**5. Discussion**

Spatial and temporal precipitation data acquisition is highly important for the regional climate studies as well as for the management of water resources and agriculture of the concerned regions. With the advancement in geo-information and remote sensing technologies, this paves the way to acquire gridded data free of cost on a different time and scale. However, the reliability and accuracy of the data must be quantified against the in-situ records at the regional scale [77]. The current study provides a comprehensive assessment of GB and SB precipitation products against the reference records over the Punjab province during annual, winter, and summer monsoon from the period of 1979–2017 and 2003–2017, respectively. Moreover, the changes in trends and detection of abrupt turning point in temporal data series of products were evaluated and compared with the reference data series. In addition, the correction factor for each product was introduced to minimize the percentage of error in the respective products.

The results indicate that the spatial and temporal performance of GPCC and TRMM precipitation product outperformed the other corresponding products in terms of statistical measurements (high value of CC and R2 with lower value of ME and RMSD) during the annual and summer monsoon season during 1979–2017 and 2003–2017, respectively. However, CRU performed better during winter estimates when compared with reference data during the whole study period. Moreover, the range of deviation among different precipitation products underestimated with different magnitude as compared with the reference data. We found large uncertainties in the magnitude and temporal variability among different products. The magnitude range of deviation was up 100–200 mm among different products within the same category and with reference data. The results of deviation are associated with the findings of [48], who reviewed the 22 global precipitation products and was reported the magnitude of deviation up to 300 mm. All of the selected products estimated the higher amount of annual precipitation in the northern Punjab and exhibited the similar pattern from north to south gradient with the range of 1500–15 mm. However, the spatial deviation was more pronounced in the northern side as compared with the central and southern part of the study region. Moreover, the pattern of underestimation was more pronounced in the GB products as compared with SB products. One of the possible explanations of underestimation is due to lower pixel resolution of GB products over the study region [28].

Furthermore, the results showed that all of the GB products performed better in terms of lower values of ME during winter following by annual and summer monsoon season. However, the SB products showed better performance during annual scale following by summer and winter monsoon period. Overall, the agreement of GB products with observed precipitation decreases with the increase in the amount of precipitation and vice versa for SB precipitation products. The performance of SB precipitation products is well concomitant with the findings of [28,78], who reported the better performance of SB products to capture the higher magnitude of precipitation and less accurate for low precipitation events over different regions of Pakistan. In contrast, [79] revealed the better performance of SB products with less rainfall events over northwest Himalaya regions. The extent of deviation in product similarity can possibly be attributed to multiple factors, which include difference in study region, elevation, land use and land cover change [28]. The pattern similarity by using Taylor diagram indicates the better performance of GPCC and TRMM products during annual and summer monsoon seasons in terms of higher CC of around 95% during the whole study period. The results are well associated with the findings of [27], who reported the better pattern similarity performance of TRMM product in terms of higher CC of around 80% against other datasets during annual time scale over the upper Indus basin. The spatial distribution of statistical metrics (ME, RMSD, and CC) indicates the better performance of GPCC and TRMM products. However, the range of deviation in GB and SB products were more obvious in the northern Punjab with the overall pattern of underestimation (GB products) and overestimation (SB products) from the north to south gradient. The possible explanation of estimated deviation in GB and SB products over the northern Punjab could be due to wind induced errors and the presence of higher amount of aerosols in the atmosphere. The higher amount of aerosols in the atmosphere intercepts with the precipitation that reduces the efficiency of ground based meteorological stations [80]. Moreover, the wind induced errors can be attributed to the efficacy of ground based stations [81] Similarly, the pattern of overestimation in SB products particularly over northern Punjab could be associated with the higher rate of evaporation in the lower reaches of the mountains. These results are well associated with the findings of [82], who indicated the overestimation of SB products in the foothills of Himalaya mountain ranges over northern Pakistan. Moreover, the deviation in accuracy of GB products over northern Punjab could be due to inter-annual variability in weather, orographic effects and fewer gauges over the mountainous region. Many studies reported such effects in different climate studies over the northern belt of Pakistan [27,28,79]. Moreover, the fewer pixels of precipitation products also showed a weaker ability to capture the accurate spatial pattern in the central and southern parts of the study area. The possible explanation of accuracy deviation in these areas could be due the rapid extent of urban areas. Most of the meteorological station installed in the non-urban domain which could not represent the exact situation of climate variability in urban areas. Several studies documented the effect of urbanization on climate variability at a spatial scale [83].

The MK trends of observed precipitation indicate insignificant increasing trends during the period 1979–2017 and 2003–2017 with the rate of 1.12 and 5.5 mm/decade, respectively. The increasing trend was faster during period 2003–17 as compared with the period 1979–2017. The MK trend of GB and SB products showed the increasing trend with different magnitude of overestimation, except UDel and CPC products, which indicates decreasing trend as compared with observed precipitation. Furthermore, the abrupt transition analysis indicates the multiple turning points in precipitation data series during the period 1980–2005 with major abrupt changes in progressive series were detected during the period of 1984–1985 (negative), 1988–1989 (positive), 1998–1999 (negative), and 2005–2006 (positive). Many studies over the Asian and Indian monsoon system exhibited similar kind of precipitation trends over different timescales. [84–86]. The negative precipitation trend detected during the period of 1984–1985 and 1998–1999 could be due to major drought period over the whole country. The results are consistent with the findings of [87,88], who reported the severe drought conditions in the country during the mid-80s and late 90s. Moreover, the gentle increase with no obvious break point has been detected in progressive data series during the period of 2005–2017. The large inter-annual variability in reference data series were well captured by GPCC and TRMM products during the period 1979–2017 and 2003–2017, respectively.

In spite of the fact that the spatial and temporal performance of GPCC and TRMM products outperformed the respective products in terms of lower values of ME, RMSD, and higher range of CC against in-situ records during the period 1979–2017 and 2003–2017, respectively. However, it is still uncertain to estimate the accurate amount of precipitation, as witnessed by the estimated mean error in the precipitation products over the whole study region. In view of significant extent of biases in the products, we determined the factor of correction of each product for their reliable utilization in hydro-climate projects on annual, winter, and summer monsoon periods over the Punjab region. The results of estimated correction factors of different precipitation products are well covenant with the findings of [27,89], who estimated the adjustment factors of precipitation products for Upper Indus Basin (UIB). However, the magnitude of the estimated correction factors is different due to different study period and region.

The spatial and temporal performance of the products depend on a multiple factors, which are intricate in the data algorithms of the products, e.g., sources of data, spatial and temporal resolution, interpolation techniques, missing gaps in data and topography, etc. [50,90]. Similarly, the number of stations, quality, and time scale of in-situ records used for the evaluation of global products are also very important for the identification of potential product for the specific regions [72]. The detection of autocorrelation in the datasets is also important for the accuracy of trends detection and abrupt changes in the climate data series. However, the presence of autocorrelation was more significant for precipitation products, particularly for high altitude regions above 4000 m [27]. The superior performance of GPCC and TRMM precipitation products over the Punjab region might be due to their better data processing procedures, number of gauge station into account, and interpolation techniques.

The precipitation products, even with their intrinsic biases and limitations, are still important for providing valuable source of information related to precipitation variability on spatial and temporal scale. The global precipitation products are also important for climate studies when there is a lack of funding or resources to go into the field and record these types of observations. Caution must also be exercised when comparing and using the GPPs; large uncertainty exists where gauge density is low [91]. The present study highlights the potential products for precipitation over the Punjab province. The evaluation results are beneficial for improving our understanding towards the use of products in arid and semi-arid regions, like Punjab province. However, the spatial and temporal discrepancies were also identified, which will be useful for the further use of these products in hydro-meteorological applications. In view of large discrepancies among the products, future studies should focus on the new methods for the better comparative accuracy and evaluation procedure of global climate products. For instance, [92] introduce a new method of evaluation and decision making by using the fuzzy logic procedure based on the integrated linguistic operated weighted average (ILOWA) method. The main

advantage of this technique is to provide sustainable products by using different linguistic terms that are primarily an easy approach for decision making. The products assessments based on integrated models could increase the level of accuracy for the appropriate selection of potential products by considering the statistical indicators.
