**1. Introduction**

Winter wheat is the main crop in the North China Plain (NCP). Due to the high irrigation demand of winter wheat, more than 70% of the irrigated water resources are used for winter wheat irrigation every year [1]. The increasing population has led to a corresponding increase in the demand for agricultural, industrial and domestic water in the NCP. The surface water resources are insu fficient, and groundwater has become the main source of water for the NCP [2]. In recent decades, the overexploitation of groundwater has led to a significant decline in groundwater levels, which increases not only environmental problems but also the pressure on agricultural food production [3,4]. Groundwater is the main source of water for NCP agriculture irrigation. Long-term dependence on groundwater for agricultural irrigation has resulted in groundwater over-exploitation, and agricultural water irrigation needs to be reduced; however, the sustainable of food crop production must also be

ensured [5,6]. Timely and e ffective monitoring of irrigation water is of grea<sup>t</sup> significance for agricultural water managemen<sup>t</sup> and water resources protection. The irrigation signal includes the time, frequency and area of irrigation. Irrigation time can be used to dynamically correct irrigation schedules, while irrigation frequency and area can be used for the estimation and dynamic monitoring of agricultural irrigation water use [7–10]. This study prepares to establish a model that can be used to detect irrigation signals and dynamically acquire irrigation information. The results of the irrigation signal will be used for the dynamic monitoring of agricultural irrigation water to achieve refined managemen<sup>t</sup> of agricultural irrigation.

With the continuous development of remote sensing technology, more remote sensing data can be used for irrigation information detection [11–15]. Compared with traditional agricultural statistical methods, remote sensing has a wide range of multifrequency, high spatial and temporal resolution advantages and has been widely used in agricultural managemen<sup>t</sup> [16–18]. Representative data sources include Moderate Resolution Imaging Spectroradiometer (MODIS), which provides 250 m, 500 m and 1 km resolution daily surface reflectance data. The richness of time series and improvement in remote sensing data spatial resolution has greatly improved the accuracy of irrigated area identification [19]. In recent research, the Normalized Di fference Vegetation Index (NDVI) has been extensively used as an e ffective indicator for irrigated area recognition based on optical remote sensing data [19–21]. An analysis of the time-varying pattern of NDVI is the primary method for identifying irrigated and non-irrigated areas. In particular, wheat and maize are a ffected by irrigation, and their NDVIs will appear to be higher than other vegetation [20,22]. Although the identification method for irrigated areas has been comprehensive, this irrigated area extraction method based on optical remote sensing data is mostly used for long-term irrigated area monitoring to analyze trends in irrigated areas over multiple years. Chen et al. [23] proposed a method for detecting irrigation extent, timing and frequency based on MODIS and Landsat remote sensing data, which is an important irrigation property for understanding the sustainability of water resources in arid and semiarid regions. The irrigation signal detection method based on the visible vegetation index must model the daily scale data, and this method is more suitable for irrigation signal detection in regions with less cloud cover. Remote sensing images of areas with more clouds are likely to miss the critical period of irrigation signal detection due to cloud pollution. Moreover, in addition to the influence of image quality, the response of vegetation to irrigation is lagged, which increases the uncertainty of irrigation timing detection.

In addition to the method of identifying the irrigated area by using vegetation index information, the change in the wetness index can also be used to identify the irrigation signal [24]. Based on the SM being higher in the irrigated area than in the non-irrigated area, some researchers have identified irrigated areas based on di fferent principles. Based on the MODIS enhanced vegetation index (EVI) and land surface water index (LWSI) ratio method, Peng et al. [25] introduced the variable EVI/LWSI threshold function to improve the detection ability of this method in di fferent rice crops under mixed rice crop patterns (single-season rice, early-season rice, and late-season rice). Abuzar et al. [26] used vegetation and thermal thresholds derived from Landsat and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data to detect the irrigated area in an Australian irrigation district based on the soil temperature in the irrigated area being lower than that in the non-irrigated area. Although di fferent researchers use SM information to detect irrigated area information from di fferent aspects, they do not use SM indicators because optical and thermal infrared remote sensing data cannot directly obtain SM information.

Active and passive microwave satellites have proven to be e ffective tools for retrieving soil water variations at regional and global scales [27–29]. NASA's Soil Moisture Active Passive (SMAP) satellite, launched on 31 January 2015, provides a new source of data for near-surface (0–5 cm) soil water monitoring on a global scale. Colliander et al. [30] validated the SMAP surface SM product through the core validation site. The results indicate that the SMAP radiometer-based SM data product meets the expected performance of 0.04 m<sup>3</sup>/m<sup>3</sup> volumetric SM (unbiased root mean square error) and that the combined radar-radiometer product is close to its expected performance of 0.04 m<sup>3</sup>/m3. Chan et al. and

Zhang et al. [11,31] evaluated the results of di fferent SMAP products in di fferent regions and obtained similar conclusions to those of Colliander. SMAP has more information improvements than previous SM satellites, which has raised interest in whether SMAP can improve irrigation monitoring [32]. Subsequently, Lawston et al. [33] explored the use of SMAP data in identifying irrigation areas and timing in the Sacramento Valley, San Luis Valley and Columbia River Valley. However, the study did not identify the irrigation timing in the irrigated area. Since the detection of the irrigated area is a combination of changes in SM over a period of time, the time scale is the entire period of the crop. Compared with optical/thermal infrared methods, SMAP's method of detecting irrigated areas has unique advantages in terms of temporal resolution and ability to directly acquire SM [34]. The SMAP data spatial resolution is a major limiting factor that a ffects its use.

Obtaining irrigation time, area and frequency will help estimate irrigation water volume and provide data support for agricultural irrigation management. Despite having the low spatial resolution, SMAP provides high temporal resolution SM products. To address the spatial resolution issues, this paper will be studied in the following three aspects: 1) Based on SMAP and meteorological data, the irrigation signal in the study area was detected, which solved the problem of optical data not being applicable in cloudy regions; 2) MODIS remote sensing data were used to downscale the detection results to solve the low spatial resolution problem of SMAP data; and 3) through an analysis of the consistency of winter wheat growth covered by SMAP pixels, the SMAP data e ffectiveness in downscaling the winter wheat irrigation results in this study area was evaluated.

## **2. Study Area**

The region of interest in this paper is located in the southern part of the Hebei Province and belongs to the NCP. The boundaries of the study area are city administrative boundaries, including Shi Jiazhuang, Baoding, Langfang, Hengshui, Cangzhou, Xingtai and Handan, with a total area of 8.9 × 10<sup>4</sup> km<sup>2</sup> (as shown in Figure 1). Although precipitation in the study area is not scarce, the distribution of precipitation during the year is extremely uneven. The study area is dominated by a temperate monsoon climate with mean annual precipitation of 472.7–889.2 mm, and 70% of the annual precipitation occurs between June and September [35]. Under the irrigation conditions of the study area in recent years, the main crop pattern is the winter wheat-summer maize double crop rotation. Winter wheat and summer maize are also the main irrigated crops in this region [36]. The lower amount of precipitation in spring is not enough to provide su fficient water for winter wheat growth, and groundwater irrigation has been the main irrigation method for winter wheat and summer maize for a long time. Winter wheat is generally irrigated 4–5 times, and precipitation has little e ffect on the number of irrigations due to the severe shortage of precipitation during the winter wheat growing period. Summer maize is usually irrigated before planting, and if e ffective precipitation has occurred before planting and the soil moisture meets the sowing requirements, the crop will not be irrigated during the growing period. The Middle Route of the South-to-North Water Transfer Project (MSWTP) was launched at the end of 2014, and this project provided a new source of water for agricultural irrigation in the NCP [37].

**Figure 1.** Study area and meteorological sites locations and the spatial distribution of SM stations.

#### **3. Materials and Methods**

The flow chart (shown in Figure 2) of this paper includes the processing of collected data (Section 3.1), selection of samples (Section 3.2.1), the application of algorithms (Section 3.2.2) and validation of accuracy (Section 3.2.3).

**Figure 2.** Flow chart for this study. Here, 5-point Mov Avg represents the 5-point moving average and Avg and Std represent the average and standard deviation, respectively. The irrigation Acc accumulates as a result of the irrigation signal.

#### *3.1. Data Collection and Pre-Processing*
