#
Effects of Climate Change on Precipitation and the Maximum Daily Temperature (T_{max}) at Two US Military Bases with Different Present-Day Climates

^{*}

## Abstract

**:**

_{max}) at two USA locations that have different climates—the Travis Airforce Base (AFB) in California [38.27° N, 121.93° W] and Fort Bragg (FBR) in North Carolina [35.14 N, 79.00 W]—are analyzed. The effects of climate change on central tendency, tail distributions, and both auto- and cross-covariance structures in precipitation and T

_{max}fields for three time periods in the 21st century centered on the years 2020, 2050, and 2100 were analyzed. It was found that, on average, T

_{max}under the Representative Concentration Pathway (RCP) 4.5 emission scenario is projected to increase for the years 2020, 2050, and 2100 by 1.1, 2.0, and 2.2 °C, respectively, for AFB, and 0.9, 1.2, and 1.6 °C, respectively, for FBR, while under the RCP8.5 emission scenario T

_{max}will increase by 1.1, 1.9, and 2.7 °C, respectively, for AFB, and 0.1, 1.5, and 2.2 °C, respectively, for FBR. The climate change signal in precipitation is weak. The results show that, under different emission scenarios, events considered to be within 1% of the most extreme events in the past will become ~13–30 times more frequent for T

_{max}, ~and 0.05–3 times more frequent for precipitation in both locations. Several analytical methods were deployed in a sequence, creating an easily scalable framework for similar analyses in the future.

## 1. Introduction

_{max}) and precipitation and their possible correlations. The second goal is to pave a road for future similar analytical efforts, which are likely to include more locations due to the scalability of the utilized framework, as well as to point to common challenges in performing this type of analysis to assure easy reproducibility. The roadmap of the entire research is shown in Figure 1. The inputs are observational T

_{max}and precipitation data, both of which were prepared by Livneh et al. (2013) [33], cover the period 1915–2011 and the downscaled general circulation model outputs (32 models) of the same variables given at the same high-resolution grid (approximately 6 km) for the period 1950–2100.

## 2. Methodology and Materials

#### 2.1. General Approach

_{max}and precipitation fields within 30-year-wide windows, for four periods, is assumed: 1985–2015 (the baseline period), 2005–2035, 2035–2065, and 2085–2115. In case of precipitation, this assumption is relatively easy to justify, as even at the larger timescales (on the order of 100 years) the climate change signal is weak and unclear, as is shown in this study. For T

_{max}, the assumption of pseudostationarity is harder to defend, as it has been well established that the Earth system is getting warmer. Yet, apart from statistical tests, there are few arguments for skipping the detrending step within each of the time windows used. First, discretized empirical cumulative distribution functions (CDFs), not parametric ones, are used to derive the relationships between historical-modeled and observational datasets and project observations in the future. Thus, no assumption about the distribution type of the residuals or trend type had to be made, i.e., the effects of potential trends are preserved within empirical CDFs. Second, all data within each 30-year wide window are treated as integral populations, regardless of the chronology of the appearance of extreme values within selected windows. In other words, data are treated as sets rather than sequences within each window. The potential presence of trends within time-windows is irrelevant for the purpose of our analysis. Furthermore, in an ideal scenario, models should capture trends as well, which would be subsequently encoded in the discretized empirical CDFs of modeled data, used for bias-correction. Consequently, CDFs that all include supposedly the same trend are, actually, used. Finally, the fact that there is one dimension (in this case temporal) inside the populations for which it was suspected that the data exhibit trending behavior is, per se, not a sufficient argument for detrending as it depends on the specific purpose and the type of analysis. The topic of interest is the overall population statistics within predefined temporal windows, not chronology, inside the windows or the tendency of the extreme values to group in any subspace of the windows. Thereafter, discretized empirical CDFs provide a sufficient and complete description of the population statistics required for applying the quantile mapping family of methods. Eventually, to preserve a rigorous scientific approach, an original two tailed Mann–Kendall test of the trend null hypothesis is performed. The result clearly shows that, at the level of significance of 0.05, there is no trend in the historical observations of T

_{max}within a 30-year window at AFB (p = 0.58) and FBR (p = 0.97), and thus the null hypothesis is not rejected. The scatter plots and linear fits are shown in Figure S1.

_{max}, absolute (difference), and for precipitation relative (ratio), climate change signals per quantile are extracted. Then, extracted climate change signals are used to project the observational dataset into the future. Next, tail-distributions using the Generalized Extreme Value theory (GEV) are analyzed. Annual maximums for T

_{max}and precipitation are extracted and their distributions separately modeled (Section 3.4). Finally, the copula theory is deployed to model the autocorrelation and cross-correlation fields at both AFB and FBR locations. Copulas are sampled using the Monte-Carlo approach to assess the risk of having collocated or coincident extreme events in both precipitation and T

_{max}fields (Section 2.5). The theoretical backgrounds of the applied techniques are given in Section 2.3.

#### 2.2. Observational Datasets

_{max}datasets prepared by Livneh et al. (2013) are used [33], downloaded from ftp://ftp.cdc.noaa.gov/Projects/Datasets/livneh/metvars. It is a publicly available, long-term (1915–2011), hydrologically-consistent dataset for the conterminous United States, intended to aid in the study of water and energy exchanges at the land’s surface. The data are provided for a 1/16th degree grid (approximately 6 km, or approximately 3.75 miles), which matches the grid in which the modeled data are given (see Section 2.3). The reported data are derived from daily temperature and precipitation observations from approximately 20,000 NOAA Cooperative Observer (COOP) stations. Data from the grid cells that are closest to the target locations (2.6 miles from AFB, and 1.8 miles from FBR) are used as representatives of the data at target locations, because the target locations are very close to the grid cells and the intention is to avoid additional smoothing effects that are usually caused by interpolation attempts (e.g., kriging).

#### 2.3. Global Circulation Models

#### 2.4. Localized Constructed Analogs (LOCA)

_{max}data (given at 1/16th degree (approximately 6 km, or approximately 3.75 miles)) were downloaded for the entire US (ftp://gdo-dcp.ucllnl.org/pub/dcp/archive/cmip5/loca/LOCA_2016-04-02/), and regional subsets were selected around two points of interest (+/− 2° in latitude and longitude), preserving only on-land data. The data from the most proximate grid cell are used as representatives of the target locations, following the same procedure that was used with the observational data (see Section 2.1).

#### 2.5. Theory

#### 2.5.1. Bhattacharyya Coefficient

_{i}and and q

_{i}are the numbers of members of samples p and q in the i-th partition, D

_{B}is the Bhattacharyya distance, and BC is the Bhattacharyya coefficient (possible range of values for BC is 0–1, and for D

_{B}0–∝).

_{max}and precipitation, in the past and future for both emission scenarios. Ideally, the PDFs that stand out in terms of the similarity to others (as it is expected that PDFs contain redundant information content) should also have either higher or lower associated beta coefficients in the MLR reconstruction of the observational CDFs using modeled CDFs, as they either bring in additional, unique information not captured by other models or they less correctly represent observed reality. Thus, the set of Bhattacharyya coefficients that represents the similarities between PDFs of all model ensemble members can be seen as a measure of consistency between different model outputs.

#### 2.5.2. Quantile Delta Mapping (QDM)

_{max}and precipitation fields within 30-year wide windows is assumed for both reference and future periods, and intra-window variability is treated as stochastic, so no long-term trends are eliminated from the data. In addition, all data that fall within the windows are modeled at once, without splitting the data into smaller chunks (e.g., months). Consequently, PDFs (and CDFs) are treated as 30-year-wide window-specific, so no detrending is involved. Discretized observational CDFs are reproduced using MLR and future “observational” CDFs are produced with future modeled CDFs while assuming stationarity of MLR beta coefficients. Thus, CDFs that result from two MLR approaches are used, one for the past and one for the future, to extract climate change signals using the same approach as Cannon et al. (2015) [23]. To preserve absolute (for T

_{max}) rather than relative (for precipitation) changes in quantiles, the climate change signal is applied additively rather than multiplicatively to create future observational equivalents.

#### 2.5.3. Generalized Extreme Value (GEV) Theory

_{max}represents a truly random variable, while precipitation represents a random variable if it is non-zero. In practice, modeling the GEV distribution of extreme events assumes taking block maxima (or minima) and fitting the GEV distribution to data.

_{max}and precipitation over 30-year-wide windows are used for the reference baseline period (1985–2015) and for future periods centered around the years 2020, 2050, and 2100 to recode for both emission scenarios. Then, fit GEV distributions are fitted to the resulting subset of data. Then, probabilities for resulting extreme values that exceed chosen thresholds are estimated (see Section 3.4).

#### 2.5.4. Copula Theory

_{max}and (non-zero) precipitation represent continuous random variables and thus fulfill the requirements for modeling using copulas.

## 3. Results and Discussion

#### 3.1. Bhattacharyya Coefficients and Model Ensemble Member Consistency

_{max}). It is observed that, for example, model CESM1-CAM5 shows that its past modeled precipitation PDF is similar to the observations (Figure 2a), but for all three future periods under the RCP8.5 emission scenario (Figure 2b–d) the modeled precipitation PDF differs greatly from the PDF of observations in the baseline period. In other words, this model implies that climate change under RCP8.5 emission scenario would bring pronounced changes in the distribution of precipitation in all three future periods.

_{max}similarly shows large differences between baseline and future, showing the most pronounced changes in T

_{max}distribution in 2015–2035 and 2035–2065, while the 2085–2115 period shows the most dissimilar PDFs to T

_{max}observational PDFs in the baseline period.

_{max}(Figure 3c,d), where only inmcm4 seems to be somewhat different under the RCP4.5 emission scenario, and CanESM2 under the RCP8.5 scenario for the 2035–2065 period.

#### 3.2. MLR and QDM

_{max}are reconstructed using 32 model CDFs and MLR machinery. Figure 4 shows two bar plots (Figure 4a,b) for precipitation and T

_{max}, with MLR beta coefficients (and intercepts) for all 32 models, together with plots (Figure 4c,d) showing how good the fits are (turquoise lines are not visible due to prefect overlap).

_{max}(1985–2011) are projected into the future using the Quantile Delta Mapping approach to extract the information regarding the climate change signal. This is done by comparing values on past (reconstructed) and future (constructed) CDFs, quantile-by-quantile, using MLR. The technique leads to a set of “observations” in the future, different within every time period (2020, 2050, and 2100, +/−15 years).

_{max}for AFB, with clearly visible changes in central tendency and the distributions induced by climate change in T

_{max}, while precipitation PDFs have less apparent changes. Table 1 shows the changes, in the form of empirical PDFs, in the central tendencies of both precipitation and T

_{max}at both AFB and FBR locations, for both emission scenarios, and for all three future periods of interest. Based on the data presented in Table 1, future weather brings higher maximum daily temperatures, between +2.2 and +2.7 °C for AFB and +1.6 and +2.2 °C for FBR, under the RCP4.5 and RCP8.5 emission scenarios, respectively. Under two emission scenarios, both locations exhibit similar differences in T

_{max}shifts in year 2100: ~0.5 °C. The change in precipitation, however, does not show a clear pattern. For the FBR location and both emission scenarios, the future weather brings less precipitation. For the AFB location, the scenarios suggest precipitation will stay more or less the same, with slightly higher precipitation in the RCP8.5 scenario.

#### 3.3. Modeling Tail Distributions Using Generalized Extreme Value (GEV) Theory

_{max}at both locations under the RCP4.5 emission scenario.

_{max}maxima for both AFB and FBR locations extracted from observations for the 1985–2011 period [33]. The uncertainty is given as the GEV parameter fitting uncertainty and does not include the uncertainty stemming from inaccuracies in the model itself (residual bias, MLR beta non-stationarity, etc.). The values from Table 2 are used as a baseline for comparison with future values.

_{max}maxima, same as for the past observational datasets, for both locations and both emission scenarios. The probabilities of heat waves are systematically increasing with temporal distance into the future for both emission scenarios. Thus, the 1-in-100-years event for the year 2100, in terms of annual maximum T

_{max}, shifted from above 318.3K (AFB) in the baseline period to above 320.6 K in the RCP4.5 and above 321.0 K in the RCP8.5 scenario, which is consistent with the change in the mean T

_{max}values of +2.2 and +2.7 K in the same period for the same emission scenarios (Table 1). Very similar consistency is found for the FBR location between the shift in mean value and the values corresponding to the chosen thresholds in GEV distribution, which reconfirms the robustness of the GEV approach, given that consistent results were obtained using different analytical procedures. On the other hand, the distribution of annual precipitation maxima shifts differently compared to the mean values. For example, the AFB location will have lower magnitude extreme precipitation events (p < 0.01) under the RCP4.5 emission scenario (+0.7%, −10% and −13% for the years 2020, 2050, and 2100, respectively) compared to the baseline, and higher magnitude under the RCP8.5 emission scenario (+15%, +7% and +18% for the years 2020, 2050, and 2100, respectively) compared to the baseline.

_{max}in the past. The results show that both AFB and FBR locations will witness events that, in the past, were considered to be in the 1% most extreme T

_{max}events ~13–30 times more often. For precipitation, the range found is ~0.05–3 times more often (see Table 4 for details).

#### 3.4. Copulas

_{max}. Two locations examined in this study are ~4000 km apart, and the intuitive expectation of having correlated extreme events is low, especially given that precipitation seasonality at these locations is very different. In addition, especially in CA (AFB), precipitation extremes happen in winter months, so the cooccurrence of the two extremes is unlikely. However, as stated earlier, one of the goals of this study was to provide a rigorous, quantitative, and scalable framework for similar analyses in the future, which would likely include more than two locations and perhaps more variables, where analogous arguments based on local meteorology do not necessarily hold. Three tasks were defined: (1) Using empirical copulas, model autocorrelation structures in T

_{max}and precipitation fields for AFB and FBR locations and for past and future periods. (2) Using empirical copulas, model cross-correlation structures between T

_{max}and precipitation fields for both locations in past and future time periods. (3) Check the trend in autocorrelation structures to detect eventual sensitivity to climate change.

_{max}, it is easy, e.g., to use CDFs and, instead of using original data, to use their quantiles. However, in the case of precipitation, due to multiple zero values, even if CDFs are used, the resulting distribution of data is not flat, and copula theory cannot be applied directly. Instead, only non-zero precipitation values are used to create copulas, and later the probabilities that precipitation will be non-zero when assessing the probabilities of joint events are separately taken into account.

#### 3.4.1. T_{max} Autocorrelation at AFB and FBR

_{max}autocorrelation by analyzing simultaneous distributions of T

_{max}values in the observational dataset at AFB and FBR. Prior to creating copula structures, T

_{max}observations are converted into uniformly distributed pseudo-observations using a rank-based approach. Figure 7a shows the marginal and joint distributions of T

_{max}observations at AFB and FBR, 7b shows a 2D histogram in tile format, 7c shows the marginal and joint distribution of pseudo-observations, and 7d shows the empirical copulas fitted into a joint distribution of pseudo-observations calculated using a 0.03 kernel bandwidth and copula densities evaluated at 100 equi-spaced points in [0,1] range along both axes. A copula is a hyperdimensional structure, and the same approach could be used for more than two locations, if necessary.

_{max}pairs at both locations, which obey the modeled copula structure generated, and these pseudo-observations are subjected to statistical analysis.

_{max}occurring only once in 20 days at both locations is 0.5872% (which means that on ~two days for every year, T

_{max}recorded at AFB and FBR will be among the 5% warmest days recorded at both sites). It is found that the probability for T

_{max}occurring only once in 50 days at both locations is 0.0796% (which means that on one day for every ~five years, T

_{max}recorded at AFB and FBR will be among the 2% warmest days recorded at both sites). It is found that the probability for T

_{max}occurring only once in 100 days at both locations is 0.0135% (which means that on one day for every ~20 years, T

_{max}recorded at AFB and FBR will be among the 1% warmest days recorded at both sites). The probability that T

_{max}occurring only once in 365 days (annual maximum) at both locations is 0.0037% (which means that on one day every ~74 years, T

_{max}recorded at AFB and FBR will be the actual annual T

_{max}maximum recorded at both sites). The probability that both sites could experience a 1-in-20-years event simultaneously is essentially zero for these two sites.

_{max}values. In this analysis, it is assumed that the copula based on actually observed values is stationary and will not change even if never-observed (no-analog-day) events happened. It is concluded that the risk of correlated heat waves at a 1-in-20 intensity at these two sites was negligible in 1985–2015.

_{max}events for those two locations.

#### 3.4.2. Precipitation Autocorrelation at AFB and FBR

^{−4}% (which means that one day every ~250 years precipitation recorded at AFB and FBR will be among 1% highest precipitations recorded at both sites).

^{−4}% (which means that one day every ~850 years precipitations recorded at AFB and FBR will be the annual precipitation maximum recorded at both sites). The probability that 1-in-20-years precipitation events will occur on the same day at these two sites is essentially zero.

#### 3.4.3. T_{max}/Precipitation Cross-Correlation at AFB

_{max}events would occur at AFB and FBR sites were examined separately. Even at a first glance, it looks very unlikely that those two extremes will co-occur, as the rainy season at AFB is happening in the winter, and, in addition to this, rain exhibits a cooling effect, but the analytical steps described above are repeated. Figure 9 shows the marginal and joint distributions of non-zero precipitation events and the corresponding T

_{max}at AFB location during 1985–2011, as well as a 2D histogram, the marginal and joint distributions of their pseudo-observations, and the modeled copula. It was found that the probability that precipitation and the corresponding T

_{max}that happen only once in 20 days is 0.0112% (which means that one day every ~25 years precipitation and T

_{max}recorded at AFB will be among 5% highest precipitations and T

_{maxs’}recorded). It was found that the probability that precipitation and corresponding T

_{max}that happen only once in 50 days is 0.001% (which means that one day every ~266 years precipitation and T

_{max}recorded at AFB will be among 2% highest precipitations and T

_{max}recorded). Finally, it was found that the probability that precipitation and the corresponding T

_{max}that happen only once in 100 days is 1.321510

^{−4}% (which means that one day every ~2100 years precipitation and corresponding T

_{max}recorded at AFB will be among 1% highest precipitations recorded at that site). The probability that precipitation that happens only once in 365 days (annual maximum) is 3.160210

^{−5}%, which means that one day every ~8700 years precipitations recorded at AFB will be the maxima for annual precipitation and the corresponding T

_{max}. Given that the rainy season in CA happens in winter, this finding is not surprising. The probability that 1-in-20-years events will occur on the same day is zero.

#### 3.4.4. T_{max}/Precipitation Cross-Correlation at FBR

_{max}that happen only once in 20 days is 0.1073% (which means that one day every ~2.5 years precipitation and T

_{max}recorded at FBR will be among 5% highest precipitations and T

_{max}recorded). It was found that the probability that precipitation and the corresponding T

_{max}that happen only once in 50 days is 0.0124% (which means that one day every ~22 years precipitation and T

_{max}recorded at FBR will be among 2% highest precipitations and T

_{max}recorded). It was found that the probability for precipitation and the corresponding T

_{max}that happen only once in 100 days is 0.0018% (which means that one day every ~150 years precipitation and T

_{max}recorded at FBR will be among 1% highest precipitations recorded at both sites).

^{−4}% (which means that one day every ~610 years precipitation recorded at FBR will be the maxima for annual precipitation and T

_{max}. Interestingly, the probability, despite being low, is much higher than the probability of a co-occurrence of these events at AFB. This is consistent with the fact that precipitation is not as strictly associated with the winter season at FBR, as in the case of AFB. The probability that 1-in-20-years events will occur on the same day is essentially zero.

## 4. Conclusions

_{max}and precipitation fields at two locations, AFB and FBR, which represent locations of first-grade military facilities were evaluated. The immediate goal was to inform infrastructural planning, and the second goal was to come up with a robust, reproducible set of methodological steps that could be deployed to analyze effects of climate change on additional or different locations.

_{max}is projected to increase on average by 2.2–2.7 °C at AFB and 1.6–2.2 °C at FBR in the years 2085–2115, based on the combined results from two emission scenarios. Given that the value is lower than the change in average surface temperature on Earth (e.g., [51]), it implies that T

_{max}and T

_{average}distributions are projected to change in slightly different ways, which is beyond the scope of this study. It was found that climate change will not induce significant changes in auto- and cross-correlation structures in precipitation and T

_{max}fields at two studied locations.

_{max}is ~13–30 and 0.05–3 for precipitation.

## Supplementary Materials

_{max}from historical observations for AFB(a) and FBR(b) (Livneh et al. (2013)) along with the best linear fit, Figure S2: Monterey County, CA historical and future temperature trajectories from Cal-Adapt.org, Table S1: List of downscaled CMIP5 GCMs used in the study.

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## References

- Parry, M.L.; Rosenzweig, C.; Iglesias, A.; Livermore, M.; Fischer, G. Effects of climate change on global food production under SRES emissions and socio-economic scenarios. Glob. Environ. Chang. A
**2004**, 14, 53–67. [Google Scholar] [CrossRef] - Patz, J.A.; Campbell-Lendrum, D.; Holloway, T.; Foley, J.A. Impact of regional climate change on human health. Nature
**2005**, 438, 310–317. [Google Scholar] [CrossRef] [PubMed] - Milly, P.C.D.; Betancourt, J.; Falkenmark, M.; Hirsch, R.M.; Kundzewicz, Z.W.; Lettenmaier, D.P.; Stouffer, R.J. Climate change. Stationarity is dead: Whither water management? Science
**2008**, 319, 573–574. [Google Scholar] [CrossRef] [PubMed] - Schweikert, A.; Chinowsky, P.; Espinet, X.; Tarbert, M. Climate Change and Infrastructure Impacts: Comparing the Impact on Roads in ten Countries through 2100. Procedia Eng.
**2014**, 78, 306–316. [Google Scholar] [CrossRef][Green Version] - Bulkeley, H. Cities and governing of climate change. Annu. Rev. Environ. Res.
**2010**, 35, 229–253. [Google Scholar] [CrossRef][Green Version] - Bellard, C.; Bertelsmeier, C.; Leadley, P.; Thuiller, W.; Courchamp, F. Impacts of climate change on the future of biodiversity. Ecol. Lett.
**2012**, 15, 365–377. [Google Scholar] [CrossRef][Green Version] - Harris, J.M.; Roach, B.; Environmental, J.M.H. The Economics of Global Climate Change; Global Development and Environment Institute, Tufts University: Massachusetts, MA, USA, 2015. [Google Scholar]
- Urry, J. Climate Change and Society; UK Polity Press: Cambridge, UK, 2011. [Google Scholar]
- Moss, R.H.; Kravitz, B.; Delgado, A.; Asrar, G.; Brandenbeger, J.; Wigmosta, M.; Preston, K.; Buzan, T.; Gremillion, M.; Shaw, P.; et al. Nonstationarity RC Workshop Report: Nonstationary Weather Patterns and Extreme Events Informing Design and Planning for Long-Lived Infrastructure; Technical Report; Pacific Northwest National Lab: Richland, WA, USA, 2017. [Google Scholar]
- Brekke, L.D.; Dettinger, M.D.; Maurer, E.P.; Anderson, M. Significance of model credibility in estimating climate projection distributions for regional hydroclimatological risk assessments. Clim. Chang.
**2008**, 89, 371–394. [Google Scholar] [CrossRef] - Van Vuuren, D.P.; Edmonds, J.; Kainuma, M.; Riahi, K.; Thomson, A.; Hibbard, K.; Hurtt, G.C.; Kram, T.; Krey, V.; Lamarque, J. The representative concentration pathways: An overview. Clim. Chang.
**2011**, 109, 5. [Google Scholar] [CrossRef] - Taylor, K.E.; Stouffer, R.J.; Meehl, G.A. An overview of CMIP5 and the experiment design. Bull. Am. Meteorol. Soc.
**2012**, 93, 485–498. [Google Scholar] [CrossRef][Green Version] - Andrews, T.; Gregory, J.M.; Webb, M.J.; Taylor, K.E. Forcing, feedbacks and climate sensitivity in CMIP5 coupled atmosphere-ocean climate models. Geophys. Res. Lett.
**2012**, 39, L09712. [Google Scholar] [CrossRef][Green Version] - Ntegeka, V.; Baguis, P.; Roulin, E.; Willems, P. Developing tailored climate change scenarios for hydrological impact assessments. J. Hydrol.
**2014**, 508, 307–321. [Google Scholar] [CrossRef] - Barros, A.P.; Lettenmaier, D.P. Dynamic modeling of orographically induced precipitation. Rev. Geophys.
**1994**, 32, 265–284. [Google Scholar] [CrossRef] - Leung, L.R.; Ghan, S.J. A subgrid parameterization of orographic precipitation. Theor. Appl. Climatol.
**1995**, 52, 95–118. [Google Scholar] [CrossRef] - Gleckler, P.J.; Taylor, K.E.; Doutriaux, C. Performance metrics for climate models. J. Geophys. Res.
**2008**, 113, D06104. [Google Scholar] [CrossRef] - Rupp, D.E.; Abatzoglou, J.T.; Hegewisch, K.C.; Mote, P.W. Evaluation of CMIP5 20th century climate simulations for the Pacific Northwest USA. J. Geophys. Res. Atmos.
**2013**, 118, 10884–10906. [Google Scholar] [CrossRef] - Pierce, D.W.; Cayan, D.R.; Thrasher, B.L. Statistical downscaling using localized constructed analogs (LOCA). J. Hydrometerol.
**2014**, 15, 2558–2586. [Google Scholar] [CrossRef] - Christensen, J.H.; Boberg, F.; Christensen, O.B.; Lucas-Picher, P. On the need for bias correction of regional climate change projections of temperature and precipitation. Geophys. Res. Lett.
**2008**, 35, L20709. [Google Scholar] [CrossRef] - Mearns, L.O.; Arritt, R.; Biner, S.; Bukovsky, M.S.; McGinnis, S.; Caya, S.S.D.; Correia, J., Jr.; Flory, D.; Gutowski, W.; Takle, E.S.; et al. The North American Regional Climate Change Assessment Program: Overview of phase I results. Bull. Am. Meteor. Soc.
**2012**, 93, 1337–1362. [Google Scholar] [CrossRef] - Sillmann, J.; Kharin, V.V.; Zhang, X.; Zwiers, F.W.; Bronaugh, D. Climate extremes indices in the CMIP5 multimodel ensemble: Part 1. Model evaluation in the present climate. J. Geophys. Res. Atmos.
**2013**, 118, 1716–1733. [Google Scholar] [CrossRef] - Cannon, A.J.; Sobie, S.R.; Murdock, T.Q. Bias Correction of GCM Precipitation by Quantile Mapping: How Well Do Methods Preserve Changes in Quantiles and Extremes? J. Clim.
**2015**, 28, 6938–6959. [Google Scholar] [CrossRef] - Ehret, U.; Zehe, E.; Wulfmeyer, V.; Warrach-Sagi, K.; Liebert, J. HESS opinions: “Should we apply bias correction to global and regional climate model data?”. Hydrol. Earth Syst. Sci.
**2012**, 16, 3391–3404. [Google Scholar] [CrossRef][Green Version] - Maraun, D. Nonstationarities of regional climate model biases in European seasonal mean temperature and pre- cipitation sums. Geophys. Res. Lett.
**2012**, 39, L06706. [Google Scholar] [CrossRef][Green Version] - Maraun, D. Bias correction, quantile mapping, and downscaling: Revisiting the inflation issue. J. Clim.
**2013**, 26, 2137–2143. [Google Scholar] [CrossRef][Green Version] - Gudmundsson, L.; Bremnes, J.B.; Haugen, J.E.; Engen-Skaugen, T. Technical note: Downscaling RCM precipitation to the station scale using statistical transformations—A comparison of methods. Hydrol. Earth Syst. Sci.
**2012**, 16, 3383–3390. [Google Scholar] [CrossRef][Green Version] - Teutschbein, C.; Seibert, J. Bias correction of regional climate model simulations for hydrological climate-change impact studies: Review and evaluation of different methods. J. Hydrol.
**2012**, 456–457, 12–29. [Google Scholar] [CrossRef] - Teutschbein, C.; Seibert, J. Is bias correction of regional climate model (RCM) simulations possible for non-stationary conditions? Hydrol. Earth Syst. Sci.
**2013**, 17, 5061–5077. [Google Scholar] [CrossRef][Green Version] - Hagemann, S.; Chen, C.; Haerter, J. Impact of a statistical bias correction on the projected hydrological changes obtained from three GCMs and two hydrology models. J. Hydrometeorol.
**2011**, 12, 556–578. [Google Scholar] [CrossRef] - Themeßl, M.J.; Gobiet, A.; Heinrich, G. Empirical- statistical downscaling and error correction of re- gional climate models and its impact on the climate change signal. Clim. Chang.
**2012**, 112, 449–468. [Google Scholar] [CrossRef] - Maurer, E.P.; Pierce, D.W. Bias correction can modify climate model simulated precipitation changes without adverse effect on the ensemble mean. Hydrol. Earth Syst. Sci.
**2014**, 18, 915–925. [Google Scholar] [CrossRef][Green Version] - Livneh, B.; Rosenberg, E.A.; Lin, C.; Nijssen, B.; Mishra, V.; Andreadis, K.M.; Maurer, E.P.; Lettenmaier, D.P. A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States: Update and extensions. J. Clim.
**2013**, 26, 9384–9392. [Google Scholar] [CrossRef] - Haas, T.C. Lognormal and moving window methods of estimating acid deposition. J. Am. Stat. Assoc.
**1990**, 85, 950–963. [Google Scholar] [CrossRef] - Tadić, J.M.; Qiu, X.; Miller, S.; Michalak, A.M. Spatio-temporal approach to moving window block kriging of satellite data v1.0. Geosci. Model Dev.
**2017**, 10, 709–720. [Google Scholar] [CrossRef][Green Version] - California Water Commission Techical Reference. 2016. Available online: https://water.ca.gov/LegacyFiles/economics/downloads/EAS_Docs/WSIP_Draft_TechRefDoc_Compiled.pdf (accessed on 21 January 2020).
- Mauritsen, T.; Stevens, B.; Roeckner, E.; Crueger, T.; Esch, M.; Giorgetta, M.; Haak, H.; Jungclaus, J.; Klocke, D.; Matei, D.; et al. Tuning the climate of a global model. J. Adv. Modeling Earth Syst.
**2012**, 4. [Google Scholar] [CrossRef] - Kharin, V.V.; Zwiers, F.W.; Zhang, X.; Wehner, M. Changes in temperature and precipitation extremes in the CMIP5 ensemble. Clim. Chang.
**2013**, 119, 345–357. [Google Scholar] [CrossRef] - Van den Dool, H.M. Searching for analogues, how long must we wait? Tellus
**1994**, 46A, 314–324. [Google Scholar] [CrossRef] - Wood, A.W.; Leung, L.R.; Sridhar, V.; Lettenmaier, D.P. Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs. Clim. Chang.
**2004**, 62, 189–216. [Google Scholar] [CrossRef] - Maurer, E.P. The utility of daily large-scale climate data in the assessment of climate change impacts on daily streamflow in California. Hydrol. Earth Syst. Sci.
**2010**, 14, 1125–1138. [Google Scholar] [CrossRef][Green Version] - Abatzoglou, J.T.; Brown, T.J. A comparison of statis- tical downscaling methods suited for wildfire applications. Int. J. Climatol.
**2012**, 32, 772–780. [Google Scholar] [CrossRef] - Bhattacharyya, A. On a measure of divergence between two statistical populations defined by their probability distributions. Bull. Calcutta Math. Soc.
**1943**, 35, 99–109. [Google Scholar] - Coles, S. An Introduction to Statistical Modeling of Extreme Values; Springer: London, UK, 2001; ISBN 1-85233-459-2. [Google Scholar]
- Rudin, W. Principles of Mathematical Analysis; McGraw Hill: New York, NY, USA, 1976; pp. 89–90. [Google Scholar]
- Charpentier, A.; Fermanian, J.-D.; Scaillet, O. The Estimation of Copulas: Theory and Practice. In Copulas: From Theory to Applications in Finance; Rank, J., Ed.; Risk Books: London, UK, 2007; pp. 35–62. [Google Scholar]
- Schmidt, T. Coping with Copulas, in: Copulas—From Theory to Application in Finance; Risk Books: London, UK, 2007; pp. 3–34. [Google Scholar]
- Sklar, A. Fonctions de Repartition a n Dimensions et Leurs Marges; Public instruction Statistics University: Paris, France, 1959; Volume 8, pp. 229–231. [Google Scholar]
- Chen, S.X. A beta kernel estimation for density functions. Comput. Stat. Data Anal.
**1999**, 31, 131–135. [Google Scholar] [CrossRef] - Brown, B.M.; Chen, S.X. Beta-Bernstein smoothing for regression curves with compact supports. Scand. J. Stat.
**1999**, 26, 47–59. [Google Scholar] [CrossRef] - Brown, P.T.; Caldeira, K. Greater future global warming inferred from Earth’s recent energy budget. Nature
**2017**, 552, 45. [Google Scholar] [CrossRef] [PubMed]

**Figure 1.**The flow diagram of the analysis. (1) Input data (observational and model data) are first converted into empirical PDFs and CDFs. (2) PDFs and CDFs are subjected to MLR to reconstruct observational CDFs using model CDFs. Next, (3) QDM is applied to observational data using past and future CDFs. (4) Once bias-corrected future data are available, we apply GEV theory and examine change in likelihood of extreme events. Finally, (5) copula theory is applied to assess whether climate change increases risk of simultaneous or collocated extreme events.

**Figure 2.**Bhattacharyya coefficients for precipitation (

**a**–

**d**) and T

_{max}(

**e**–

**h**) modeled PDFs compared to the baseline period (1985–2015). (

**b**,

**f**) correspond to period centered around year 2020, (

**c**,

**g**) centered around 2050, and (

**d**,

**h**) centered around 2100. Two column plots (

**b**–

**d**for precipitation and

**f**–

**h**for T

_{max}) correspond to two emission scenarios, RCP4.5 (first column) and RCP8.5 (second column).

**Figure 3.**Selected Bhattacharyya coefficient plots for precipitation (

**a**,

**b**) and T

_{max}(

**c**,

**d**) PDFs inter compared between model ensemble members. (

**a**) Precipitation PDF inter-comparison for 2035–2065, RCP4.5. (

**b**) Precipitation PDF inter-comparison for 2035–2065, RCP8.5. (

**c**) T

_{max}PDF inter-comparison for 2035–2065, RCP4.5. (

**d**) T

_{max}PDF inter-comparison for 2035–2065, RCP8.5.

**Figure 4.**MLR beta coefficients (and intercepts) (

**a**,

**b**) obtained by reconstructing observational precipitation and T

_{max}CDF using 32 model outputs downscaled by LOCA, for the 1985–2011 period. Figure 4c,d shows observational and reconstructed CDFs for precipitation (

**c**) and T

_{max}(

**d**).

**Figure 5.**Exemplary CDFs for precipitation (

**a**,

**b**) and PDFs for T

_{max}(

**c**,

**d**) for the AFB location, comparing the 2035–2065 period to the baseline period.

**Figure 6.**(

**a**) GEV PDF for annual precipitation maximum values for 2035–2065, AFB, under RCP4.5 emission scenario, (

**b**) GEV PDF for annual T

_{max}maximum values for 2035–2065, AFB, under RCP4.5 emission scenario, (

**c**) GEV PDF for annual precipitation maximum values for 2035–2065, FBR, under RCP4.5 emission scenario, and (

**d**) GEV PDF for annual T

_{max}maximum values for 2035–2065, FBR, under RCP4.5 emission scenario.

**Figure 7.**(

**a**) Marginal and joint distribution of observations of T

_{max}at AFB and FBR during 1985–2011 period. (

**b**) 2D histogram shown in tile format. (

**c**) Marginal and joint distributions of T

_{max}pseudo observations, prepared as input to copula-modeling. (

**d**) Fitted empirical copula.

**Figure 8.**(

**a**) Marginal and joint distribution of precipitation observations at AFB and FBR during 1985–2011 period. (

**b**) 2D histogram shown in tile format. (

**c**) Marginal and joint distributions of precipitation pseudo-observations, prepared as input to copula-modeling. (

**d**) Fitted empirical copula.

**Figure 9.**(

**a**) Marginal and joint distribution of observations of non-zero precipitation and corresponding T

_{max}at AFB, during 1985–2011 period. (

**b**) 2D histogram shown in tile format. (

**c**) Marginal and joint distributions of non-zero precipitation and T

_{max}pseudo-observations, prepared as input to copula modeling. (

**d**) Fitted empirical copula.

**Figure 10.**(

**a**) Marginal and joint distribution of observations of non-zero precipitation and corresponding T

_{max}at FBR, during 1985–2011 period. (

**b**) 2D histogram shown in tile format. (

**c**) Marginal and joint distributions of non-zero precipitation and T

_{max}pseudo-observations, prepared as input to copula-modeling. (

**d**) Fitted empirical copula.

**Table 1.**Changes in the average daily precipitation and the maximum daily temperature (T

_{max}) values (future-past) for three future periods centered around the years 2020, 2050, and 2100, derived from bias-corrected future projections using Multi-Linear Regression (MLR) and past observations, for two emission scenarios (RCP4.5 and RCP8.5).

AFB | FBR | |||||||
---|---|---|---|---|---|---|---|---|

RCP4.5 | RCP8.5 | RCP4.5 | RCP8.5 | |||||

Year (Centered Around) | Precipitation (mm/day) (Delta) | T_{max} (°C) (Delta) | Precipitation (mm/Day) (Delta) | T_{max} (°C) (Delta) | Precipitation (mm/Day) (Delta) | T_{max} (°C) (Delta) | Precipitation (mm/Day) (Delta) | T_{max} (°C) (Delta) |

2020 | +0.1 | +1.1 | +0.2 | +1.1 | −0.5 | +0.9 | −0.3 | +0.1 |

2050 | −0.0 | +2.0 | +0.0 | +1.9 | −0.4 | +1.2 | −0.3 | +1.5 |

2100 | −0.1 | +2.2 | +0.0 | +2.7 | −0.5 | +1.6 | −0.4 | +2.2 |

**Table 2.**Exceedance probability thresholds and associated values taken from observations (1985–2011) of the Travis Airforce Base (AFB) and Fort Bragg (FBR) locations, based on the Generalized Extreme Value (GEV) distribution for precipitation and T

_{max}. The uncertainty is reported as 95% Confidence Intervals (CI).

Observations, 1985–2011 | ||||
---|---|---|---|---|

AFB | FBR | |||

Exceedance Probability (1985–2015) | Precipitation (mm/Day) (95% CI) | T_{max} (K) (95% CI) | Precipitation (mm/Day) (95% CI) | T_{max} (K) (95% CI) |

P (0.05) | 70.4 (43.9–131.3) | 317.2 (315.1–320.3) | 94.9 (66.3–152.4) | 313.7 (311.5–317.6) |

P (0.02) | 80.8 (46.0–183.1) | 317.9 (315.4–322.1) | 109.0 (70.1–204.7) | 314.1 (311.6–319.3 |

P (0.01) | 88.4 (47.1–233.4) | 318.3 (315.6–323.5) | 119.8 (72.5–254.9) | 314.3 (311.6–320.4) |

**Table 3.**Exceedance probability thresholds and associated values taken from bias-corrected modeled precipitation and T

_{max}values for three future periods (centered around the years 2020, 2050, and 2100), for both AFB and FBR locations under two emission scenarios, based on GEV distribution. The uncertainty is reported as 95% Confidence Intervals (CI).

RCP4.5 | RCP8.5 | |||||
---|---|---|---|---|---|---|

Year (Centered Around) | Exceedance Probability | Precipitation (mm/Day) (95% CI) | T_{max} (K) (95% CI) | Precipitation (mm/Day) (95% CI) | T_{max} (K) (95% CI) | |

AFB | 2005–2035 | P (0.05) | 73.8 (45.7–137.8) | 318.2 (316.1–321.8) | 80.2 (47.7–159.5) | 318.2 (316.2–321.6) |

P (0.02) | 83.0 (47.5–185.5) | 318.9 (316.3–324.1) | 92.7 (50.0–227.8) | 319.0 (316.4–323.6) | ||

P (0.01) | 89.1 (48.4–229.2) | 319.4 (316.4–325.9) | 101.7 (51.2–295.5) | 319.5 (316.6–325.2) | ||

2035–2065 | P (0.05) | 67.3 (44.3–109.1) | 319.1 (317.0–322.6) | 76.5 (46.2–138.0) | 319.1 (317.0–322.5) | |

P (0.02) | 74.7 (46.5–135.1) | 319.8 (317.3–324.9) | 87.6 (49.9–183.3) | 319.9 (317.3–324.5) | ||

P (0.01) | 79.5 (47.6–155.9) | 320.3 (317.4–326.6) | 95.2 (50.4–223.6) | 320.3 (317.4–325.9) | ||

2085–2115 | P (0.05) | 66.0 (43.4–103.2) | 319.3 (317.2–322.7) | 80.45 (46.4–155.8) | 319.9 (317.8–323.2) | |

P (0.02) | 72.8 (45.7–123.5) | 320.1 (317.5–324.9) | 94.3 (49.5–219.4) | 320.6 (318.1–325.0) | ||

P (0.01) | 77.0 (46.9–138.5) | 320.6 (317.7–326.6) | 104.3 (51.1–280.1) | 321.0 (318.2–326.4) | ||

Year (centered around) | Exceedance Probability | Precipitation (mm/day) (95% CI) | T_{max} (K) (95% CI) | Precipitation (mm/day) (95% CI) | T_{max} (K) (95% CI) | |

FBR | 2005–2035 | P (0.05) | 95.8 (65.9–140.6) | 314.5 (312.4–318.5) | 91.3 (64.1–136.3) | 313.9 (311.7–317.8) |

P (0.02) | 105.6 (69.9–164.4) | 314.9 (312.4–320.0) | 103.2 (68.4–168.8) | 314.3 (311.8–319.5) | ||

P (0.01) | 111.6 (72.1–181.4) | 315.0 (312.5–321.1) | 111.4 (71.0–195.6) | 314.5 (311.8–320.6) | ||

2035–2065 | P (0.05) | 93.5 (63.4–134.7) | 314.9 (312.7–318.7) | 92.5 (65.1–134.8) | 315.3 (313.1–319.1) | |

P (0.02) | 103.4 (69.6–157.5) | 315.3 (312.8–320.2) | 103.1 (69.4–161.4) | 315.8 (313.3–320.8) | ||

P (0.01) | 109.6 (72.0–173.8) | 315.5 (312.9–321.3) | 110.1 (71.8–181.8) | 316.1 (313.3–322.0) | ||

2085–2115 | P (0.05) | 94.6 (66.4–135.5) | 315.3 (313.1–319.4) | 93.0 (65.3–134.2) | 315.9 (313.7–319.8) | |

P (0.02) | 104.1 (70.5–157.1) | 315.7 (313.2–321.1) | 103.2 (69.6–158.0) | 316.4 (313.9–321.4) | ||

P (0.01) | 109.9 (72.8–172.2) | 315.9 (313.2–322.2) | 109.6 (72.0–175.4) | 316.6 (313.9–322.5) |

**Table 4.**The ratio of probability of the extreme event in the future that corresponded to p < 0.01 in the past and its probability for three future periods (centered around the years 2020, 2050, and 2100), for both AFB and FBR locations under two emission scenarios, based on GEV distribution. The uncertainty is reported as 95% Confidence Intervals (CI).

AFB | FBR | |||||||
---|---|---|---|---|---|---|---|---|

Precipitation (mm/Day) (95% CI) | T_{max} (K) (95% CI) | Precipitation (mm/Day) (95% CI) | T_{max} (K) (95% CI) | |||||

RCP4.5 | RCP8.5 | RCP4.5 | RCP8.5 | RCP4.5 | RCP8.5 | RCP4.5 | RCP8.5 | |

2005–2035 | 1.09 (0–16.5) | 2.77 (0–20) | 4.35 (0–21.4) | 4.63 (0–21.5) | 0.30 (0–10) | 0.47 (0–8) | 7.58 (0–35.5) | 1.86 (0–26) |

2035–2065 | 0.20 (0–11) | 1.87 (0–17) | 11.35 (0–31.1) | 11.63 (0–31.5) | 0.23 (0–9) | 0.33 (0–8) | 11.83 (0–38) | 17.65 (0–41.5) |

2085–2115 | 0.06 (0–10) | 2.97 (0–19) | 13.60 (0–33.4) | 22.45 (0.7–42.5) | 0.21 (0–9) | 0.25 (0–9) | 19.98 (0–45) | 31.04 (0–52.5) |

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Tadić, J.M.; Biraud, S.C.
Effects of Climate Change on Precipitation and the Maximum Daily Temperature (T_{max}) at Two US Military Bases with Different Present-Day Climates. *Climate* **2020**, *8*, 18.
https://doi.org/10.3390/cli8020018

**AMA Style**

Tadić JM, Biraud SC.
Effects of Climate Change on Precipitation and the Maximum Daily Temperature (T_{max}) at Two US Military Bases with Different Present-Day Climates. *Climate*. 2020; 8(2):18.
https://doi.org/10.3390/cli8020018

**Chicago/Turabian Style**

Tadić, Jovan M., and Sébastien C. Biraud.
2020. "Effects of Climate Change on Precipitation and the Maximum Daily Temperature (T_{max}) at Two US Military Bases with Different Present-Day Climates" *Climate* 8, no. 2: 18.
https://doi.org/10.3390/cli8020018