Drought affects all forest ecosystem properties. It is therefore essential to improve drought monitoring by understanding its spatial diversity with relation to tree species diversity, resilience, and resistance [1
]. Apart from light, water availability is the most important factor determining tree growth [2
]. Water shortage influences the biophysical properties of the vegetation, and the size of this effect depends on the drought duration. For example, water stress results in reduced photosynthesis rates and stomatal closure, thereby inhibiting the growth of all plant cell types [3
]. In forest management, it is therefore crucial to obtain up-to-date information about the actual tree habitat condition and its changes caused by drought.
A comparison of remote sensing images of vegetation indices from two periods can indicate a spatial distribution of changes. However, adequate change detection is a challenging issue. Several methods are available to not only determine differences between images, but also to designate an appropriate threshold to distinguish regions with and without changes [4
]. Of these methods, image differencing is the most popular one, where an image from one period is subtracted, pixel by pixel, from an image from another period [5
Research on the impact of drought on ecosystems, taking into account different variables, is important not only for the Central European areas examined in this study, but also for other moderate climate zone regions, including Scandinavian countries, in which adverse impacts of drought on species mortality [6
] and distribution [7
] of forest complexes have been observed. For example, a 10-year summer drought cycle in southern Finland halted the growth of a 12-year-old oak (Quercus robur
L.) stand, irrespective of the health class [8
]. It is therefore important to enhance the long-term resistance of such stands to water stress, especially as species ranges shift to northern Europe (Scandinavian countries) and to greater longitudes [9
] in the context of global climatic changes. It is expected that in the timeframe of 50–70 years, the mean drought duration in Europe will become considerably longer. Climate models indicate that at present, in northern Europe, the mean period without precipitation is increasing and that from 2070 to 2099, droughts will be markedly longer than in 1961 to 1990 [10
]. For example, the maximum duration of the period without precipitation in eastern Finland may increase from 27 to 39 days (as an average from four models) [11
]. Our understanding of the present and future drought lengths in Europe should be analyzed in the historical context. Droughts were more prolonged during the Medieval Warm Period (10th–12th century). The occurrence of megadroughts in north-central Europe was similar in timing, duration, and relative intensity to droughts in the North America [12
]. The most extreme drought in southern England, but also in Central Europe and Italy, was noticed in 1921 [13
]. Other periods of historical droughts were noticed in 1540 (widespread occurrence of moderate to extreme droughts in central Europe), 1616 (including “dried-up rivers”, precipitation deficit, and excessive warmth), 1741 (drought over Ireland contributed to the severity of the “Irish famine”), and the great drought of 1893 over the British Isles and also continental Europe [14
Also, research on drought response differences between mixed and pure stands has been done during the past decade. Pine–beech mixed stands were indicated to be an effective stand model for climate change in drought-prone sites in the Mediterranean region [15
]. Species complementarity may also enhance growth in mixed forests, but competition effects between tree species may override complementarity advantages at the drought-prone sites [16
]. Research on temporal shifts between competition and facilitation in mixed forests of beech–oak and beech–spruce in Central Europe showed that the promotion of mixed stands in forest management can improve forest resilience in terms of growth in the face of climate change [17
], but this type of forestry should consider the site’s moisture and fertility, as well as climatic conditions [18
]. If we assume that water is one of many resources necessary for forest growth, research done during the last two decades has pointed out that species mixing frequently improves resource supply, uptake, use efficiency, and, as a result, also tree and stand growth [19
]. If two species differ in their strategies of space occupation, complementarity and reduction in competition for resources in mixed versus pure stands is possible [22
]. This effect may increase when one species exerts a positive effect and facilitates the other [23
]. It is important for research studies to take into account the dominant species of the forest stand(s) being considered. In our case, four species may affect the response for drought in 2015.
Drought tolerance varies among forest tree species. In mixed temperate forests, Scotch pine (Pinus sylvestris
L.) shows the ability to tolerate long-term and short-term periods of drought, which is related to the morphology, phenology, and adaptability of this species [24
]. Q. robur
demonstrates less drought tolerance abilities. In comparison with other oak species in the temperate climate zone, this species has higher water requirements and is more vulnerable to water stress [25
]. On the other hand, Q. robur
can also be found on sites with low water availability, for example, on sandy, nutrient-poor soils [27
] or on plateaus and the exposed slopes of well-drained limestone hills [28
]. In such habitats, pedunculate oak (Q. robur
) shows a high drought tolerance, which ensures its survival, and is important for the acclimation of Central European oak species to drought stress [29
]. Data from previously established literature indicate that pedunculate oak copes better with drought than red oak Quercus rubra
L. (Q. rubra
]. The increased drought tolerance of oak habitats that are generally lacking in diversity may have been caused by the coexistence of pedunculate oak with an admixture of red oak, which has adapted to the limited water resources found in the habitats that it often appears. Despite the fact that both Q. rubra
and Q. robur
are considered to be highly vulnerable to water stress, they can occupy habitats with varied water availability, although Q. rubra
can also occur in extremely dry habitats [30
]. The most drought sensitive among interested trees seems to be Betula pendula
Roth (B. pendula
) species. Previous literature indicates that birches are less drought tolerant than Q. robur
]. Under natural conditions, B. pendula
(silver birch) often resides in cool and wet regions, including peat-bogs, stream and lake banks, cool and wet forests, and the slopes of cool bays [35
]. For this species, water deficit may cause significant stress and is dangerous for both adult individuals and nursery seedlings [37
This article analyzes the relationship between the stand response to short-term drought and the species diversity measured by remote sensing methods. It is assumed that species diversity is an important element influencing stand resilience to different stressors [38
]. Variations in the water regime and the related water shortages caused by reduced precipitation levels are some of the most important factors that contribute to the development of fungal diseases of trees [39
]. Fungal infections, such as the one caused by the genus Phytophthora
(the HESOFF Life + project, European Commision grant no. LIFE11 ENV/PL/000459), or the Dutch elm disease caused by invasive fungal species [41
] (ELMIAS Life + project, European Commision grant no. LIFE12 NAT/SE/001139), pose some of the greatest threats to natural habitats with ecological and economic values. For this reason, it is important to explore the specific behavior of individual habitats in response to a drought-affected season.
Forest stands selected as HESOFF project test areas were affected by Phytophthora
pathogen. The aim of the study was to estimate the influence of phosphite treatments on the health of the affected forests. In situ and remote sensing measurements were conducted between 2012 and 2015. The results of the measurements, carried out by traditional methods, unambiguously demonstrated the occurrence of drought in 2015. The effects of agricultural drought were the greatest in the Wielkopolska region, where its duration exceeded 100 days. The soil drought was also the most intensive there compared with the whole country and the soil water deficit lasted for more than 30 days. The climatic water balance showed that the annual water deficit in 2015 in Wielkopolska—the macroregion where our test site is located—exceeded 100 mm [43
In the summer of 2016, field surveys were organized. The residents of the surrounding lands confirmed exceptional water shortages in their wells in the previous season (on average, 2.5 times lower water levels). The drought impact affected surveys of the health of oak stands that were conducted every month of the growing season and this was a reason that we decided to estimate this impact. Consequently, vegetation seasons in 2014 and 2015 were selected for this study.
To analyze the stand response to drought, 15 forest complexes near Leszno, Poland, were designated and, using remote sensing techniques, their species diversity and the changes in vegetation indices between 2014 and 2015 were measured. Subsequently, taking into account the species diversity index and other habitat variables, the impacts of drought on the condition of the examined stands were analyzed.
Based on our results, we can draw conclusions about the impact of drought on stand resistance, assessed via the following factors: habitat conditions, species diversity, and the dominant species in the stand. The highest correlations between the differential indices and the Shannon-Wiener index were observed for GNDVI (+0.74), BNDVI (+0.68), and SAVI (+0.68).
We found no relationship between the differential indices and the indices demonstrating tree and understory coverage. However, a statistically significant relationship was found for the variables characterizing the dominant species in a given stand for all analyzed ΔVIs. The resistance to drought increases with decreasing diameter at breast height (DBH), height, and abundance.
4.1. Species Diversity
Divisions MD50 and SO27 were characterized by the highest resistance to drought. These habitats had the youngest stands and thus the lowest DBH, age, and abundance. At the same time, division MD50 showed the highest species diversity index (88.24%), while SO27 had the largest share of Scotch pine. It should be noted that Nasiłowska et al. (2017) [49
] indicated both species diversity and the share of Scotch pine within a division as factors enhancing the response to drought. In turn, the values of indices decreasing within the ranges (−0.79 to −0.99), (−0.67 to −0.81), and (−0.68 to −0.87) could be seen for divisions DBB 126, DBB 129, and DBB 96, evidencing a negative trend in the development of the local vegetation. These were the divisions with the lowest species diversity index values (all below 9.0%) and Q. robur
as the dominant species.
When comparing the indices H of all divisions with the corresponding differential indices, a mean correlation coefficient of +0.66 (±0.042) was obtained, confirming a decisive effect of H parameter on the stand response to drought. It should be pointed out that the study on the impact of species diversity on the stand response to drought could not consider the simultaneous effects of factors such as forest type or dominant species, and it was only a simultaneous analysis of the index H and the other habitat variables that ensured a full description of the reasons for the behavior of stands under stress conditions. However, our results also show exceptions to this general trend.
Two divisions with a low diversity index, i.e., DBB 130a (H = 0.064) and DBB 139 (H = 0.076), demonstrated higher drought resistance than a division with higher diversity, i.e., DBB 96 (H = 0.26). This could have been caused by a poor understory in division DBB 96 and a small percentage share of additional species in the oak stand (<7.0%). This example shows that one must not interpret the index H uncritically, as it is sensitive to the number of scarce species in a stand.
4.2. Dominant Species
The percentage share of the determined tree species was significantly (p-value < 0.05) correlated with ΔVIs for most indices only for Q. robur (−0.63 ± 0.04) and B. pendula (+0.55 ± 0.02), as well as for all indices for P. sylvestris (+0.63 ± 0.096). This result indicates that the resistance of forest stands to drought increased as the percentage share of Q. robur diminished and the shares of B. pendula and P. sylvestris increased.
Two notable observations in the literature regarding Q. robur
may be confirmed by our results. Q. robur
monocultures are less resistant to drought than Q. rubra
and P. sylvestris
monocultures. Oak stands with Q. robur
share >98% had mean ΔVIs below −0.46, for one Q. rubra
monoculture the ΔVI was equal to +0.11 and for two P. sylvestris
monocultures the ΔVIs were higher than +0.66. Very interesting is the influence of soil type on the almost pure oak stands. The lowest values of the differential indices, just below the interpretation threshold (−0.99 for the ΔBNDVI), were found in DB126, a 126-year-old homogenous (H
= 49.0%) oak stand. Among the analyzed oak divisions with low diversity, it was the only one situated on gley soils. In non-drought periods, these soils favor the growth of wet forests and cause the development of more shallow root systems as a result of gleying [92
], thus creating ecosystems that are more vulnerable to water stress. During drought periods, these soils are characterized by an exceptional water deficit, leading to water stress in plants [93
The second phenomenon that was observed on our test site was reducing foliage in favor of small roots, which is probably part of the process of the acclimation of Central European oak species to drought stress. This is confirmed by the highest significance of the correlation between the percentage share of Q. robur and the NDMI differential index.
Comparison between Q. robur monocultures and mixed stands show large difference between corresponding ΔVIs: below −0.46 for monocultures and near 0.00 or +.0.57 for mixed stands. Mixed stands with relative small admixtures of Q. rubra or P. sylvestris seem to be most resistant to drought (>0.57). For an oak–pine mixed stand, but with inverse proportions, the ΔVI was average (+0.04).
In areas DBB 130b and DBB 130c, where Q. robur
dominated with a share of Q. rubra
(17.7% and 11.16%, respectively), a higher drought tolerance was found than in areas with higher biodiversity, such as DBB 96. Despite the fact that, as indicated in the literature, it is a species that typically avoids drought [88
], our research demonstrated that, when found alongside the addition of Q. rubra
, Q. robur
appeared to have positive effects on the drought resistance of habitats. This study found that areas overgrown by pedunculate oak, with an admixture of Q. rubra
(DBB 130b and DBB 130c), had a higher drought tolerance than areas without Q. rubra
. This might also be related to the habitat requirements of both oak species. Q. robur
is a species with high soil requirements, as it prefers rich soils, such as brown and lessive soils and black earths, whereas Q. rubra
develops well on poorer soils, such as clayey sands and sandy soils [94
]. On the Krotoszyn Plateau, the areas under study were dominated by low-fertile gley soils, to which red oak was considerably better adapted than pedunculate oak. Vivin et al. (1993) found that the richness/fertilization of a habitat had a significant impact on the growth and development of Q. robur
and Q. rubra
under water deficit conditions [95
]. Under the impact of a prolonged drought, the mortality of young Q. robur
specimens was higher than that of Q. rubra
and, on this basis, the authors noted that Q. rubra
was a species with a higher tolerance to drought stress than Q. robur
. They also found that both species coped better with drought stress if mineral fertilizers were provided, which was probably related to the osmoregulation mechanism in the case of a low water potential [95
Our results confirmed that P. sylvestris monocultures are drought resistant. The worse difference index observed for the pine stand with oak admixture (21.65%) should be the subject of future research.
These observations will help add to the data from previous literature, since an assessment of the vulnerability of trees to drought is of major importance for improving the forecasts of the dieback of forests and species under the impact of climate change. This topic is very complex, because favorable climatic periods that allow for abundant tree growth may result in structural overshoot of aboveground tree biomass and when water and temperature stress occurs premature leaf senescence and partial canopy dieback to whole-tree mortality may occur as a consequence, thus reducing canopy leaf area during the stress [96
]. On the other hand, historical data suggests that dry forests are experiencing increasing drought-induced mortality, but this observation is not absolutely true for all of the forest types, and the spatial variability of this phenomenon is very large [97
]. Certain forest tree species demonstrate substantially different needs in terms of water supply. In particular, this is the case with oaks [98
]. For example, a direct comparison between Quercus petraea
(Matt.) Liebl. (Q. petraea
), Q. robur
, and Q. rubra
has shown that Q. petraea
is more resistant to drought than the other two species [95
4.3. Monitoring Water Stress Using Differential Indices
Among the analyzed indices, only ΔGNDVI accommodated the green channel and is thus the only index sensitive to the chlorophyll content in cellular structures. The chlorophyll content indicates the condition of the vegetation and its photosynthetic potential; the higher the content, the more energy can be absorbed and used for plant development. A reduced share of this pigment in cellular structures can be seen in the lower reflectance in green light. As indicated by the literature, its value should be independent of changes in pigments other than chlorophyll [57
], which is responsible for light absorption; its efficiency depends on stressors. The ΔGNDVI, which shows the drought impact, correlates most strongly with biodiversity; we therefore conclude that this green pigment and its quantity in the habitat determines and differentiates the spectral response to a greater extent than other factors do.
In addition, this index should be independent of the impact of the soil visible between leaves or the impact of the atmosphere, which is particularly important for forest areas [57
]. Moreover, its relationship with the diversity index H
is stronger than that for classical indices, which are applied on a standard basis to eliminate the impact of the substrate (Table 4
), such as SAVI [56
] and MSAVI [52
Considering all these factors, this index is increasingly used to investigate stress in vegetation [99
]. Our results indicate that it is useful for forests and suggest that it should be used more frequently as an alternative to NDVI.
4.3.1. Quercus sp. L.
The ΔMSAVI differential index is distinguished by its strongest correlation (displaying the highest statistical significance) with the percentage share of the dominant species Q. robur
. It has been designed to enhance its sensitivity to changes in the condition of vegetation and, at the same time, to eliminate soil impact, making it less dependent on the quantity of the biomass than the original SAVI. This factor clearly affects the drought response of oak. However, when approached in general terms (Table 3
), the variation of both ΔMSAVI and ΔSAVI is similar (with a standard deviation of 0.81). This indicates the greater usefulness of the ΔMSAVI index for investigating forest environments dominated by Quercus
4.3.2. Pinus sp. L.
The percentage share of the species P. sylvestris
showed the most significant statistical correlation with the ΔNDMI differential index. This coniferous tree species occurred in several analyzed habitats and dominated three of them, i.e., SO100, SO105, and SO27 (Table 1
). Among the three analyzed indices based on mid-infrared, it demonstrates by far the strongest relationship with drought. Therefore, when comparing the results in Table 3
, it can be considered that trends of change shown by ΔNBR and ΔNBR2 are less reliable than those calculated using ΔNDMI. Research performed to date has indicated the usefulness of indices based on mid-infrared for investigating coniferous forests [102
]. The present study demonstrates that the index calculated on the basis of bands 5 (880 nm) and 6 (1610 nm) more adequately shows the condition of a stand with a large share of pine than the other two indices (NBR and NBR2), considering the second reflectance maximum in mid-infrared of 2200 nm (band 7). It is possible to better explore the behavior of these three indices and their usefulness for investigating drought impacts for a larger number of divisions including this species, whereas this study considers only three of them.