a case study on global warming

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List

Climate change effects on desert ecosystems: A case study on the keystone species of the Namib Desert Welwitschia mirabilis

Pierluigi bombi.

1 Institute of Research on Terrestrial Ecosystems, National Research Council, Monterotondo, Rome, Italy

Daniele Salvi

2 Department of Health, Life and Environmental Sciences, University of L’Aquila, Coppito, L’Aquila, Italy

Titus Shuuya

3 Gobabeb Namib Research Institute, Walvis Bay, Namibia

Leonardo Vignoli

4 Department of Science, University of Roma Tre, Rome, Italy

Theo Wassenaar

5 Department of Agriculture and Natural Resources Sciences, Namibia University of Science and Technology, Windhoek, Namibia

Associated Data

Data cannot be shared publicly because they regard endangered populations of a species that is potentially sensitive to illegal collections. Data are available from the CNR-IRET Institutional Data Access (contact via ti.rnc.teri@aireterges ) for researchers who meet the criteria for access to confidential data.

Deserts have been predicted to be one of the most responsive ecosystems to global climate change. In this study, we examine the spatial and demographic response of a keystone endemic plant of the Namib Desert ( Welwitschia mirabilis ), for which displacement and reduction of suitable climate has been foreseen under future conditions. The main aim is to assess the association between ongoing climate change and geographical patterns of welwitschia health, reproductive status, and size. We collected data on welwitschia distribution, health condition, reproductive status, and plant size in northern Namibia. We used ecological niche models to predict the expected geographic shift of suitability under climate change scenarios. For each variable, we compared our field measurements with the expected ongoing change in climate suitability. Finally, we tested the presence of simple geographical gradients in the observed patterns. The historically realized thermal niche of welwitschia will be almost completely unavailable in the next 30 years in northern Namibia. Expected reductions of climatic suitability in our study sites were strongly associated with indicators of negative population conditions, namely lower plant health, reduced recruitment and increased adult mortality. Population condition does not follow simple latitudinal or altitudinal gradients. The observed pattern of population traits is consistent with climate change trends and projections. This makes welwitschia a suitable bioindicator (i.e. a ‘sentinel’) for climate change effect in the Namib Desert ecosystems. Our spatially explicit approach, combining suitability modeling with geographic combinations of population conditions measured in the field, could be extensively adopted to identify sentinel species, and detect population responses to climate change in other regions and ecosystems.

Introduction

Climate change is one of the strongest threats for ecosystems worldwide. Variations in the density of species, range changes, and extinction events have been documented at local and global level [ 1 – 3 ]. Furthermore, changes in species diversity, ecosystem functioning, and service provision are expected for the future as a consequence of climatic pressures on natural populations [ 4 – 6 ]. In Africa, deep impacts by climate change have been forecasted for animals [ 7 – 9 ], plants [ 10 – 12 ], and biodiversity in general [ 4 , 13 ].

Desert ecosystems are predicted to be one of the most vulnerable ecosystems to global climate change [ 14 – 16 ]. Rising temperature, decreasing rainfall, and increasing atmospheric CO 2 , are expected to strongly affect the structure and function of desert ecosystems [ 15 , 17 ]. Desert-adapted species are vulnerable to climate change [ 18 ] and among them endemic plant species are particularly susceptible to the loss of suitable habitat [ 19 ]. The negative effect of climate change on desert plants has been demonstrated worldwide [ 20 ].

In the arid regions of southern Africa, projections of climate change impacts on species persistence indicate a high vulnerability of endemic plant diversity to climate change [ 19 , 21 ]. For example, climate-linked increases of mortality have been observed for the quiver tree ( Aloidendron dichotomum Klopper & Gideon 2013; [ 22 ], and a potential decrease of climatic suitability was recently pointed out by Bombi [ 23 ] for welwitschia (common name for Welwitschia mirabilis Hooker 1863). Welwitschia mirabilis is regarded as a living fossil, representing an ancient lineage of gymnosperm plants, and it is recognized as a symbol of the Namib Desert biodiversity. This species has a peculiar morphology, being a long-living dwarf tree with only two leaves growing throughout its entire life [ 24 ]. This is also a keystone species for the Namib desert ecosystems, where it provides food, water, and refuge for many animal species, including mammals, reptiles, and insects [ 25 , 26 ]. Welwitschia mirabilis is endemic to the central and northern Namib Desert, ranging between the Kuiseb River in Namibia and the Nicolau River, north of Namibe, in Angola [ 27 , 28 ]. In this area, welwitschia plants occur in four separated sub-ranges, three in western Namibia ( Fig 1 ) [ 29 ] and one in south-western Angola. Bombi [ 23 ] showed that populations living in the three Namibian subranges have experienced and will in the future face rather different climatic conditions. The same study predicted a significant reduction of climatic suitability in the northernmost Namibian subrange (which lies in a transition zone between desert and arid savanna) under current climate change scenarios. In particular, the ongoing rise of temperature can drive the local climate out of the realized niche for the northern populations, thus increasing their extinction risk [ 23 ]. Although these findings were potentially important for conservation planning, the study was based on low spatial resolution data (2.5 arcmin) available at the national scale, thus limiting its utility for targeting individual populations.

In the main map, the black polygon indicates the study area, the red polygons show the known species distribution, and the blue polygons represent the boundaries of the observed extent of occurrence in Northern Namibia. In the inset map the red circle indicates the general location of the study area in Africa. Map tiles by Stamen Design, under CC BY 3.0.

Many species respond to climate change by changing their distribution range [ 30 – 32 ]. These changes have been generally described as poleward and/or upward movements to track suitable temperature conditions along latitudinal and altitudinal gradients [ 33 – 35 ]. However, in many cases documented geographic patterns of response are complex and do not align with simple latitudinal and altitudinal shifts [ 36 ]. Indeed, the assumption of simple, uni-directional distribution shifts does not account for intricate interactions among temperature, precipitation, and species-specific tolerances and can lead to substantial underestimation of the effect of climate change on species distributions [ 37 ]. To overcome these drawbacks, one promising approach to quantify possible range changes is based on the comparison of the species-specific spatial pattern of climatic suitability variation (the expected responses), generated by predictive models, with the pattern of appropriate metrics of population conditions measured in the field (the observed responses) [ 38 ]. This approach can increase our ability to identify the effect of climate change on species dynamics.

The main aim of this study was to determine whether the observed geographic combination of population condition (plant health, reproductive status, and size) of welwitschia in northwestern Namibia is associated with ongoing climate change. Secondly, we tested if the same pattern follows a latitudinal or altitudinal gradient in agreement with the assumption of a poleward or upward range shift. More specifically, we wanted to first validate the projections of potential impacts of climate change on W . mirabilis with field-based data. Then, we wanted to assess whether the simple assumption of a poleward/upward range shift is suitable for detecting climate change effects. To do this, we compared the geographic combination of population conditions, measured in the field, with the expected pattern of response, estimated by ecological niche models. If climate change is affecting welwitschia populations, we expected the worst population condition in sites where climatic suitability is decreasing and the best where suitability is increasing. Moreover, if a poleward/upward range shift is the major response to climate change, we could expect a latitudinal or altitudinal trend in the observed patterns of response. Since potential divergent responses to climate change by intraspecific lineages have been observed before [ 39 ] and different realized niches were described for each distinct Namibian subrange [ 23 ], we focused on populations in the northern subrange and considered them as an independent ecological unit, with its own climatic niche and with its (sub)specific expected response. By testing our main and secondary hypotheses, we hope to inform the long-term conservation of W . mirabilis and further contribute to the scientific debate on the climate change impacts on biodiversity.

Materials and methods

Field data collection.

During May 2019, we carried out a field expedition in the northernmost Namibian subrange of W . mirabilis , as defined by the ’Digital Atlas of Namibia’ [ 29 ], in order to obtain information relevant for the species conservation. This study was authorized by the National Commission on Research, Science and Technology of Namibia (Research Permit RCIV00032018) and performed in public lands, managed by the Orupembe, Sanitatas, and Okondjombo Communal Conservancies. During the expedition, we spent 10 full days, in a team of six persons, searching for welwitschia plants across the northernmost Namibian subrange by (1) driving at low speed along the available tracks (more than 330 km) while recording the presence of plants in a ~30 m wide transect on each side of the vehicle, and (2) walking across potentially suitable habitats (more than 65 km) in ~60 m wide transects on each side, in both valley bottoms and hill slopes. Doing this, we explored comprehensively more than 65 km 2 (330 km x 30 m x 2 sides + 65 km x 60 m x 2 sides x 6 persons). The starting points and spatial extent of our walking transects were informed by the knowledge of our local team members, who have an intimate knowledge of the area. We are confident that the combination of local knowledge and systematic transects extending beyond the known range have allowed us to establish the extent and characteristics of the majority of this sub-range.

We collected detailed data on four categories of plant traits: plant location, health condition, reproductive status, and plant size. More specifically, we recorded the precise coordinates (using a handheld GPS) (1), the sex (2), and the presence/absence of cones (3) for almost all the individual plants we observed (just a few, unreachable plants were excluded). The health condition (4), the stem diameters (minimum (5) and maximum (6) along the two main axes of the stem), and the mean leaf length (along the curved trajectory of the leaves) (7) were measured in sites with a sufficient number of plants (> 25 for variable 4 and > 35 for variables 5–7) for reducing the effect of chance on categorical and numerical variables. In sites with more than 60 plants, we considered a random subset of ~60 plants. We ranked health condition on a four-point scale (dead, poor, average, good) based on leaf color (see S1 Fig ). Although this is a relatively coarse scale, the brightness of the green color and the ratio of red/brown to green together are a remarkably consistent and accurate indicator of good health condition as measured by photosynthesis efficiency [ 40 ]. The green color of the leaf is associated with the chlorophyll content and the photosynthetic efficiency of the tissues [ 41 , 42 ], which is influenced by environmental stress [ 43 , 44 ]. An estimate of health condition such as the above is both a direct reflection of the environmental (including climatic) stress that the plant experiences and an index of the likelihood that its resistance to parasites might be compromised [ 45 , 46 ]. We expected that changes in local climate will be visible in its leaf color as a quick proxy of plant health.

Observed pattern of response

For each welwitschia stand (defined a posteriori , through a GIS-based analysis, as a group of plants separated from the other groups by a distance larger than the intra-group mean distance), we calculated three categories of synthetic indicators of population response (derived from plant health, reproductive status, and size) from the field-measured data. For each stand, we calculated the proportion of plants that were dead or in poor, average, and good condition to the total number of plants in the stand. We also calculated the reproductive status (the proportion of plants in the stand that had cones) and the plant size (average stem major axis, stem minor axis, and leaf length). The presence/absence of cones were used as a proxy of population recruitment potential instead of other, more common methods (e.g. plant size) in order to gather a trend of the last few years (after 2000) in plants with an extraordinary low growth rate.

In order to test a previously proposed gradient in plant size, health condition, and reproductive status from hill slopes and valley bottoms [ 27 ], we compared these variables for plants growing in the drainage systems with those in steeper locations.

Expected pattern of response

We used a spatially explicit approach based on ecological niche modeling to estimate the geographic combination of plant response expected as a consequence of climate change. To do this, we defined our study area as a bounding box three times larger than the latitudinal and longitudinal extent of the previously known subrange of welwitschia in northern Namibia [ 29 ]. Inside this study area, we fitted models on 1000 pseudo-presence/absence points by using historical (1950–2000) climate data from the WorldClim databank, version 1.3 [ 47 ] at the spatial resolution of 30 arcsec (about 1 km) in the R -based [ 48 ] biomod2 Package [ 49 ]. In order to control the model-associated uncertainty, we adopted an ensemble forecasting approach [ 50 ]. In particular, we used Generalized Linear Models [ 51 ], Generalized Additive Models [ 52 ], Generalized Boosting Models [ 53 ], Classification Tree Analyses [ 54 ], Artificial Neural Network [ 55 ], and Random Forest [ 56 ] methods, which are widely used and recognized as robust methods.

Pseudo-presence/absence points were randomly generated across the study area and classified as presence or absence points based on their position inside or outside the species extent of occurrence, generated as a minimum convex polygon from our detailed distribution data. Doing this, we overcame the problems related to the small number of real sites of presence and to the spatial autocorrelation due to the non-homogeneous distribution of the real sites [ 57 ]. Multicollinearity among predictors was reduced by discarding those with variance inflation factor higher than five [ 58 ]. Thus, four variables were retained for modeling (i.e. Temperature Seasonality, Mean Temperature of the Driest 3-months period, Precipitation Seasonality, Precipitation of the Warmest 3-months period) (See S2 Fig ). Each independent model was projected into the study area under historical climatic conditions and three-fold cross-validated by calculating the true skill statistic (TSS) [ 59 ]. Finally, we generated a single consensus model of historical suitability. More specifically, we used the six individual models to predict a continuous value of suitability (between 0 and 1), so the final consensus suitability is the TSS-weighted average of six values for each 30 arcsec pixel. Suitability ranges between 0 (no suitability) and 1 (perfect suitability). Future climate suitability was predicted by projecting the models into future climatic conditions across the study area. We used Global Climate Models (GCMs) elaborated by 19 research centers within the Coordinated Modelling Intercomparison Project Phase 5 (CMIP5), which represented the basis for the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC-5thAR) [ 60 ]. GCMs were elaborated for the four Representative Concentration Pathways (RCP2.6, RCP4.5, RCP6.0, and RCP8.5) for 2050 (average for 2041–2060) and downscaled at the same spatial resolution as the historical climate data using WorldClim as baseline climate [ 61 ].All the 59 available scenarios of future (2050) climate from CMIP5 were utilized for projecting the models. We compared the historical climate data and the future climatic scenarios by calculating the Multivariate Environmental Similarity Surfaces (MESS) [ 62 ] in order to estimate the reliability of predictions. The MESS calculation estimates the extent to which the predictor variables in the study area are similar to the conditions experienced by the species in the presence sites. Negative MESS values indicate areas where at least one variable is outside the range of the experienced conditions and thus where model predictions can be less robust [ 62 ].

Suitability variation over time was calculated as the difference between future and historical suitability (Suitability variation = future suitability–historical suitability) and assigned to the observed plant stands on the basis of their location. Suitability variation ranges between -1 (perfectly historically suitable and perfectly unsuitable in the future) and 1 (perfectly historically unsuitable and perfectly suitable in the future). Following our approach, the predicted suitability variation represents an ongoing process, ranging between 2000 and 2050, and the population data collected in the field represent a snapshot approximately in the middle of this process.

Association between observed and expected patterns

Test of model-based pattern.

For each variable, we tested the linkage between the observed and the expected patterns of responses at the stand level (i.e. measured synthetic indicators vs modelled suitability variation) by adopting a null-model approach [ 63 – 65 ]. First, we quantified the observed correlation between measured values and expected suitability variation in the same sites ( r obs ) by calculating the Pearson r. Second, we generated in R (as all the other analyses) 30,000 random permutations of the measured values and calculated the simulated correlation with the expected suitability variation for each permutation ( r sim ). Third, we calculated the probability of the null hypothesis that the observed correlation was drawn at random from the distribution of the simulated correlations [ 66 ]. Finally, in order to control the familywise error rate due to multiple comparisons, we corrected our p values adopting the approach proposed by Benjamini & Hochberg [ 67 ]. These corrected p values ( p corr ) measure the level to which the suitability variation (corresponding to the expected response) due to climate change explains the actual responses observed in the different stands.

In addition, we tested whether the observed responses follow a general and simple geographic combination. We tested the hypothesis of a latitudinal (equator-to-pole) or altitudinal (low-to-high elevation) range shift, as if often assumed for detecting climate change impacts on species distributions. To do this, we adopted the same approach used for testing the linkage between the observed and the model-based expected patterns of responses. We contrasted each measured variable with the stand latitudes and altitudes. We quantified the correlation between the measured variable and the latitude/altitude, calculated the probability that the observed correlation comes randomly from the simulated correlations after 30,000 random permutations, and corrected our p values with the Benjamini & Hochberg [ 67 ] approach. With this approach, we assumed that populations at higher latitude/elevation should be in better general conditions (i.e., lower proportion of plants in poor conditions and of dead plants and higher proportion of plants in good conditions and of plants with cones) if climate change effect is following a simple geographic gradient. As a result, we obtained an estimation of the extent to which climate change effects can be explained as a simple geographic gradient.

Overall, we recorded 1330 plants within the known distribution of W . mirabilis in northern Namibia. These plants are clustered in 12 distinct stands, which are scattered across the central part of the known range at elevations between 806 and 991 m above sea level. On the basis of our field effort and the expert knowledge of our local team members, we are confident that the recorded/observed individuals, and their resulting extent of occurrence, represent the majority of plants in this northern-Namibian sub-range. The area of each recorded stand varied from 2000–825,000 m 2 (for a total surface of about 1.5 km 2 ) and the number of plants per stand varied between four and about 400. The extent of occurrence (estimated as the minimum convex hull) of welwitschia in the area covers about 215 km 2 and the inter-stand distance varied from 1.8 to 30 km. Even though we searched extensively throughout the study area, we defined a markedly smaller extent of occurrence than the distribution map previously published for the species in northern Namibia [ 29 ], which was based on very general information.

The available climatic models revealed that the temperature historically available in the current extent of occurrence of welwitschia in northern Namibia is becoming almost completely unavailable in the same area ( Fig 2A ). In particular, annual mean temperature within the stands is rising about 1.5–2.5°C, with strong variations among the different CMIP5 scenarios. In contrast, the total annual precipitation is likely remaining relatively stable ( Fig 2B ), with small reductions or increases forecasted by different scenarios.

Density plot of historical data (in green) and expected future values (in orange) for annual mean temperature (A) and annual precipitation (B) in the welwitschia extent of occurrence. Density plots use kernel density estimates to show the probability density function of the variables.

Observed variation of plant parameters

The most common class of health condition was ‘average’, with 50% of all the plants and a range between 32% and 74% across individual stands being found in this status ( Fig 3B ). Plants in ‘poor’ condition were 32% (range: 11–50%), but only 10% of all plants were in a ‘good’ condition (range: 0–30%) ( Fig 3A and 3C respectively). Seven percent of all plants were dead (range: 0–30%) and 56% (range: 10–90%) had cones ( Fig 3E and 3D respectively). Not all individuals could be sexed, but among those that were, 56% were males, with a sex ratio (males/females) ranging between 0.6–1.7 across stands ( Fig 3F ). Stem major axis and stem minor axis were highly variable, ranging from 2 to 100 cm (18.8 ± 14.1 cm; range: 10–33 cm) and from 0.3 and 55 cm (10.3 ± 9.8 cm; range 4.6–22 cm), respectively ( Fig 3G and 3H respectively). Leaf length varied from almost 0 cm (completely browsed plants) up to 93 cm (18.7 ± 13.4; range 11–40 cm) ( Fig 3I ) (see S1 Table ).

Density plots of measured values for: proportion of plants in poor conditions (A), proportion of plants in average conditions (B), proportion of plants in good conditions (C), proportion of plants with cones (D), proportion of dead plants (E), sex ratio (F), stem major axis (G), stem minor axis (H), and leaf length (I). Density plots use kernel density estimates to show the probability density function of the variables.

The previously proposed difference between plants growing in valley bottoms and on hill slope (Kers, 1967) was not verified. Indeed, most of the tested variables were not significantly different for the two groups of plants (stem minor axis: Mann-Whitney U = 10773, p = 0.49; leaf length: U = 10364, p = 0.1968; health condition: Mann-Whitney U = 189, p = 0.32; reproductive status: χ 2 = 0.05, p = 0.82). The plant stem major axis only was different ( U = 13120, p = 0.02) but in the opposite direction respect to the proposed pattern, with plants on slopes bigger than plants on valleys (mean: 14 ± 7.85 cm and 11 ± 8.48 cm respectively).

Expected pattern of species response

The strongest reduction of climatic suitability is expected in the eastern half of the extent of occurrence of welwitschia, as well as in some areas extending further north and south to the extent of occurrence ( Fig 4 ). On the other hand, our models predict an increase in climatic suitability to the northwest of the current extent of occurrence ( Fig 4 ). The MESS analysis indicated that all the pixels in the species extent of occurrence have positive values (minimum: 1.10, mean: 7.19 ± 2.47), suggesting that geographical extrapolation on future predictions are reliable (see S3 Fig ). As a result, all the recorded stands are expected to be facing suitability reductions in the period 2000–2050, with variability between almost no reduction and complete reduction.

Expected suitability variation from climate change (red and green shades indicate negative and positive variations respectively) calculated as the difference between future and historical suitability. Black polygon indicates the study area, and the blue polygon represents the boundaries of the observed extent of occurrence in Northern Namibia.

Observed vs expected patterns

Stronger predicted reductions of climatic suitability in the stand sites are associated with lower plant health condition, fewer plants with cones, and an increased number of dead plants. More specifically, the proportion of plants in poor condition in each stand increases with the reduction of suitability ( Fig 5A ). In contrast, the proportion of plants in average and good condition decreases as suitability variation decreases ( Fig 5B and 5C ). The proportion of plants with cones (i.e. a proxy of the potential population recruitment) is lower in stands where stronger reductions of climatic suitability are expected ( Fig 5D ). At the same time, the proportion of dead plants (i.e. population mortality) is negatively correlated with the predicted variation of climatic suitability ( Fig 5E ). However, neither the number of plants per stand (i.e. population size) ( Fig 5F ) nor plant body size ( Fig 5G, 5H and 5I ) is correlated with the suitability variation.

Plots of the proportion of plants in poor (A), average (B), and good conditions (C), the proportion of plants with cones (D), the proportion of dead plants (E), the number of plants (F), the stem major (G) and minor axis (H), and the leaf length (I) as functions of the expected suitability variation in the stands. In all the plots, blue dots are values for plant stands, colored lines and areas represent the regression lines and the confidence intervals respectively, and the texts on the top of the plots indicate the results of the null-model correlation analyses (Pearson r and corrected p values for multiple comparisons). Note that suitability variation values (X-axis) are all negative; thus, the reduction of suitability increases from right to left.

The observed geographic combination of species response does not follow any simple geographic gradient. Indeed, the latitude of welwitschia stands is not correlated with any measured variable ( Table 1 ). Similarly, altitude and the measured variables are not correlated ( Table 1 ). Overall, no latitudinal or altitudinal variation is occurring as a response to climate change.

Results of the null-model correlation analyses (Pearson r and corrected p values for multiple comparisons).

The observed pattern of population conditions of welwitschia plants in northern Namibia is consistent with the expected response under climate change, and specifically with predicted variations of climatic suitability. These results strongly suggest that ongoing climate change is affecting the status of welwitschia populations in the area and producing significant changes (i.e. contraction) of the species distribution at the local scale, which translates into a threat for the long-term conservation of the species. On the other hand, the geographic combination of welwitschia response to climate change does not follow a latitudinal or altitudinal gradient. Therefore, the potential impact of climate change on this species would have passed undetected using the common testing approach based on poleward/upwards range shift.

Inter-stand variations of multiple parameters (i.e. plant health conditions and proxies for population trends) are highly correlated with estimated variation in climatic suitability, strongly suggesting a negative effect of the ongoing climate change on welwitschia trees. In this light, the high correlation between the variation of climatic suitability and plant conditions (poor, average, and good) can support a link among climate change, the distribution of plants, and the variation of plant health with observed increase of individuals in poor conditions and the reduction of plants in average or good conditions. The loss of climatic suitability could also affect future population trends, by affecting recruitment (as suggested by the observed reduction of plants with cones) and mortality (as suggested by the observed increase of dead plants). Although we measured static parameters of population condition, the geographic combination of these parameters (observed in 2019) is coherent with the dynamism of a range shift from areas that were suitable in the past (before 2000) to areas that will be suitable in the future (2050). Indeed, bad population conditions such as poor plant health, low reproduction potential, and high mortality provide a proxy of negative population dynamics, which are typically associated with the trailing edge of species range. This study indicates a clear relationship between observed and expected patterns of populations response in welwitschia. The responsivity of welwitschia to climate change as measurable by proxies of population status could make this species a ‘sentinel’ of climate change effects for Namib ecosystems. The application of the approach used in this study is thus encouraged to identify sentinel species in other desert environments around the world for an effective biomonitoring of climate-linked biotic changes.

The visual estimation of plant condition can be considered a rough estimation of chlorophyll content of the leaves and thus of the plant’s photosynthetic efficiency [ 41 , 42 ]. Alterations of photosynthesis is a well-known effect of environmental stress [ 43 , 44 ]. Heat stress in particular inhibits photosynthesis in tropical and subtropical plants [ 68 , 69 ]. This effect can be stronger in arid environments, where water shortage can hamper the leaf temperature mitigation [ 70 ]. On the other hand, studies on other populations of W . mirabilis showed that rainfall is followed by an increase in the plant’s health status [ 40 ]. As a result, we hypothesize that the observed worsening of plant condition is associated with the complex interaction between the significant increase in temperature, which is the main predicted climate alteration occurring in the area ( Fig 2 ), and the constant but limited water availability in the desert environment. However, specifically designed experiments would be needed to tease apart the different possible forces that could cause the observed responses.

Our results are consistent with the study of Bombi [ 23 ] carried out at the national level and at a much coarser spatial resolution. That study suggested a potential impact of climate change on welwitschia populations in northern Namibia and, predicted a general reduction of climatic suitability for W . mirabilis as well as potential effects on population recruitment and thus on population structure. The author postulated that adult welwitschia plants are likely to survive the expected reduction in climate suitability. In contrast, the correlation we observed between mortality and climate change would indicate a less optimistic scenario, with a progressive reduction of plant health, which translates in a potential long-term reduction of population size. This evidence could support the hypothesis that climate is changing faster and/or is becoming too hot/arid even for W . mirabilis . In this regard, it is worth noting that W . mirabilis is not a typical desert-adapted plant (it has C3 metabolism and a relatively high water demand [ 71 ]), and its ancestors probably occupied much more mesic habitats (probably even forests). It is likely that the current range fragmentation was the result of strong aridification during the Tertiary and Quaternary [ 72 ]; with current climate change predictions pointing towards further desertification (increase of 2°C in annual mean temperature is expected by 2050). Furthermore, this evidence should encourage specific management plans for northern Namibian populations and suggests that climate change should be considered among welwitschia conservation issues.

Quantitative data on plant physiological status (e.g. leaf growth rate, photosyntetic efficiency, water use efficiency), on substrate characteristics such as soil moisture profiles, and on population demographic parameters (e.g. annual recruitment, plant growth, annual mortality) are required to obtain a more detailed picture of the occurring alterations and to clarify the possible mechanistic linkage with climatic stress. Repeated physiological and demographic measurements in different sites would make it possible to follow plant responses over time. The activation of programs for the long-term monitoring of the species in the region would be particularly helpful, allowing critical situations to be detected at early stages and planning of effective recovery measures. Obviously, long-term monitoring in this remote area would be difficult and would require the involvement of local communities as well as the provision of significant resources by local and international agencies aimed at the conservation of desert ecosystems in Namibia.

Despite the great interest in W . mirabilis , which is the only living representative of an ancient lineage of gymnosperms [ 73 ], and its key-role in the Namib Desert ecosystems, several aspects of the species distribution and biology are still to be clarified for a science-based conservation strategy. First, the real level of geographic and genetic isolation of the different subranges should be verified in order to identify intra-specific evolutionary and conservation units. Second, an effort to census and make available the current knowledge on species distribution, demography, and conservation should be undertaken. Third, an analysis of climate change impacts should be extended to the other subranges and a science-based assessment of the conservation status should be made at local and global level. This set of measures could significantly contribute to conservation measures for the species that are effective in the long term.

The geographic combination of response we observed in welwitschia is more complex than the simple poleward/upward shift that was often observed for other species [ 33 , 35 , 74 ]. In the case of W . mirabilis populations of northern Namibia, the observed pattern of population conditions, which can represent a response to climate change, follows local contingencies rather than a simple latitudinal or altitudinal trends ( Table 1 ). This could be associated with the small scale of the study, which emphasized the role of local factors (e.g. land morphology, dominant winds, recurring fog), but is also in agreement with previous large-scale studies. Previous studies pointed out that specific responses to climate change can be divergent [ 36 ] and that assuming a simplified poleward/upward species movement can result in climate change impacts being underestimated [ 37 ]. In our specific case, the linkage between climate change and population conditions, which is suggested by our results, would have been completely undetected with a simplified, but frequently used approach based on the assumption of poleward/upward shifts.

The comparison of the expected pattern of response to ongoing climate change, as estimated by suitability modeling, with the observed patterns of population conditions, as measured in the field, appeared as a powerful approach for detecting impacts of climate change on wild species. This approach, proposed by Bombi et al. [ 38 ], allowed the identification of climate change as potentially a major driver of the geographical pattern of welwitschia health, reproductive status, and size we observed in the field. Clearly, such an approach is prone to a certain level of false negative when other factors, not directly related to climate change (e.g. wind, herbivory, parasites), interact, blurring the pattern generated by climate change [ 38 ]. In addition, our small sample size could further increase the probability of type II errors. On the other hand, the probability of false positives is very low and not influenced by non-climatic factors [ 38 ] or sample size, strengthening the meaning of the detection of a climate change-related pattern. This study underlines the importance of considering species responses to climate change as an emergent property of the different effects on individual populations. At a higher biodiversity level, ecosystem responses to climate change can be considered as an emergent property of the effects on individual species. Such a hierarchical relationship provides direction for the application of spatially explicit approaches, such as the one used in this study, to multiple species and across diverse ecosystems. We therefore advocate for the implementation of a large-scale program for the identification of sentinel species of climate change effects. This would allow the detection, estimation, and monitoring of climate change impacts on biodiversity, improving the long-term conservation of species at the ecosystem level.

Our study confirms that desert-adapted species can be vulnerable to the effects of climate change. Although some studies have hypothesized minor impacts on desert ecosystems by climate change [ 75 , 76 ], growing evidence, provided through different approaches, shows severe effects of climate change on species, communities, and ecosystems in arid regions worldwide [ 18 , 77 , 78 ]. Altogether these studies, in agreement with our results, indicate that desert ecosystems are likely to suffer from biodiversity loss with intensifying global warming as result of a reduction of environmental suitability for the endemic biota [ 22 ].

Supporting information

Examples of plants in the four health condition classes (A: Dead; B: Poor; C: Average; D: Good).

Color shades indicate the extent to which the predictor variables in each pixel are similar to the conditions experienced by the species in the presence sites. The black line separates zones with positive values to those with negative values. Negative MESS values indicate areas where at least one variable is outside the range of the experienced conditions and thus where model predictions can be less robust. The blue polygon represents the species extent of occurrence in the area.

S1 Appendix

Acknowledgments.

We wish to thank our Himba guides and friends Riatunga Koruhama and Mavekaumba Tjiposa; Karen Nott of IRDNC, who facilitated access to the area; Vera De Cauwer of the SCIONA project (EuropeAid/156423/DD/ACT/Multi), for the constructive collaboration. We are also grateful to the Okondjombo Communal Conservancy.

Funding Statement

This study was supported by the Mohamed bin Zayed Species Conservation Fund ( www.speciesconservation.org ) to PB (Project N 182519816).

Data Availability

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 27 December 2023

The urban heat island under extreme heat conditions: a case study of Hannover, Germany

• Nadja Kabisch 1 ,
• Finja Remahne 1 ,
• Clara Ilsemann 1 &
• Lukas Fricke 1  

Scientific Reports volume  13 , Article number:  23017 ( 2023 ) Cite this article

2018 Accesses

7 Altmetric

Metrics details

• Climate sciences
• Environmental sciences
• Environmental social sciences

Global warming has resulted in higher frequencies of climate extremes, such as drought periods or heat waves. Heat waves are intensified in urban areas due to the urban heat island effect. Studies are inconclusive as to whether the urban heat island effect is intensified during heat waves. Using the city of Hannover, Germany, as a case study, we analysed the intensity of the urban heat island under unprecedented summer heat conditions in the years 2018, 2019 and 2020, which were among the hottest in Germany since weather records began. We compared the intensity of the urban heat island across these years with the non-heat year of 2017. Differences were analysed for various inner-city urban locations and an urban park, while accounting for their distinct land use and land cover characteristics. We identified the urban heat island effect across all years investigated in the study and also found a significant intensified urban heat island effect during the years of unprecedented heat, when night-time temperature minima are considered. The urban heat island was identified on a lower level, however, with maximum daily temperatures when compared to the non-heat year. The lowest intensity of the urban heat island was found for the urban park site, highlighting the need for more city-wide greening strategies, including tree-covered and open green spaces, to provide all residents with the cooling services of green spaces.

Similar content being viewed by others

Disappearing cities on US coasts

Leonard O. Ohenhen, Manoochehr Shirzaei, … Robert J. Nicholls

Expert review of the science underlying nature-based climate solutions

B. Buma, D. R. Gordon, … S. P. Hamburg

Meta-analysis shows the impacts of ecological restoration on greenhouse gas emissions

Tiehu He, Weixin Ding, … Quanfa Zhang

Introduction

For the past several decades, many regions around the world have been impacted by extreme climate events as a result of climate change, including extreme precipitation, droughts and heatwaves 1 . The impacts of these extreme events and, particularly, their co-occurrence with one another, such as heatwaves occurring under drought conditions, have been shown to result in severe environmental, economic, social and health challenges, as seen, for instance, in China 2 and Europe 3 . Health-related challenges associated with heat periods include increased mortality and morbidity. For example, the summer heatwave in Europe in 2003 resulted in more than 70,000 heat-related deaths 4 and the heatwave in 2022 in more than 60,000 estimated deaths, predominantly in Spain and Germany 5 . Excess summer deaths were shown for Germany for the previous extreme heat years of 2018, 2019 and 2020, with approximately 20,000 heat-related deaths for this period 6 . In particular, the year 2018 was highlighted as the warmest year in Germany since the beginning of systematic weather records, with an annual average temperature of 10.5 °C being 2.3 °C higher than the long-term mean (1961–1990: 8.2 °C). Also, the years of 2019 (10.3 °C) and 2020 (9.9 °C) were exceptionally warm 7 . In fact, exceptionally high temperatures during these periods were identified for cities like Berlin and Hannover.

Cities are further impacted by the urban heat island (UHI) effect. The UHI is expressed by elevated temperatures within cities compared to their rural surroundings 8 . Higher temperatures in urban areas are caused by the specific characteristics of the urban landscape including a high degree of impervious surface, an accumulation of concrete material, the three-dimensionality of urban structures, limited shares of green and blue spaces, and anthropogenic heat storage and release, such as from cars, industry and air conditioning. The UHI is, thus, a result of the spatial arrangement of the built environment, the urban land use and land cover (LULC), overall thermal characteristics, and additional heat released from anthropogenic sources, all aggravated by ongoing urbanisation processes 9 .

As extreme climate events such as heat waves will occur in higher frequencies globally 1 , the possibility that the UHI effect may be intensified or amplified by heat waves (such as the above introduced heat waves occurring in Germany 2018, 2019 and 2020), through synergistic or combinative effects, should be examined. Scientific studies that analysed potential synergistic effects of UHI and heat wave conditions have used observational air temperature data from weather stations 10 , 11 , 12 or land surface temperatures (LST) derived from remote sensing 13 (for an overview of these different methodological approaches deriving UHI related effects see Lai et al. 14 and Zhang et al. 15 ). These studies mostly confirm the overall assumption of an intensified UHI during heat wave periods, e.g. for the cases of Madison, in the USA 16 , Melbourne and Adelaide, in Australia 10 , and Athens, in Greece 17 . For the Madison case, the authors found an intensification of the UHI for day-time and night-time during heat periods and that the UHI during day-time was stronger, with 39 days experiencing temperatures over 32.2 °C compared to only 9 days for the long-term mean 16 . In Melbourne, the UHI was identified to be more pronounced by up to 1.4 °C during heatwaves and in Adelaide by up to 1.2°C 10 .

The extent of an UHI intensification under heat was, however, shown to be very different depending on the respective climate zones 18 , local environmental conditions and LULC structures, and temporal aspects. For example, coastal cities, such as Athens, showed an amplification effect of the UHI under heat conditions with an average UHI by up to 3.5 °C, depending on the background wind field, such as sea breeze conditions which induce a cooling effect at coastal stations, intensifying the extent of the UHI 17 . In terms of LULC, urban green spaces are usually promoted as a nature-based solution to mitigate high temperatures 11 , 19 , 20 , 21 . Less is known, however, about the heat mitigation effect of urban green spaces during heat wave conditions. Finally, differences in the intensity of the UHI were reported with different temporal aspects, such as when focussing on different time periods throughout the day 17 , 22 . For example, synergistic effects of the UHI and a heat wave period were found particularly during the night but with a lower or even reversed effect during the day 10 , 23 , 24 . Some studies could not, however, identify any significant difference of the UHI intensity during heat wave periods compared to non-heat wave periods 25 and most of the studies only assess one single heat event 16 , 17 .

In conclusion, the relationship between the UHI and heat, and their synergistic direction, is not fully conclusive and tends to depend on local, methodological, and temporal factors. The main aim of this study, then, is to assess the intensity of the UHI during several consecutive years of unprecedented extreme heat and to compare different urban built environments and an urban park location with the air temperature in rural areas. In particular, the following research questions are addressed:

What is the air temperature difference between certain urban areas and the rural surroundings in years of extreme summer heat conditions versus a non-extreme year? Are these potential differences significant? To what extent does the potential intensity of the UHI differ when diversely-structured areas around urban measurement stations, including an urban park location, are considered?

Using the large German city of Hannover as a case, our objectives are to i) analyse air temperature data from five measurements stations in the rural, urban and urban park areas of the city for the summer time period of the years 2017–2020; ii) compare daily mean, maximum and minimum air temperature values for the summer period of the year 2017 as a non-heat year with the years of 2018, 2019 and 2020, known as years of extreme heat in Germany; iii) analyse the local LULC around the measurement stations to derive potential UHI intensity influencing factors.

Methodology

Case study: hannover, germany.

The city of Hannover is the capital of the Federal State of Lower Saxony and is located close to the southern border of the Northern German Lowland in the transition area to the Lower Saxony Mountain and Hilly Region. The city is a large city counting 543,247 inhabitants in 2021, with a population density of 2,661 inhabitants/km 2 and a total area of 204.15 km 2 26 . The city has been growing, with a population growth rate of 7% between 2008 and 2018 and further growth is expected by 2.8% until 2029 27 .

Hannover is characterised by a diversity of different landcover categories within its 13 urban districts. A share of 35% of the city areas are dedicated as green spaces, including urban forests, parks, cemeteries, gardens, agricultural areas, etc. Blue spaces including lakes, canals, rivers, etc., accumulate to around 3.1% of the total city area (Fig.  1 ).

Hannover case study with land cover and location of the five measurement sites. Data source: Urban Atlas (EEA, 2018). Note: (1) LH = Langenhagen (rural site); (2) HE = Herrenhausen (urban site), (3) WD = Weidendamm (urban site); (4) MB = Marianne-Baecker Allee (urban site); (5) KP = Kattenbrookspark (urban park site).

Hannover, like many other cities and regions in Germany, experienced unprecedented heat conditions during the summers of 2018, 2019 and 2020, when compared to previous years. Summer mean air temperature, yearly mean air temperature and sum of precipitation for the city of Hannover for the period 2010–2021 are presented in Table 1 . The years of 2018–2020 stand out as the three hottest and also driest consecutive years. Similarly, the summer mean air temperatures in Germany for the years under study were: 2017: 17.9, 2018: 19.3, 2019: 19.2 and 2020: 18.2 28 . The year 2017, thus, is used as a comparison in our study, as a non-heat year when compared to the following consecutive years with unprecedented summer heat.

Data from a meteorological measurement campaign of the German Weather Service (GWS, DWD) 30 are used in this study (Table 2 ). The meteorological measurement data were recorded with the help of temporary stationary measurements at various locations in urban and rural areas of the city of Hannover. Measured data included the following meteorological parameters: air temperature, relative humidity, wind speed, wind direction and, at selected locations, incoming global radiation, measured at a height of two meters over a period of three and a half years (June 2017–December 2020) 30 . As explained above, the focus of the years 2018–2020 is based on the fact that these years were shown to be the hottest years in Germany since weather records but also on data availability. The meteorological measurement campaign for the five stations ended in December 2020. Overall data loss was low (less than 1%), and data gaps or inconsistencies were reported only for one measurement station (HE urban) for some days in June 2018 (for the period 1.6–4.6; 7.6–10.6 and for the 12.6.–30.6.2018).

For our analyses, we used the data for the summer period. The meteorological definition of the summer is from 1 June to 31 August. A summer day is defined as a day with a maximum air temperature ≥ 25 °C and a heat day is defined as a day with a maximum air temperature ≥ 30 °C. Finally, tropical nights are defined as nights where the minimum night temperature is ≥ 20 °C between 18:00 and 06:00 UTC 30 . We define the UHI intensity as the commonly used 2 m-height air temperature (near-surface air temperature) at each urban station minus the air temperature at the rural station, as per previous studies 23 . When using the maximum daily or minimum daily air temperatures, we also refer to them as day-time or night-time temperatures 31 .

The five measurement sites included three inner-city urban sites: Weidendamm (WD urban), Marianne-Baecker-Allee (MB urban), Herrenhausen (HE urban); an urban park site Kattenbrooks Park (KP urban park); and the rural site at Langenhagen (LH rural) close to the airport. The two stations HE and LH are under permanent use. The locations of the five stations are shown in Fig.  1 .

The measurement station at Weidendamm (WD urban) is located in the Nordstadt district—a dense inner-city residential area. The Marianne-Baecker-Allee (MB urban) station is in the commercial area of the Linden-Süd district and can also be regarded as an inner-city location representing urban conditions. The Herrenhausen (HE urban) station is located on the premises of the Institute for Meteorology and Climatology of the Leibniz University Hannover. The measurement station in Kattenbrooks Park (KB urban park) is located within an open lawn area of an urban park. The measurement station representing the rural site is Langenhagen (LH rural), which is located northwest of the city. The location of the site complies with the internationally defined standards of the World Meteorological Organization (WMO) 30 .

To assess the LULC around the measurement stations, the Urban Atlas data of the European Environment Agency 2020 for Hannover was used with a 10 m minimum mapping width. A 3D-building model and a Digital Orthophoto were additionally used to analyse the local environmental and urban structures around the measurement stations (Table 1 ). The Digital Orthophoto with a grid width of 10 m presents the basis of the terrain heights of the buildings. The building model includes block models so that buildings are represented as rectangular and with flat roofs. The Digital Orthophoto (DOP), with a ground resolution of 20 cm covering  a total area of 2 km  \(\times\)  2 km, was used to calculate the Normalized Difference Vegetation Index (NDVI). The NDVI is an indicator for surrounding greenness 35 and was previously used as an vegetation-indicating parameter in studies on the urban climate environment 36 .

Data analysis

We applied descriptive statistics (mean, minimum and maximum, standard deviation) and spatial analyses in a Geographical Information System (GIS) to analyse the (differences in the) intensity of the UHI over the years and to show potential spatial–temporal differences and local specifics of the surrounding LULC.

A parametric test, the paired t-test, was used and considered appropriate for investigating the statistical significance in the outcome environmental variables between the rural site (LH rural), the different inner-city urban sites and the urban park site for each year of interest. A one-way analysis of variance (ANOVA) with post hoc tests (Bonferoni) for multiple comparisons were used to explore whether the UHI-intensity differs significantly between the years of interest, that is, to identify if there are significant differences in the UHI-intensity levels (measured by daily mean air temperature, max. air temperature and min. air temperature) of the non-extreme year (2017) compared to the extreme heat years (2018, 2019 and 2020). As a part of performing the ANOVA, tests for the required distributional assumptions were conducted which included the Levene statistics to assess the equal variance assumption. Significance for both the paired t-tests and the ANOVA was considered at the p < 0.05 level as applied in earlier studies 37 . Statistics were performed using IBM SPSS Statistics 28.0 (IBM Corporation, Armonk, NY, USA).

To assess the LULC in close vicinity of the measurement sites, a buffer of 100 m around the measurement stations was created to identify the percentage of the different LULC classes, the number and percentage of buildings of different heights, and percentage of the area within NDVI value classes. Buildings were grouped into three classes according to their height. The first class grouped buildings from 0 m to ≤ 13 m tall, including both detached buildings with no more than two units of use and detached buildings, such as those used for agriculture and forestry. The second class includes buildings such as single-family houses and high-rise buildings with heights between > 13 m and ≤ 22 m. The third class grouped high-rise buildings at a height > 22 m.

Air temperature differences between the inner-city urban, the urban park and the rural measurement sites

Table 3 provides an overview of the environmental conditions during the summer periods for 2017–2020 in the city of Hannover. Mean air temperature was lowest in the non-heat year of 2017 for all measurement sites with lowest values of 17.8 °C for the rural site LH and the urban park site KP. Highest maximum air temperatures were measured for the heat year 2019, with the inner-city urban station HE, for instance, recording a maximum temperature of 39.6 °C in 2019. Mean values for relative humidity were usually highest for the rural site LH (with one exception of 2020 where the urban park (KP) area showed slightly higher values). 2017 stands out as the year with highest humidity values measured at all stations compared to the other years. Nearly all values measured at the urban and urban park stations were significantly different compared to the values at the rural station (LH rural).

Figure  2 shows the total number of heat days and tropical nights for all measurement stations for the years 2017, 2018, 2019 and 2020. The number of heat days was lowest for the non-heat year 2017 (between 2 heat days at LH and 6 at HE) and highest for 2018 (between 22 heat days at LH and 32 at HE) for all measurement stations. Numbers are higher respectively for the inner-city urban sites compared to the rural site LH. The total numbers differ, however, with regard to the tropical nights. For 2017, the inner-city urban site of WD is the only measurement station where tropical nights were detected (2 in 2017). During the heat years of 2018–2020, the total number of tropical nights was two up to four times higher for the inner-city urban stations compared to the rural station LH and also compared to the urban park station, indicating an UHI effect during night-time summer conditions for the years of unprecedented heat. The highest number of topical nights was shown for the year 2018 at all measurement stations (with 3 tropical nights at the urban park station KP, 4 at the rural station LH and 13 at the urban station WD).

Total number of heat days (in yellow) and tropical nights (in blue) measured at the rural, urban and urban park measurement sites for the summer periods of 2017, 2018, 2019, 2020. Note: LH = Langenhagen, WD = Weidendamm, MB = Marianne-Baecker Allee, HE = Herrenhausen, KP = Kattenbooks Park.

To assess whether the UHI intensity is significantly different between all years of interest, Fig.  3 shows boxplots of daily mean air temperature, maximum and minimum air temperatures and also highlights indication of significance by different significance levels.

Boxplots of the mean differences of daily mean air temperature, maximum air temperature and minimum air temperature between the rural and the urban sites for each year of interest. Whiskers indicate 1.5 × interquartile range. Significant differences are indicated between the years with * at the 0.05 level, ** at the 0.01 level, and *** at the 0.001 level. Note: In cases when Levene statistics revealed that the variances for the variables were not equal, Welch statistic was reported and Tamhane’s T2 – post-hoc test were used as a robust multiple comparison tests, not assuming equal variances. Note: LH = Langenhagen, WD = Weidendamm, MB = Marianne-Baecker Allee, HE = Herrenhausen, KP = Kattenbooks Park.

Significant differences for the daily mean air temperature were observed for the inner-city urban sites for the non-heat year of 2017 but on a lower level compared to 2018 (Fig.  3 , boxplots a–d). Significant differences were found between 2017 and the heat years when minimum air temperatures indicating night-time conditions are considered. The greatest differences were identified for 2018 compared to 2017 for the urban sites WD, MB and HE, reaching up to around 2.5 K (median value, Fig.  3 , boxplots i–l), indicating that the UHI intensity is pronounced during the night in years of extreme heat when compared to a non-heat year. This association of a synergistic effect of the UHI and heat wave periods is, however, not detectible and even inverted when considering the differences in the daily maximum temperatures (Fig.  3 , boxplots e–h). In 2017, the UHI intensity is even on higher levels for maximum air temperatures compared to the heat years of 2018, 2019, and 2020, with significant differences particularly for the urban sites of WD and HE. The difference between the years is non-significant when the urban park site is considered.

Land use and land cover (LULC) structures around the measurement sites

The percentage shares of LULC within a buffer of 100 m around the measurement sites are shown in Fig.  4 . LULC around the rural measurement station LH consists of open areas such as pastures, arable land and land denoted as airport. The inner-city urban measurement sites are mostly surrounded by industrial, commercial and transport infrastructure areas. The share of urban fabric is around 16% at the urban station WD and 1% at the urban station HE. More than three quarters of the LULC close to the urban park station (KP) is classified as green urban areas and around 20% as transport infrastructure.

Percentage share of LULC in a 100 m buffer around the measurement sites. Note: LULC classes are summarized based on Urban Atlas 2018 classification. Urban fabric includes “Continuous urban fabric” and “Discontinuous medium density urban fabric”. Note: LH = Langenhagen, WD = Weidendamm, MB = Marianne-Baecker Allee, HE = Herrenhausen, KP = Kattenbooks Park.

The respective LULC is also reflected in the NDVI values illustrated in Fig.  5 . Percentage of areas with higher NDVI values indicating more vegetation cover are shown for the rural site (LH) and for the urban park site (KP). The area around the inner-city urban measurement stations is mostly sealed, as shown by a high share of the area having NDVI values close to or below 0.05. The inner-city urban stations HE and WD also show a high share of building area with more than 50% of the area covered by smaller buildings below 13 m height.

Buildings with different heights within a 100 m buffer around the measurement stations (left) and NDVI values (right). Note: (1) LH = Langenhagen (rural site); (2) HE = Herrenhausen (urban site), (3) WD = Weidendamm (urban site); (4) MB = Marianne-Baecker Allee (urban site); (5) KP = Kattenbrookspark (urban park site).

Synthesis of results

Table 4 presents a synthesis of the main variables with regards to the extent of the UHI intensity in our study. It shows the data for 2017 compared with 2018 as a heat year. The rural station LH and the urban park station KP are presented with the lowest values for the maximum and minimum air temperatures for both years. The highest measured air temperature values are indicated for the inner-city urban stations, with HE showing the highest maximum daily values, and WD showing the highest minimum air temperatures for both years. The number of heat days and tropical nights were lowest for 2017 and 2018 for the rural and the urban park stations. The number of heat days was highest with the urban station HE (32 in 2018), while the number is slightly lower for the urban station WD (27 in 2018). At this measurement station (WD), the number of the tropical nights is, however, highest (13 in 2018 vs. 10 at HE). Also, at this measurement station (WD), the relative humidity was lowest in both years. This difference between the two urban stations WD and HE may be explained by the different LULC structures at WD compared to HE. At WD, the area within 100 m shows the highest share of impervious surface (indicated by very low NDVI values < 0.05, Fig. 5 ) and the highest share of buildings, in particular buildings taller than 13 m (Table 4 ).

Based on these LULC conditions, we conclude that the area around the urban station HE is heating up to higher levels during the day compared to WD, but also cools down more during the night. At the urban station WD, the daily air temperature is heating up to a slightly lower degree, but also cools down less during the night due to a much slower heat release. The slower heat release can be explained by the high share of impervious areas and building cover. In particular, the comparatively higher share of buildings with heights of 13 m and above and the high share of impervious surfaces prevent nocturnal cooling. The tall buildings provide, however, more shading throughout the day which can explain that the maximum heat and number of heat days is comparatively lower here at WD compared to the other inner-city urban stations.

In this study, we analysed the extent of the UHI effect for the city of Hannover by comparing three years of unprecedented summer heat (2018–2020) with a non-heat year (2017). Using data from five measurement stations located in a rura areal, three inner-city urban areas and an urban park area, we identified the UHI phenomenon for all years under study and for all measurement stations. When comparing air temperatures of the non-heat year with the extreme heat years, we found a greater UHI intensity during the nights of the extreme heat years compared to the non-heat year when minimum air temperatures are considered. This association was detectible especially for the very hot and very dry summer of 2018. Conversely, when day-time temperatures are considered with maximum air temperatures, the UHI intensity was on a lower level during the heat years compared to the non-heat year. The identified UHI intensity associations were on a lower and partly non-significant level for the urban park location indicating the local cooling potential of urban green spaces to mitigate the UHI.

The years of 2018, 2019 and 2020 have been highlighted as years with exceptional, unprecedented heat in Europe and Germany 6 , 38 . In particular, 2018 was indicated as the warmest and also driest year in Germany since weather records. We identified the greatest amplification of the UHI intensity under heat conditions for night-time conditions for 2018. The extreme conditions particularly in 2018 may be related to the co-occurrence of the heatwave and drought period which could have led to further synergistic effects amplifying the heat conditions. Elsewhere, in Europe 38 and China 2 , the co-occurrence of heat and drought have been shown to amplify the effects of both on biodiversity and vegetation conditions, such as by limiting the provision of ecosystem services.

For Hannover, we identified an air temperature amplification effect of the UHI for the night-time conditions for the heat years. For day-time conditions (maximum temperatures), we saw a reduced intensity of the UHI compared to the non-heat year. For the city of Berlin, Germany, during the heatwaves of 1996 and 2006 24 , an amplified UHI was also identified for night-time conditions with an intensity of 1.29 K for 1994 and 0.83 K for 2006. When day-time temperatures were considered, air temperature measured at the Berlin city locations were even lower than those in the rural areas. Another study in Berlin found a non-significant day-time UHI intensity between normal conditions and heat days 39 . In the city of Guangzhou, China, the UHI intensity under heat waves was found to be amplified during the day and also the night but that it was also more significant and stronger at night 23 . Others have found that the UHI is stronger during heat days for both day-time and night-time conditions, such as in an Australian study for Adelaide and Melbourne 10 .

The different UHI intensities for day-time and night-time conditions may be related to the characteristics of the local urban environment including the LULC close to the measurement stations. Urban structures with urban materials, high density of buildings, and high rates of impervious surfaces contribute to a high heat storage capacity during the day. The thermal admittance and release of heat in the form of sensible heat is, then, also higher during the night. Together with a usually more stable nocturnal boundary layer and a decreased vertical mixing, the UHI intensity shows highest values at night. This effect is reinforced during extreme heat periods 23 which we also found for Hannover, particularly in 2018. Similarly to our results of the inner-city urban location of WD having the highest share of taller buildings, research for the city of Guanzhou found that local climate zones with higher shares of taller and compact buildings had higher air temperatures but on a slightly lower level when compared to other urban climate zones 23 . The differences were explained by possible shading effects of taller buildings.

For Hannover, we showed reduced air temperatures with lowest minimum values during the day and the night for the rural measurement station but also for the urban park station, indicating the importance of urban green spaces to mitigate local UHI effects and the resulting impacts on human health. The UHI effect was nearly completely diminished for the urban park site. The effect of vegetation, such as green spaces, on local air temperatures and increased relative air humidity was also found in a study of the city of Würzburg, Germany 11 . In this study, Rahman et al. highlighted, however, that the cooling effect of a green space in urban areas is dependent on the local built environment and urban morphology, which may reduce wind speed, hinder ventilation and vertical mixing counteracting a potential cooling effect. For Hannover, the urban park measurement site is located in an open, lawn covered park area with a very low number of buildings surrounding it. The usual blocking of the natural wind flow in dense inner-city areas is reduced here but allows heat release during the night. The urban park site in Hannover can, thus, be regarded as a local cooling island under heat particularly during the night. This cooling effect is local, however, and has a limited effect on the wider residential area 38 and has also been referred to as the “tree cooling pond effect” 40 . Referring to the median radiant temperatures, Zhang et al. (2023) showed that the cooling effect of individual trees was mostly identified for the crown-canopied spaces with very limited effects when distance from the trunk increased in more sunny places 40 .

For a more climate resilient urban planning, our results highlight the need to introduce structurally diverse urban green spaces over an entire city area at as many locations as possible. Trees provide a cooling function particularly during the day even under hot and dry conditions due to shading effects 38 . Open areas such as lawn and grassland areas promote heat release and, thus, cooling during the night 38 . The combination of both, tree covered and open areas, for structurally diverse urban green spaces has been introduced as the “savannah approach” 38 , 41 . When developing and maintaining urban green spaces, also the wise selection of climate (heat and drought) adapted tree species with larger canopies could help to mitigate the UHI 40 . As space in growing cities is contested, novel concepts such as pocket parks, smaller parks in street corners, brownfield redevelopments or additional green infrastructure elements such as green roofs and facades could complement a wide and divers urban green space network 38 , 42 .

Limitations

Our analyses focussed on summer periods and used data of fixed reference dates (1 June and 31 August). Temperature extremes, however, potentially occur earlier (in May) or later (September) in the year. Due to this temporal focus, we may have missed out some additional hot days or tropical nights. The measurement stations also had some missing or invalid data, although with the overall data loss was below 1%.

Our study is based on data from five fixed measurement stations. These locations have different site conditions, so that a comprehensive picture of the complexity of the entire city cannot be provided. The data includes, however, several inner-city urban sites and an urban park site and the respective environmental conditions in terms of LULC were described and temperature differences explained in detail. Other approaches apply the concept of the footprint effect of an UHI using remote sensing data and land surface temperature which enables the application of different buffer zones to detect temperature differences 43 . Still the usability of remote sensing data is depending on cloud cover, etc. Finally, we used the NDVI as an indicator for LULC, in particular to show unsealed, vegetated areas. The values of the NDIV are, however, depending on the fitness of the vegetation and lower values can occur when vegetation is impacted e.g. during drought periods.

Conclusions

In this study, we analysed the extent and intensity of the UHI in the city of Hannover during unprecedented heat conditions. We compared data for the extreme summer heat periods of 2018–2020 with data from the non-heat year 2017. We identified an UHI for all years under study but a greater UHI intensity within the heat years compared to the non-heat year when night-time temperatures are considered. Conversely, we found a lower UHI intensity when focussing on the daily temperatures during the heat vs. non-heat years. These UHI intensity associations are strongest for the inner-city urban measurement sites located in dense urban districts. The measurement site at the urban park location showed a lower UHI and partly non-significant difference among the different years.

The results indicate the importance of implementing and maintaining urban green spaces in cities as local cooling zones which will become even more important under ongoing global warming. With increasing temperatures and higher frequencies of extreme heat events, the provision of structurally diverse urban green spaces with tree and tree-canopy covered areas to provide shading during the day but also open spaces to allow cooling during the night can support heat mitigation measures to regulate the local urban climate.

Data availability

Data will be made available on request but based on permission of the Deutsche Wetterdienst (see Acknowledgements section).

IPCC. Summary for Policymakers. in Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (ed. Masson-Delmotte,V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfeld, O. Yelekçi, R. Yu, and B. Z.) (2021). https://doi.org/10.1017/9781009157896.001 .

He, B. J., Wang, J., Zhu, J. & Qi, J. Beating the urban heat: Situation, background, impacts and the way forward in China. Renew. Sustain. Energy Rev. 161 , 112350 (2022).

Article   Google Scholar  

Romanello, M. et al. The 2021 report of the Lancet Countdown on health and climate change: Code red for a healthy future. Lancet 398 , 1619–1662 (2021).

Article   PubMed   Google Scholar  

Robine, J.-M. et al. Death toll exceeded 70,000 in Europe during the summer of 2003. C. R. Biol. 331 , 171–178 (2008).

Ballester, J. et al. Heat-related mortality in Europe during the summer of 2022. Nat. Med. 29 , 1857–1866 (2023).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Winklmayr, C., Muthers, S., Niemann, H., Mücke, H.-G. & an der Heiden, M. Heat-related mortality in Germany from 1992 to 2021. Dtsch. Arztebl. Int. (2022) https://doi.org/10.3238/arztebl.m2022.0202 .

German Weather Service (Deutscher Wetterdienst DWD). Klimatologischer Rückblick auf 2022: Das sonnenscheinreichste und eines der beiden wärmsten Jahre in Deutschland. https://www.dwd.de/DE/leistungen/besondereereignisse/temperatur/20230123_klimarueckblick-2022.html (2023).

Roth, M. Understanding urban heat islands. in The Routledge Handbook of Urban Ecology (eds. Douglas, I. et al.) (Routledge, 2021).

IPCC. Regional fact sheet - Urban Areas. Clim. Chang. 2021 Phys. Sci. basis 1–2 (2021).

Rogers, C. D. W., Gallant, A. J. E. & Tapper, N. J. Is the urban heat island exacerbated during heatwaves in southern Australian cities? 441–457 (2019).

Rahman, M. A. et al. Spatial and temporal changes of outdoor thermal stress: Influence of urban land cover types. Sci. Rep. 12 , 671 (2022).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Schatz, J. & Kucharik, C. J. Seasonality of the Urban Heat Island Effect in Madison. Wisconsin. J. Appl. Meteorol. Climatol. 53 , 2371–2386 (2014).

Article   ADS   Google Scholar  

He, B. J., Zhao, Z. Q., Shen, LDu., Wang, H. B. & Li, L. G. An approach to examining performances of cool/hot sources in mitigating/enhancing land surface temperature under different temperature backgrounds based on landsat 8 image. Sustain. Cities Soc. 44 , 416–427 (2019).

Lai, D., Liu, W., Gan, T., Liu, K. & Chen, Q. A review of mitigating strategies to improve the thermal environment and thermal comfort in urban outdoor spaces. Sci. Total Environ. 661 , 337–353. https://doi.org/10.1016/J.SCITOTENV.2019.01.062 (2019).

Article   ADS   CAS   PubMed   Google Scholar  

Zhang, J. Khoshbakht, M. Liu, J. Gou, Z. Xiong, J. & Jiang, M. A clustering review of vegetation-indicating parameters in urban thermal environment studies towards various factors. J. Thermal Biol. 110 , 103340. https://doi.org/10.1016/j.jtherbio.2022.103340 (2022).

Schatz, J. & Kucharik, C. Urban climate effects on extreme temperatures in Urban climate effects on extreme temperatures in Madison. Environ. Res. Lett. 10 , 1 (2015).

Founda, D. & Santamouris, M. Synergies between Urban Heat Island and Heat Waves in Athens (Greece), during an extremely hot summer (2012). Sci. Rep. 7 , 10973 (2017).

Article   ADS   PubMed   PubMed Central   Google Scholar  

Zhao, L. et al. Interactions between urban heat islands and heat waves. Environ. Res. Lett. 13 , 034003 (2018).

Raymond, C. M. et al. A framework for assessing and implementing the co-benefits of nature-based solutions in urban areas. Environ. Sci. Policy 77 , 15–24 (2017).

Kabisch, N. et al. Nature-based solutions to climate change mitigation and adaptation in urban areas: Perspectives on indicators, knowledge gaps, barriers, and opportunities for action. Ecol. Soc. 21 , art39 (2016).

Bowler, D. E., Buyung-Ali, L., Knight, T. M. & Pullin, A. S. Urban greening to cool towns and cities: A systematic review of the empirical evidence. Landsc. Urban Plan. 97 , 147–155 (2010).

Ward, K., Lauf, S., Kleinschmit, B. & Endlicher, W. Heat waves and urban heat islands in Europe: A review of relevant drivers. Sci. Total Environ. 569–570 , 527–539 (2016).

Article   ADS   PubMed   Google Scholar  

Luo, F., Yang, Y., Zong, L. & Bi, X. The interactions between urban heat island and heat waves amplify urban warming in Guangzhou, China: Roles of urban ventilation and local climate zones. Front. Environ. Sci. 11 , 1–13 (2023).

Gabriel, K. M. A. & Endlicher, W. R. Urban and rural mortality rates during heat waves in Berlin and Brandenburg. Germany. Environ. Pollut. 159 , 2044–2050 (2011).

Article   CAS   PubMed   Google Scholar  

Ramamurthy, P. & Bou-Zeid, E. Heatwaves and urban heat islands: A comparative analysis of multiple cities. J. Geophys. Res. Atmos. 122 , 168–178 (2017).

Landeshauptstadt Hannover. Strukturdaten der Stadtteile und Stadtbezirke 2022 . (2023).

Landeshauptstadt Hannover. Bevölkerungsprognose 2019 bis 2030. https://www.hannover.de/content/download/814425/file/Band 134 - Bevölkerungsprognose 2019 bis 2030.pdf (2020).

German Weather Service (Deutscher Wetterdienst DWD). Deutscher Klimaatlas. https://www.dwd.de/DE/klimaumwelt/klimaatlas/klimaatlas_node.html (2023).

German Weather Service (Deutscher Wetterdienst DWD). Climate Data Center. https://cdc.dwd.de/portal/ (2023).

German Weather Services (Deutscher Wetterdienst DWD) Regonales Klimabüro Hannover. Untersuchung zum Stadtklima der Landeshauptstadt Hannover . (2022).

Scott, A. A., Waugh, D. W. & Zaitchik, B. F. Reduced Urban Heat Island intensity under warmer conditions. Environ. Res. Lett. 13 , 064003 (2018).

EEA (European Environment Agency). Land Use/Land Cover 2018 of Functional Urban Area – DE013L1_Hannover – DE (DE013L1_HANNOVER_UA2018). https://land.copernicus.eu/local/urban-atlas/urban-atlas-2018 (2020).

LGLN (Landesamt für Geoninformation und Landesvermessung Niedersachsen). Digitales Orthophoto (DOP). Details. https://opengeodata.lgln.niedersachsen.de/#dop (2022).

LGLN (Landesamt für Geoinformation und Landesvermessung Niedersachsen). 3D-Gebäudemodell (LoD1). https://opengeodata.lgln.niedersach-sen.de/#lod1 (2021).

Gascon, M. et al. Normalized difference vegetation index (NDVI) as a marker of surrounding greenness in epidemiological studies: The case of Barcelona city. Urban For. Urban Green. 19 , 88–94 (2016).

Zhang, J. et al. A clustering review of vegetation-indicating parameters in urban thermal environment studies towards various factors. J. Therm. Biol. 110 , 103340 (2022).

Zhang, J., Gou, Z. & Shutter, L. Effects of internal and external planning factors on park cooling intensity: Field measurement of urban parks in gold coast. Australia. AIMS Environ. Sci. 6 , 417–434 (2019).

Kraemer, R. & Kabisch, N. Parks under stress: Air temperature regulation of urban green spaces under conditions of drought and summer heat. Front. Environ. Sci. 10 , 1 (2022).

Fenner, D., Holtmann, A., Meier, F., Langer, I. & Scherer, D. Contrasting changes of urban heat island intensity during hot weather episodes. Environ. Res. Lett. 14 , 124013 (2019).

Zhang, J., Gou, Z., Zhang, F. & Yu, R. The tree cooling pond effect and its influential factors: A pilot study in Gold Coast, Australia. Nature-Based Solut. 3 , 100058 (2023).

Meltzer, L. Consideration of Climate Change in the Design of Parks and Open Spaces. A study within the framework of the INTERREG IVC Project ‘Hybrid Parks’. (2014).

Kabisch, N., Korn, H., Stadler, J. & Bonn, A. Nature-Based Solutions to Climate Change Adaptation in Urban Areas—Linkages between Science, Policy and Practice . (Springer Nature, 2017). https://doi.org/10.1007/978-3-319-56091-5 .

Zhou, D., Zhao, S., Zhang, L., Sun, G. & Liu, Y. The footprint of urban heat island effect in China. Sci. Rep. 5 , 2–12 (2015).

Google Scholar  

Download references

Acknowledgements

We would like to thank DEUTSCHER WETTERDIENST (DWD) with the Department of Climate and environment consultancy, in particular Gabriele Krugmann for kindly providing the data of the climate measurement campaign in Hannover. We would like to thank Rupert Legg for polishing the English of the manuscript.

Open Access funding enabled and organized by Projekt DEAL.

Author information

Authors and affiliations.

Institute for Physical Geography and Landscape Ecology, Leibniz University Hannover, Schneiderberg 50, 30167, Hannover, Germany

Nadja Kabisch, Finja Remahne, Clara Ilsemann & Lukas Fricke

You can also search for this author in PubMed   Google Scholar

Contributions

N.K.: conceptualization, original draft preparation; data analysis. C.I.: data analysis, figure preparation. F.R. data preparation and analysis, L.F.: conceptualisation, data preparation. All authors reviewed the manuscript.

Corresponding author

Correspondence to Nadja Kabisch .

Ethics declarations

Competing interests.

The authors declare no competing interests .

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Kabisch, N., Remahne, F., Ilsemann, C. et al. The urban heat island under extreme heat conditions: a case study of Hannover, Germany. Sci Rep 13 , 23017 (2023). https://doi.org/10.1038/s41598-023-49058-5

Download citation

Received : 15 July 2023

Accepted : 04 December 2023

Published : 27 December 2023

DOI : https://doi.org/10.1038/s41598-023-49058-5

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.