Sampled plantations: Based on the information provided by the Oficina Nacional Forestal (National Forestry Office) and the Fondo Nacional de Financia-miento Forestal (National Forestry Financing Fund) of the government of Costa Rica regarding the disease in plantations of Gmelina arborea, sampling was conducted in 16 sites between 2 and 6 years old, distributed in various regions of the country (Fig. 1). Two of these plantations showed no symptoms of the disease (sites 2 and 4), whereas diseased trees were evident in the remaining 14 plantations.

Sampling within the plantations: A 14 m radius circular plot (615 m ²) was established. Trees in that area were counted and the projection per hectare was performed (stand density). In addition, the diameter at 1.3 m height (DBH) and the dominant height (h) represented by the three highest trees within the plot, were measured. Healthy trees were identified within the plot, as were trees showing any symptom of the disease, the progress of which was evaluated. According to Arguedas (2004) and Salas (2015) the symptoms of the disease in G. arborea are leafless crown or foliar withering (Fig. 2a), canker occurrence (Fig. 2b), sprout occurrence (Fig. 2c) and black exudation in the stem (Fig. 2d). Then, four classes were defined (Table 1).

Determination of incidence and severity: The incidence of the disease and the average index of severity were determined in each plot, following the methodology proposed by Araúz (1998) and Salas (2015).

Soil evaluation: Ten sampling points were ran-domly selected within each assessment plot. Two soil samples were obtained at each point, employing a manual drill from depths of 0 to 10 cm (profile A) and 10 to 20 cm (profile B). The ten soil portions were mixed, and then the samples per layer were divided into four equal portions and one portion was randomly selected. The soil was characterized according to order and texture. The samples were measured for their pH in water at a 1:2 ratio, and for their proportions of Ca, Mg and K; the effective cation exchange capacity (ECEC), which corresponds to the sum of pH+Ca+Mg+K; the percent base saturation calculated as (pH/ECEC)×100; and the amounts of P, Zn, Cu, Fe and Mn. In addition, the bulk density (g/cm ³), clay (< 2 µm), silt (<2-50 µm) and sand (>50 µm) proportions of the soil in the sampled plantations were determined. For this purpose, soil samples of 80 cm ³ were gathered at two depths: from 5 cm to 10 cm and from 10 cm to 15 cm; these samples were dried at 110 °C for 24 hours to reach dry weight. Then, the density, which corresponds to the ratio of the dry weight (g) and volume of the cylinder (cm ³), was calculated.

Climatic characterization: First, the climate in the regions where plantations are located was characterized following Holdridge’s world life zones classification (Holdridge, 1967). Then, the geographic coordinates and elevation (masl) were obtained for each sampled plantation. This was performed by means of the Global Positioning System (Garmin GPS). In parallel, the permanent meteorological stations of the Instituto Meteorológico Nacional closest to the sampled plantations were located. Information was gathered about average temperature and monthly precipitation within a period of 10 years (2004-2014).

A vertical temperature gradient of 0.6 °C per 100 m and the difference in elevation between each plantation based on the elevation of each meteorological station were used to adjust the temperature of each plantation. To adjust the precipitation, an interpolation of the data was performed using the QGis® Analysis of Spatial Interpolation.

Later, the hydric balance was calculated for each plot following the Thornthwaite & Mather (1957) method. Data on the annual potential evapotranspiration, actual evapotranspiration and water deficit in the soil were gathered by means of Thornthwaite’s table for soil moisture retention, according to the effective depth and soil texture in each plot. Additionally, Holdridge’s annual potential evapotranspiration was calculated as (annual biotemperature ∗ 0.58), and the ratio of potential evapotranspiration and average annual temperature equal to Holdridge’s annual biotemperature (Holdridge, 1967) were also calculated.

Fungal identification: Wood with symptoms is presents a dark color (Fig. 3a-b), whilst healthy wood does not (Fig. 3c). Trees affected by the Gmelina death disease were cut in six plantations (Fig. 3d). Cross sections were extracted from the stems one meter from the base of the tree (Fig. 3e-g). During cutting, black areas with a strong odor were observed on the stems, perhaps because of a tree pathogen (Fig. 3a-b). A cross section of each tree was cut at breast height for later identification of the pathogen. The cross sections were washed with distilled water and then three small sections from the heartwood were cut from the periphery of the affected areas (black, strong smell), with the help of a sterile scalpel (Fig. 3h-i).

The three samples per tree were placed on separate Petri dishes containing potato-dextrose-agar (PDA; Difco laboratories, Detroit) and incubated for 22 days at 25 °C under continuous darkness. Several colonies developed during that period. Mycelia were collected from the periphery of the colonies and again placed on a Petri dish to grow into PDA. Then, fungal purification was again performed by hyphal tip transfer to Petri dishes with PDA, which were stored in the dark at 25 °C for 22 days. The mycelial samples of each purified culture for phylogenetic analysis were identified by extracting the DNA employing the Murray & Thompson´s method (White et al., 1990). Then, a PCR was conducted and the product was sequenced for one sample for each Petri dish. A total of 18 sequences were obtained (3 trees × 6 plantation or sites × 1 Petri dish). Each sample was added to a 50 µL solution containing 25 pmol of each primer. The primers used were ITS4 and ITS5. PCR products were sequenced by Macrogen. To analyze the obtained sequences, the NCBI Blast tool was employed for comparison with the GenBank database ( http://www.ncbi.nlm.nih.gov/genbank/) to identify the isolated fungi.

Root morphology evaluation: Two trees were chosen in each of the plantations where the disease was present, one visibly healthy and the other infected. For each selected tree, two points were located for sampling at the edge of the dripping area (the edge of the crown). At these points, 20 cm ³ of soil with roots was extracted at ground level. The samples were washed with distilled water to eliminate the organic matter and then placed in glass containers with ethanol (70%) to prevent deterioration. Then, two representatives 5 cm long subsamples were extracted from the terminal parts of the absorption roots. Using an Epson Expression 1680 scanner, a digital image was captured of the roots and was processed afterwards with WinRhizo Basic V 2005-b® software. The following variables resulted from the image processing: root total length, root average diameter, number of tips, number of bifurcations and number of crossings.

Evaluation of the wood properties: For the present study, we followed the methods proposed by Jeremic et al. (2004) and Moya et al. (2009), where 9 trees with symptoms of the disease were selected from a 3-year-old plantation in Penshurst, Limón, Costa Rica. The average annual temperature at this location is 26 °C and the annual precipitation is 3078 mm. The average DBH of the trees was 21 cm and the average total height was 19 m. One-meter-long logs were cut from each selected tree, until reaching the commercial height (13 cm diameter). A cross section approximately 4 cm thick was extracted at the end of each log. Five one-meter-long logs with stained and unstained regions were chosen. Then, the samples for determining wood properties were extracted first from zones affected by the disease, corresponding to wood infected by Gmelina death disease. Later, samples of unaffected wood were extracted from the same log or cross section, at the same distance from the heartwood. The properties evaluated were as follows:

a) Physical properties: Green wood density (green weight/green volume), specific gravity according to ASTM D-143 (ASTM, 2003a) and moisture content according to ASTM D-4442 (ASTM, 2003b). A total of 30 samples of dimensions 2.5 cm × 2.5 cm × 2.5 cm were extracted from the discs of infected and non-infected wood. The infected and non-infected wood samples were collected at the same distance from the heartwood. Each sample was measured for its green weight and green volume (their condition at the moment of extraction of the sample) and then placed in an oven for 24 hours at 105 °C to determine its final weight. This information was used to determine the physical parameters.

b) Mechanical properties: From the 2.7 cm thick boards obtained from the 1 m long logs, approx. 30 samples of dimensions 2.7 cm × 2.7 cm × 1 m were obtained and dried at 22 °C and 66% relative humidity to reach an equilibrium moisture content of 12%. The mechanical properties evaluated were maximum stress in compression, modulus of rupture and modulus of elasticity in static bending. All tests were performed following the ASTM D-143 standard (ASTM, 2003a).

c) Resistance to fungal decay: 240 wood samples measuring 2.5 cm × 2.5 cm × 2.5 cm were cut and divided into three groups: 80 with infected sapwood, 80 with unstained sapwood and 80 with infected heartwood. The samples were inoculated by two strains of the fungi, Trametes versicolor (white decay) and Lenzites acuta (brown decay), following the procedure set in the ASTM-D 2017-81 standard (ASTM, 2003c). This test yields the percentage of weight loss of the samples after 16 weeks of exposure to fungal attack.

d) Inorganic ash composition of wood: Chips of wood were collected from healthy and infected samples from the cross sections of the sampled trees and then ground and sieved through 40- and 60-sized meshes. Ash analysis was performed using the methods of Sparks (1996) and included determination of the mass of N, P, Ca, Mg, K and S, whose concentrations were reported as percentages, and Fe, Mn, Cu, Zn and B, whose concentrations were reported as milligrams of inorganic element per gram of ash.

e) Wood color: Color was measured in the samples to measure the mechanical properties of the infected and uninfected wood. These samples were conditioned to 12% moisture content and were evaluated in the tangential face. To measure color, the standardized color measuring system CIE L∗a∗b∗ was used with a miniScan XE Plus spectrophotometer, following the procedure described by Moya et al. (2012).

A Pearson correlation matrix was used for determining the most correlated incidence and severity for further analysis. Next, incidence and severity were correlated with the tree morphology (DBH and h), stand density, tree age, soil and climatic characteristics. A forward stepwise analysis was applied to establish the effect that tree morphology, tree age, climate characteristics and soil characteristics in the profile from 10-20 cm had on the percentage of incidence and severity of the disease and on the various damage classes. The analysis of variance, ANOVA, was also performed to determine the level of significance of symptomatic and asymptomatic wood; it was applied to characteristics of the roots and the physical and mechanical characteristics, content of macro- and microelements, loss of weight during fungal attack and different color parameters. Afterwards, the presence of significant differences among the media of these parameters of normal wood and symptomatic wood was verified by means of the Tukey test ( p<0.05). The statistical software SAS 8.0 was used to perform the statistical analysis of the data.

The incidence of the disease varied from 0% to 58% of the total number of trees in the plantation. When sorted the 16 sampled plantations from highest to lowest incidence and severity, three different inflection points could be observed. According to this, four groups were established: (i) plantations where incidence was equal to 0 (plantations 2 and 4); (ii) a second group with values of incidence from 3% to 4% (plantations 13, 7 and 9); (iii) a third group with values between 17% and 35% (plantations 1, 8, 6, 15, 12, 16, 10 and 11) and (iv) a fourth group with values of incidence between 48% and 58% (plantations 3, 5 and 14) (Fig. 4a). For severity, it ranged between 0% and 28% (Fig. 2b) and four groups were determined as well: (i) the first group includes healthy plantations (plantations 2 and 4), (ii) the second group showed severities between 1% and 4% (plantations 13, 7, 9, 8 and 6), (iii) the third group showed severities between 10% and 13% (plantations 1, 12, 16, 3, 15 and 10) and (iv) the last group showed values between 21% and 28% (plantations 5, 11 and 14) (Fig. 4b).

The percentage of the damage class A (wilting, exudation and small sprouting) was higher in plantations 1, 3, 6, 8, 12 and 16, while the percentage of damage class B (wilting, small leaves, 50% loss of foliage, canker or swelling, exudation and sprouting) was higher in plantations 5 and 10. For damage class C, the percentage was higher in plantations 14 and 15. Finally, plantation 11 showed the highest number of trees with damage class D (Table 2).

The DNA sequence analysis of the six types of wood from samples collected from symptomatic trees revealed the presence of three species of fungi: Fusarium solani was present in 3 sites (2, 3 and 4) showing a BLAST match of 99% in the GenBank database (AF178407.1, KF255997.1 and AF178407.1, respectively). Rhizomucor variabilis was present in sites 5 and 6 with a BLAST match of 99% with HM623314.1 of GenBank database. Finally, Chaetomella raphigera, with a 98% BLAST match with AY487076.1 of the GenBank database, was present only in the xylem of the tree in site 1.

The amount of clay in the soil layer between 10 cm and 20 cm accounted for 66% of the variation of incidence of the disease, followed by K in the same layer explaining 22% of the variation (Table 3). The amount of clay was positively correlated with the incidence of the disease (Fig. 5a). In the case of K, contrariwise, the correlation with the incidence was negative (Fig. 5b). However, this relation was influenced by two points, 0.5 and 0.6 in K content. If these two points were removed, then the relationship would become non-significant ( R ²=0.02). Altogether, clay and K content explained 87% of the total variation of the incidence (Table 3).

With respect to severity, it was first related to stand density, which explained 33% of the variation, after which the amount of clay in segment 10-20 cm explained 27% of the variation (Table 3). Once more, K appeared in the 10-20 cm profile and explained 21% of the variation. All these variables explained 81% of the variation of severity (Table 3). The stand density was negatively correlated with severity (Fig. 5c); the amount of clay in 10-20 cm profile was again positively correlated with severity (Fig. 5d).

The average number of tips, forks and crossings and the root length was statistically lower in in-fected trees than in healthy trees. Meanwhile, no significant difference was observed between both conditions of sampled trees regarding root diameter (Fig. 6).

Gme-lina arborea trees were evident in both the sapwood and the heartwood (Fig. 3). The affected region was differentiated from the surrounding tissue by its darker color and occasional sap secretion (Fig. 3). Fermentation was another characteristic of these dark regions as it causes a rancid to fetid rumen-like odor. Affected wood showed light (L∗) tonality, statistically lower than unaffected wood, in both green and dry wood (Fig. 7a). A higher tonality was also observed in redness (a*), although this was statistically higher in the dry condition (Fig. 7b) compared to the wet condition. Finally, less yellowish tonality (b∗) was observed in green wood in relation to normal wood (Fig. 7c).

Wood properties: it was observed that affected wood presented a statistically higher value of density in green condition, higher moisture content, less resistance to compression stress, less modulus of elasticity and module rupture in the bending test compared to unaffected wood (Table 4). The specify gravity was not statically different in affected and unaffected wood (Table 4). Regarding the durability of the wood to white-decay ( T. versicolor) and brown-decay ( L. acuta) fungal attack, there was no statistical evidence that symptomatic wood was less durable than normal wood (Fig. 7d-e).

On the other hand, the evaluation of micro- and macronutrients of wood ash showed that the amount of Mg and Fe is statistically higher and quantities of Ca, K and Zn are statistically lower in symptomatic wood compared to asymptomatic wood (Table 4). The amounts of remaining nutrients were not statistically different between symptomatic and asymptomatic wood. For other chemical characteristics, the pH and electrical conductivity were statistically lower in wood from trees with “Gmelina death disease” than in normal wood (Table 4).

Gmelina plantations have been established in Costa Rica in sites with high stand densities and high contents of clay in the profile 10 to 20 cm deep, which are inappropriate conditions that increase susceptibility to Gmelina death disease. Different solutions may be implemented for this problem. One solution is related to the selection of suitable sites where the amount of clay is below 40%, to allow the soil to conveniently drain the levels of rainfall characteristic of reforestation sites in Costa Rica. The second solution refers to managing infected trees at early ages to grow healthy trees capable of combating pathogens.

In summary, the quantity of clay present in the 10-20 cm layer offers conditions for the development of the Gmelina death disease. Lack of management of plantations also promotes the development of the disease (high severity) though increased stand density. The characteristics of climate had no effect on the disease. Although two climatic variables, Holdridge’s potential annual evapotranspiration (PAEH) and Thornthwaite’s potential evapotranspiration (PET), showed a relationship with the incidence in the trees and symptoms of the disease, it was not statistically significant. The visual quality of the wood was affected by Gmelina death disease. The affected area turned black, and mechanical resistance was also reduced. The natural durability of the wood, however, was not affected. There were changes in ash composition, specifically in the levels of Mg, Fe, Ca, K and Zn. An appropriate selection of sites, with the amount of clay below 40%, allows adequate growing of trees in commercial plantations in Costa Rica, and adequate management at early ages allows to grow healthy trees capable of combating pathogens.