of steppe, herb with or without sparse (less than 20%) woody vegetation having shortgrass or hummock vegetation and a dry or cold climate.

(18)

7 Mineral extraction sites

Areas with open-pit-extraction of construction material (sand pit, quarries) or other mineral (open-cast mines). Includes flooded gravel pits, except for river-bed extraction.

8 Dump sites Public, industrial or mine dump sites. 1 Urban Comprised mainly of construction, lawn with or without parks with plantings, weeds vegetation and a cold to hot climate.

9 Construction sites

Spaces under construction development, soil or bed rock excavations, earthworks.

10 Green urban areas

Areas with vegetation within the urban fabric, including parks, cemeteries with vegetation, and mansions and their grounds. Green urban areas concern all laid out vegetated areas greater than 25 ha which are weither situated inside or in contact with urban fabrics. Greenery with strips of lanes and pathes may be found within these areas created for recreational use.

30 Cool Crops and Towns

Comprised of farms with or without pasture, settlements with summer crop vegetation and a snowy winter climate.

11 Sport and leisure facilities

Camping ground, sport ground, leasure parks, golf courses, race courses, etc. Includes formal parks not surrounded by urban areas.

12 Non-irrigated arable land

Cereals, legumes, fodder crops, root crops and fallow land. Includes flowers and tree (nurseries

cultivation) and vegetables, with more than 75 % of area under rotation system.

93 Grass Crops Comprised of cereal, hay or grass greater than 60 percent cover with or without settlements, having annual tops vegetation and a cool-hot, wet-dry climate.

13 Permanently irrigated land

Crops irrigated permanently or periodically, using a permanent infrastructure (irrigation channels, drainage network). Most of these crops could not be cultivated without an artificial water supply.

31 Crops and Towns

Comprised of farms with or without pasture and towns, having a mixed crop vegetation with or without irrigation and a mild to hot, wet to dry climate.

14 Rice fields Land prepared for rice cultivation. Flat surfaces with irrigation channels. Surfaces periodically flooded.

36 Rice Paddy and Field

Comprised of plain or terrace paddies, having 1-3 crops per year, often flooded or irrigated.

15 Vineyards Areas planted with vines, where vineyard parcels exceed 50 % of the area and/or they determine the land use of the area.

16 Fruit trees and berry plantations

Parcels planted with fruit trees or shrubs: single or mixed fruit species, fruit trees associated with

permanently grassed surfaces.

56 Forest and Field

Comprised of forest with cropland, shrub or grassland, having mixed or deciduous vegetation and a mild-hot climate. Around 40% to 60% consist of woody

(19)

vegetation. 17 Olive groves Areas planted with olive trees,

including mixed occurrence of olives trees and vines on the same parcel, where olive parcels exceed 50 % of the area.

56 Forest and Field

Comprised of forest with cropland, shrub or grassland, having mixed or deciduous vegetation and a mild-hot climate. Around 40% to 60% consist of woody vegetation.

18 Pastures Dense grass cover, of floral composition, dominated by graminaceae, not under a rotation system. Mainly for grazing, but the folder may be harvested mechanically.

19 Annual crops associated with permanent crops

Non-permanent crops (arable land or pastures) associated with permanent crop on the same parcel.

Permanent crops are either in juxtaposition with arable

lands/pastures or located along the border of the parcels. The occupying rate of non-permanent crops is more than 50 %.

20 Complex cultivation patterns

Juxtaposition of small parcels of diverse annual crops, pasture and/or permanent crops. Arable land,

pasture and orchards each occupy less than 75% of the total

surface area of the unit.

35 Corn and Beans Cropland

Comprised of rowcrops: maize, soy, cover with alternation vegetation with or without irrigation, with less than 20% natural vegetation. 21 Land principally occupied by agriculture with significant areas of natural vegetation

Areas principally occupied by agriculture, interspersed with significant natural areas. Agricultural land occupies between 25 and 75% of the total surface of the unit.

55 Cool Fields and Woods

Comprised mainly of crop, herb and shrub, with 20 to 40% consisting out of woody vegetation. with summer green or evergreen vegetation and a snowy climate.

22 Agro-forestry areas

Annual crops or grazing land under the wooded cover of forestry species.

56 Forest and Field

Comprised of forest with cropland, shrub or grassland, having mixed or deciduous vegetation and a mild-hot climate. Around 40% to 60% consist of woody vegetation.

23 Broad-leaved forest

Vegetation formation composed principally of trees, including shrub and bush under storeys, where broad-leaved species predominate. With a crown cover of more than 30 % or a 500 subjects/ha density for plantation structure, broad-leaved trees

represent more than 75 % of the planting formation

26 Deciduous Broadleaf Forest

Broad-leaved trees represent more than 75% of the formation. Comprised of medium or tall forest, having cold-deciduous vegetation and a mild-hot climate. More than 60% of the vegetation should consist out of trees.

24 Coniferous forest

Coniferous trees represent more than 75 % of the formation. Three heights

27 Conifer Forest

Coniferous trees represent more than 75% of the formation. More than 60% of

(20)

under normal climatic conditions are higher than 5 m

the vegetation should consist out of trees.

25 Mixed forest Vegetation formation composed principally of trees, including shrub and bush under storeys, where neither broad-leaved nor coniferous species predominate. With a crown cover of more than 30 % or a 500 subjects/ha density for plantation structure. The share of coniferous or broad-leaved species does not exceed 25 % in the canopy closure.

24 Mixed Forest Comprised of low-medium forest woodland and situated in a mild-hot climate. Having conifer and broadleaf vegetation, from which neither of both vegetation cover more than 75% or less than 25% of the area.

26 Natural grasslands

Natural grasslands are areas, where herbaceous vegetation (maximum height is 150 cm and gramineous species are prevailing) which cover at least 75 % of the surface covered by vegetation.

93 Grass Crops Comprised of cereal, hay or grass greater than 60 percent cover with or without settlements, having annual tops vegetation and a cool-hot, wet-dry climate.

27 Moors and heathland

Vegetation with low and closed cover, dominated by bushes, shrub and herbaceous plants. Temperate shrubby area vegetation: includes dwarf forest trees with a 3 m maximum height in climax stage.

64 Heath Scrubs Comprised of dwarf-tall scrub on an often acid soil. Having evergreen-mixed

vegetation with less than 20% woody vegetation and a mild to cold climate.

28 Sclerophyllo us

vegetation

Bushy evergreen sclerophyllous vegetation, including maquis and garrigue mattoral and phrygana.

79 Wet

Sclerophylic Forest

Comprised of tall open Eucalypt forest, having broadleaf evergreen vegetation and a mild, wet climate

29

Transitional woodland-shrub

Bushy or herbaceous vegetation with scattered trees. Arborescent

mattorals which are pre- or post-formation of broad-leaved evergreen forest with a usually thick evergreen shrub stratum. Composed of

evergreen oaks, olive trees, carob trees or pines the crown cover density of which is less than 30 % of the surface area.

47 Dry Woody Shrub

Comprised of shrub with or without low or open woodland, with less than 40% tree vegetation, having evergreen vegetation and a dry or cool to warm climate.

30

Beaches, dunes, sands

Beaches, dunes and expanses of sand or pebble in coastal or continental locations, including beds of stream channels with torrential regime, with less than 10% vegetation.

50 Sand Desert Comprised of blowing sand dunes and flats, having patchy oasis seepage vegetation, with a vegetation cover less than 4% and a windy climate.

31 Bare rocks

Scree, cliffs, rocky outcrops, including active erosion, rocks and reef flats situated above the high-water mark, with less than 10% vegetation.

8 Bare Desert Comprised of ephemeral or sparse xerophytes with low cover vegetation (less than 4%) and an extreme drought climate.

32

Sparsely vegetated areas

Scattered vegetation, composed of gramineous and/or ligneous and semi-ligneous species, covers the surface area. Vegetated layer covers between 15 % and 50 % of the surface.

(21)

33b Burnt areas

Areas affected by recent fires, still mainly black.

34

Glaciers and perpetual snow

Land covered by glaciers or permanent snow fields.

12 Glacier Ice Ice sheet and other permanent ice.

35

Inland marshes

Non-forested areas of low-lying land flooded or floodable by fresh,

stagnant or circulating water. Covered by a specific low ligneous, semi-ligneous or herbaceous vegetation.

45 Marsh Wetland

Comprised of graminoid or herb with or without shrubs and trees. Having a mixed vegetation and a marine or cold-hot climate.

36 Peat bogs

Peatland consisting mainly of decomposed moss and vegetable matter. The accumulated sediments must contain at least 20% organic matter.

44 Mire, Bog, Fen

Comprised of sphagnum or graminoid with dwarfshrub, having evergreen or cold-deciduous vegetation and a cold climate.

37 Salt marshes

Vegetated low-lying areas, above the high-tide line, susceptible to flooding by sea water. Often in the process of filling in gradually being colonized by halophytic plants.

45 Marsh Wetland

Comprised of graminoid or herb with or without shrubs and trees with mixed vegetation and a marine or cold to hot climate.

38c Salines

Salt pans, active or in process of abandonment. May include sections of salt marsh exploited for the production of salt by evaporation.

12 Glacier Ice Ice sheet and other permanent ice.

39

Intertidal flats

Generally un-vegetated expenses of mud, sand or rock lying between high and low water mark.

15 Sea Water Saline surface water

40

Water courses

Natural or artificial water courses serving as water drainage channels. Includes canals.

14 Inland Water Fresh surface water

41

Water bodies

Natural or artificial stretches of fresh water.

14 Inland Water Fresh surface water

42

Coastal lagoons

Stretches of salt or brackish water in coastal areas, which are separated from the sea by a tongue of land or other similar topography.

43 Estuaries

The mouth of a river, within which the tide ebbs and flows.

44

Sea and ocean

Zones seaward of the lowest tide limit.

a : Because low sparse grassland is one of the few classes containing a low LAI, while still including seasonality, it is translated for areas with low amount of vegetation. Gao (2015) also uses this class for areas with little to none vegetation development.

b : Burnt areas are not indicated in Global Ecosystems. However, burnt areas are often rapidly revegetated by grasses and similar herbaceous plants.

(22)

c : Due to the fact that all non-vegetated land covers have zero vegetation, only the parameters roughness length and albedo differ for these types of land cover. Since salines have a high albedo just like ice sheets, I choose to translate the CLC salines class into the GE ice sheet class.

Appendix 2.2

Fuzzy based translation of Global Ecosystems to Corine Land Cover Classification inside the Netherlands:

Corine LC class 1st Olsen class Agreement Percentage 2nd Olsen class Agreement Percentage 3rd Olsen class Agreement Percentage Total percentage 1 1 90 - - - - 90 2 93 42 1 22 30 19 83 3 93 41 1 22 30 13 76 4 93 40 30 18 1 14 72 5 1 23 93 12 2 11 36 6 93 63 30 13 - - 76 7 93 61 30 10 - - 71 8 93 49 30 14 94 10 73 9 93 34 30 22 94 10 66 10 1 40 93 33 30 10 83 11 93 45 30 24 - - 69 12 93 59 30 14 94 13 86 16 93 34 56 25 30 25 75 18 30 70 93 20 - - 90 20 30 45 83 30 56 12 87 21 30 43 93 41 - - 84 23 93 48 30 24 - - 72 24 93 32 30 30 60 12 74 25 30 38 93 33 94 10 81 26 93 44 30 25 - - 69 27 93 47 30 17 94 12 76 30 93 62 - - - - 62 35 30 43 93 26 14 11 80 36 94 44 93 39 - - 83 37 15 46 93 26 55 11 83 39 15 - - - - 40 14 - - - - 41 14 - - - - 42 15 - - - - 43 15 - - - - 44 15 - - - -

Classification outside the Netherlands: Corine LC class 1st Olsen class Agreement Percentage 2nd Olsen Agreement Percentage 3rd Olsen Agreement Percentage Total percentage

(23)

class class 13 56 30 31 27 30 10 67 14 94 23 31 23 56 16 62 15 56 29 31 21 51 18 68 17 56 60 31 25 - - 85 19 56 53 31 17 51 12 82 22 56 45 31 23 46 16 84 28 31 31 56 26 46 13 70 29 47 71 - - - - 71 31 53 40 12 11 64 11 62 32 17 31 53 21 - - 52 33 31 49 46 23 - - 72 34 12 60 53 18 - - 78 38 15 37 51 32 8 12 81

The green highlighted CLC classes are classes containing only water, from which we are certain there is no large vegetation present as the CLC map has a higher resolution than the GE map. CLC classes highlighted in blue are classes of the classification for the fuzzy based translation was conducted on the agreement of CLC-GE classes in Scandinavia. The orange highlighted CLC classes were classified using the percentage agreement of the CLC-GE classes in Spain. The Alps were not used for the translation of the classes.

Appendix 2.3. Spatial extents of the compared regions. Netherlands Extent Degrees Top 54o North Bottom 50o North Left 3o East Right 8o East Alps Extent Degrees Top 48o North Bottom 45o North Left 5o East Right 12o East Spain Extent Degrees Top 48o North Bottom 39o North Left 17o West Right 1o West Scandinavia Extent Degrees Top 66o North Bottom 57o North Left 3o East Right 19o East Europe Extent Degrees Top 75o North Bottom 35o North Left 20o West Right 30o East