پايش خشكي كشاورزي در مقياس مزرعه مبتني بر تصاوير دورسنجي مايكروويو رطوبت خاك

Field-Scale Agricultural Drought Monitoring Based on Microwave Imagery of Soil Moisture

وقوع خشكسالي در كشاورزي صرفا با اندازه‌گيري تغييرات بارش قابل رصد نيست بلكه متغيرهاي ديگري همچون رطوبت خاك نيز در آن نقش دارند. در ميان روش‌هاي مختلف دورسنجي، طيف الكترومغناطيس مايكروويو محدوديتهاي فيزيكي ساير امواج راديومتري در اندازهگيري رطوبت خاك را ندارد. با اين تفاوت كه دادههاي مايكروويو رطوبت خاك غالبا داراي ابعاد پيكسل بسيار بزرگ (بيش از 10 كيلومتر) هستند و اين موضوع كاربرد آنها در مقياسهاي كوچك را با مشكل مواجه ميسازد. در اين پژوهش به منظور محاسبه شاخص خشكي كشاورزي در مقياس مزرعه، ابتدا با استفاده از دادههاي اندازهگيري ميداني رطوبت در محدوده دشت نيشابور طي سالهاي 1396 تا 1398، واسنجي دادههاي بازيابي رطوبت خاك سنجنده AMSR2 انجام شد. سپس با كمك تصاوير سنجنده موديس روابط خطي ريزمقياس نمايي تصاوير رطوبت خاك استخراج شده و ابعاد تصوير از 25 كيلومتر به 1000 متر كاهش يافت. در گام بعدي از شاخص خشكي كشاورزي SMADI كه تلفيقي از خصوصيات پوشش گياهي، رطوبت خاك و دماي سطح زمين است براي پايش خشكي كشاورزي در مقياس مزرعه استفاده شد. به منظور ارزيابي نتايج، شاخصهاي آماري ضريب تعيين ( )، ميانگين قدرمطلق خطا (MAE) و ريشه ميانگين مربعات خطا (RMSE) در سه كاربري اراضي منتخب شامل زراعت ديم (R1)، مرتع متوسط (R2) و مرتع فقير (R3) بررسي شد. شاخص­هاي MAE و RMSE در بازه 1.6 تا 4 و شاخص در بازه 0.73 تا 0.84 قرار گرفت. نتايج نشان داد كه الگوريتم استفاده شده در ريزمقياس نمايي و همچنين برآورد شاخص خشكي كشاورزي SMADI به خوبي قادر به بازتاب اندركنشهاي بين بارش، رطوبت خاك، پوشش گياهي و تغييرات پروفيل دمايي كانوپي است و اين ويژگي كاربرد آن را در تحليلهاي هواشناسي كشاورزي توجيه و تقويت ميكند.

Introduction  Drought analysis in agriculture can not only be achieved by measuring precipitation changes but also by using other parameters such as soil moisture. Due to the fact that soil moisture affects plant growth and yield, it is often considered for monitoring agricultural drought. Remote sensing data are often provided from three sources: microwave, visible and thermal. Most satellite soil moisture-based algorithms rely on passive microwave images, active microwaves, or a combination of data from several different sensors. Among the various remote sensing methods, the microwave electromagnetic spectrum has fewer physical limitations than other spectrum in measuring soil moisture. However, microwave soil moisture data often have very large pixel dimensions (more than 10 km), making it difficult to use them on a small scale. Materials and Methods  In this study, in order to calculate the agricultural drought index at the field-scale, AMSR2 Retrieval data were calibrated first using field moisture measurement data in the Neishabour plain during 2017 to 2019. During the research period, 560 soil samples (20 samples in 28 shifts) were collected and soil moisture was measured in the laboratory of the Department of Water Science and Engineering, Ferdowsi University of Mashhad. LPRM_AMSR2_ SOILM3_001 is one of the third level products of the AMSR2 sensor, which is produced on a daily basis with a spatial resolution of 25 × 25 km2. Land surface parameters including surface temperature, surface soil moisture and plant water availability were obtained by passive microwave data using the Land parameter Retrieval Method (LPRM). Then, by using Modis sensor images (NDVI and LST), linear downscaling equations were extracted. The dimensions of the AMSR2 images were reduced from 25 kilometers to 1000 meters using these equations. In next step, SMADI Agricultural Drought Index, which is a combination of vegetation characteristics, soil moisture and land surface temperature, was used to monitor agricultural drought at the field-scale. Statistical indicators such as coefficient of determination (R^2), mean absolute error (MAE) and root mean square error (RMSE) were also used to evaluate the statistical performance. Results and Discussion By visual analysis of the role of vegetation and land unevenness, it was found that these two factors affect the regression relationships extracted for calibration of remote sensing data. The RMSE and MAE values for the regression equations used in the calibration process were calculated in the range of 1.6 to 4%, which can be considered acceptable in comparison with the mean values of the soil moisture data (15 to 20). The results showed that changes in SMADI index in three land use zones including rainfed cultivation (R1), medium rangeland (R2) and poor rangeland (R3) have experienced a similar trend to precipitation changes, illustrating that precipitation is one of the most effective factors in major changes in SMADI agricultural drought index fluctuations. It was also observed that SMADI index changes with a delay of 1 to 8 days compared to the precipitation changes in all three zones. In all three zones, the SMADI index followed a similar trend to in-situ soil moisture changes. At mot 80% of the changes in SMADI-R1 index can be explained by in-situ SM-R1, and the rest of the changes were related to other environmental factors or measurement error. This decreases to 68% in the R3 zone. It should be noted that soil moisture monitoring can more accurately reflect the impact of environmental factors on the changes in agricultural drought index such as SMADI than other variables; because the rainfall recorded at the meteorological station does not necessarily occur uniformly throughout the study area. On the other hand, any amount of precipitation will not necessarily lead to an effective change in soil moisture storage. This also renders assessment of the performance of agricultural drought indicators difficult. Conclusion  Examination of statistical indices of coefficient of determination (R2), mean absolute error value (MAE) and root mean square error (RMSE) showed that the algorithm used in downscaling as well as estimating SMADI agricultural drought index is well able to reflect the interactions between precipitation, soil moisture, vegetation and changes in canopy temperature profile. This feature justifies and strengthens its application in agrometeorological analysis.