بررسی اثرات تغییر اقلیم بر نوسانات تراز آب دریاچه ارومیه

دریاچه ارومیه به‌عنوان یکی از بزرگ‌ترین دریاچه‌های کشور در شمال غرب ایران واقع شده است. با توجه به تغییرات بارش و دما و رخداد خشک‌سالی‌ها و ترسالی‌های شدید در این حوضه و ویژگی‌های مرفولوژیک آن (شیب کم سواحل) سطح تراز آن دچار نوسانات و تغییرات زیادی می‌شود. بطوریکه در سال‌های اخیر توسعه طرح‌های منابع آبی و به‌ویژه احداث سدهای مخزنی بر روی رودخانه‌های منتهی به دریاچه، تاثیر زیادی بر ورودی آب و درنهایت تراز آب آن داشته است. در این مطالعه با استفاده از روش شبکه عصبی مصنوعی، شبیه‌سازی نوسانات دریاچه مورد بررسی قرارگرفته است. نقش تغییر اقلیم بر تراز آینده دریاچه تحت سناریوهای a2 و b2 با استفاده از ریزمقیاس نمایی خروجی مدل‌های اقلیمی به کمک الگوی ریزمقیاس نمائی larswg تحلیل شده است. نتایج تحقیق حاضر نشان می‌دهد که تراز دریاچه طی سال‌های آینده روند کاهشی را طی خواهد کرد.

Evaluation of climate change impacts on Urmia lake water level fluctuations

Introduction:Lake Urmia, the largest territorial lake, has a great role on climate balance in the region. In recent decades, agricultural and industrial activities, infrastructural and water resources development projects has been extremely propagated. Thereby, such activities have a remarkable impact on ecological condition and water level of Lake Urmia. Although climate change also plays a great role on water level. Due to the importance of the issue, various studies have been carried out in recent years about the effect of climate change on water level changes.so in this paper attempt to simulate the fluctuations of the lake  using artificial neural network approach and finally evaluate the effects of future climate change on lake levels. Methods: Analyses in this research are done in 3 steps. In the first step, the climate change scenarios of temperature and rainfall resulted from a AOGCM model are generated under two scenarios A2 and B2 for the future period 201–2060. In the second, the outputs of this model are downscaled by the method of changefactor LARS downscaling. Thirdly, the  Urmia lake level   changes is modeled by the artificial neural network, and then the values are downscaled and introduced to the network to, finally, the lake  water level values be computed at two emission scenarios. 2.1.Generation of climate change scenariosTo study climate change for the future periods, generation of climate scenarios is necessary, the most convincing way to generate which is utilizing the output of AOGCM models that are based on physics laws and mathematical formulae to be solved in a 3D network on the surface of the earth. To simulate the climate of the globe, the main climate processes atmosphere, ocean, the earth’s surface, scale ice, and biosphere are simulated in tributary models separately, and then those of atmosphere and ocean are matched together to form AOGCM models (IPCCTGCIA, 1999). In the present work, effects of climate change on lake levels have analyzed by outputs of HadCM3 model  using A2 and B2 scenarios. 2.2.Downscaling outputs of  HadCM3 Model Local Down scaling:In this study, the method of “change factor downscaling” is used for local downscalingof AOGCM model. In this method the usual monthly ratios are obtained fromhistorical series, and the climate variables simulated by AOGCM are derived from the cell in which the region under study is placed. First of all, the climate change scenarios for temperature and rainfall are generated. To compute the scenarios for each model, the “difference” values for temperature, and “ratio” for rainfall for longterm average per month in the period 2011–2030 and 20112060 as well as the simulated base period (19611990) is computed for each cell by the computational network. Statistical downscaling:For comparison between the observed data and those from probability distribution andmean, the model LARSWG uses Chisquare (χ2) test and t test, respectively. The testsare based on the assumption that the observed and simulated meteorological are the same. The test survey the null hypothesis to the effect that the two distributions or the two means are similar, so that the difference is not significant. In this research, the LARS model was calibrated by the observed data of minimum and maximum temperature as well as precipitation in the period 1961–1990. Afterwards, the LARS model was performed for climate change scenarios of temperature and precipitation in the region, and then the time series of temperature and precipitation for the period 2011–2060 was figured out. The results of downscaled temperature and precipitation for Urmia station is shown in Fig 1 and 2 respectively.   Fig1.: Downscaled temperature for Urmia station during 20112030  Fig 2.: Downscaled precipitation for Urmia station during 20112030 Simulation of climate change impacts on Urmia lake water level changesAn  feed forward artificial neural network was used for simulation of lake level changes. To get the best result, various input models were defined and assessed some of which. Since the most important factors affecting water level of the lake are precipitation, entering discharges to the lake, temperature, and surface evaporation they are applied as inputs and biases in the network design. The final selected model to determine future lake level constitutes of inputs such as precipitation in month, total inflow in month  and evaporation in month. Also the  future total inflow to the lake was estimated by a feed forward network using monthly precipitation, monthly temperature and also number of Julian month.In this research, 70 % of data of last period 1961–1980 was dedicated to the validationstage, and the remaining 30 % for the validation stage 1980–1990. The Fig 3 shows the simulated monthly inflow  to lake and observed flow during validation period.  Fig3. Simulated and observed inflow during validation period. Simulation of monthly inflow to the Urmia lake for the future period and comparison with theobserved period indicates the decline of average runoff into the lake Moreover, the results say that the changes of the modeled runoff in the humid period (May to April) are more than the dryperiod (November to May). Also, the results certify a decrease in the average futurerunoff in scenario A2 is more than scenario B1 relative to the base period. Also the simulated water level of Urmia lake shows that the lake level will decrease over the coming years and this decreasing  are more than double in A2 scenario as compared to B2 scenario as shown in Fig 4.  Fig4. Simulation of Urmia lake level change during 20112100