ارزیابی محصولات پایگاه‌های بارش شبکه ای در گام های زمانی متفاوت در یک حوضه آبریز کوهستانی (مطالعه موردی: حوضه آبریز سد دز)

Introduction

Precise precipitation measurement is crucial in various sectors, including water management, climate research, agriculture, and disaster response. It is essential for sustainable development and the well-being of societies worldwide. Although rain gauge data provide accurate precipitation measurements at station scales, they are inadequate for estimating precipitation over larger areas or locations where weather stations cannot be established. In addition to rain gauge observations, gridded data can offer reliable support for precipitation estimates. Effective monitoring and accurate assessment of precipitation in mountain catchments are vital for understanding hydrological processes, predicting water availability, and assessing the impacts of climate change in these sensitive regions. Gridded precipitation products yield valuable insights into precipitation's spatial and temporal variability, facilitating informed decision-making in water resource management. This paper aims to evaluate the performance of gridded precipitation products at different time scales in mountain catchments, providing a comprehensive analysis of their reliability and suitability for hydrological applications. By comparing and assessing various gridded datasets, this study seeks to enhance the understanding of precipitation patterns and improve the accuracy of precipitation measurements in mountainous areas. Gridded data are available in the form of cells for all points within their spatial coverage, and are employed to estimate climatic parameters, including precipitation. This study evaluates gridded precipitation data products at annual, monthly, and daily scales, both with and without seasonal differentiation.

Materials and Methods

In this study, the ERA5 and PERSIANN-CDR products were utilized in the Dez Dam watershed. ERA5 is the latest generation of reanalysis datasets (5th generation) developed by the European Centre for Medium-Range Weather Forecasts (ECMWF). Key improvements over ERA-Interim include a better spatial grid, higher temporal resolution, more vertical levels, a new numerical weather prediction (NWP) model, and an expanded dataset for data mining. This dataset covers the period from 1950 to near real-time. Daily precipitation data from ERA5 with a spatial resolution of 0.25 degrees was used, along with precipitation estimates from the PERSIANN-based satellite algorithm, which combines infrared (IR) and passive microwave (PMW) data. Each of these datasets is derived from data collected by geostationary (GOE) and low Earth orbit (LOE) satellites. The algorithm employs artificial neural networks to extract data from super-cold pixels and the brightness temperature of various pixels, thus estimating precipitation levels for each pixel. Additionally, the monthly GPCP data product was used, ensuring high-resolution spatial and temporal patterns, with modified PERSIANN data demonstrating a good correlation with GPCP data on a monthly scale. These products were evaluated against rain gauge data in the Dez Dam basin using evaluation and diagnostic indices for daily scale assessments, a Taylor diagram for monthly evaluations, along time series data and box plots for annual scales.

Results and Discussion

The results show that in the estimate of annual precipitation, the results indicate that considering the RMSE values at the four observation stations, the performance of both models in estimating annual precipitation is not satisfactory. Additionally, for daily, monthly, and yearly precipitation estimates, the best predictive model at the Tele Zang and Dez Dam stations is the P-CDR product, whereas at the Tang-e-Panj Bakhtiari and Sepid Dasht Sazdar stations, the best dataset is ERA5. Accordingly, the highest RMSE values for daily, monthly, and annual scales are 2.8, 76.098, and 542.24 millimeters, respectively. The best daily NSE value corresponds to the ERA5 product at the Sepid Dasht Sazdar station with a value of 0.482. Furthermore, the highest correlation coefficient for the ERA5 data is at the Sepid Dasht station, with a value of 0.729. In terms of Probability of Detection (POD) precipitation detection indicators and False Alarm Ratio (FAR), the best values of 0.568 and 0.169 corresponded to the ERA5 dataset at the Tang Panj Bakhtiari station. It was concluded that the number of non-rainy days in observational data exceeds that of all datasets across all stations, with ERA5 data being closer to observed values in representing the number of non-rainy days than the P-CDR data. However, in estimating rainy days (i.e., when precipitation is greater than 0 mm but less than 3 mm), all datasets and stations indicated estimated precipitation values greater than observed data.

Conclusion

This study evaluated the gridded precipitation data from ERA5 reanalysis products and PERSIANN-CDR satellite-rain gauge data in the Dez mountain watershed. Verified data from the Khuzestan Water and Electricity Organization were used across four meteorological stations: Tele Zang, Dez Dam, Tang Panj Bakhtiari, and Sepiddasht Caesar during the period from 2008 to 2019. For annual assessments, the estimated annual rainfall derived from gridded data was compared with observational data from the investigated stations, and the distribution of all datasets was displayed using box diagrams. For monthly assessments, a Taylor diagram illustrating gridded precipitation products was created to determine the appropriateness of each dataset. Additionally, for the daily evaluation of precipitation data, two methods were employed: one that assessed the data generally without considering seasonal differences, and another that evaluated daily meteorological data by separating the seasons. Four indices—Mean Bias Error (MBE), Root Mean Square Error (RMSE), BIAS, and correlation (R)—were calculated for daily evaluation, along with three classification indices—POD, FAR, and Critical Success Index (CSI). The results indicated that for estimating annual precipitation, the ERA5 reanalysis dataset was the most accurate, outperforming satellite-rain gauge data. It was also found that the P-CDR provided the best daily and monthly rainfall estimates at the Tele Zang and Dez Dam stations, while the ERA5 dataset performed best at the Tang Panj Bakhtiari and Sepiddasht Caesar stations.