Although Climate Information Services (CIS) has been promoted in Ghana and also in Bangladesh (see Gbangou et al. 2020 and Kumar et al., 2021), this system only provides information on the recent and forecasted meteorological variables, primarily precipitation and temperature. Soil moisture that plays an important role in the soil-plant-atmosphere system is still missing. Understanding soil moisture condition is key in agriculture practice because the plant establishment and growth are directly impacted by the soil moisture stored in the soil layer. For small-holder farmers, having access to soil moisture information when practicing rainfed agriculture would help them in the decision-making process and managing the effects of climate change on agriculture.
Providing CIS with soil moisture module is challenging due to missing soil moisture observations. Then how the soil moisture forecast information can be developed in the existing CIS when the soil moisture data is not available in e.g., Ghana and Bangladesh? In the Waterapps, we will use a simple bucket/water balance model to estimate the soil moisture condition (Figure 1). Then the soil moisture condition for the coming days will be estimated using the water balance model in combination with weather forecast data. The idea is to create a simple estimation of SM forecasts that can be run on a low-cost server within seconds. Therefore, the water balance model has only 1 soil layer and water exchange within soil layers is neglected (vertical and horizontal to deep soil layer/groundwater). Soil moisture is calculated using the water balance equation as follow:
DSM = inputs of water - losses of water = (P+I+C)-(ET+D+RO)
Where DSM is change in soil moisture, P is the rainfall, I is irrigation, C is the water from the groundwater, ET is the evapotranspiration, D is the water loss to deep drainage, and RO is surface runoff. In our simple model, we neglect the input water from groundwater and irrigation, and we also neglect the water loss to deep drainage and to surface runoff. In the end, the change in soil moisture is estimated from the difference between precipitation as input and evapotranspiration as output (DSM=P-ET).
Figure 1. Schematization of a bucket/water balance model
The available soil moisture at time step i is calculated as follow:
SMi = SMi-1+DSM
Where SMi-1 is the soil moisture in the previous time step, which is estimated and inputted to the apps by the farmers or using the data obtained from remote sensing products (see Figure 2 for example). In doing so, we have trained the farmers on how to estimate the soil moisture condition at the field by feel and appearance using the method introduced by USDA. So far, we conducted training on how to measure the soil moisture condition by feel and appearance in three communities in Tolon and Savelugu Districts in northern Ghana. Another method to measure the soil moisture condition is by using a soil moisture sensor that has become cheap and cheaper nowadays.
Figure 2. Detailed soil moisture data for Belgium obtained from Vandersat.
The soil data can be obtained from the ISRIC database but for some locations in northern Ghana, we took soil samples and measured them in the laboratory. To forecast the soil moisture condition, we use the water balance model with the forecasted precipitation (P), forecasted Evapotranspiration (ET), and observed soil moisture condition as input. The ET will be calculated using the Thornthwaite method, which requires only temperature data.
While the training to measure soil moisture condition by feel and appearance in the northern region of Ghana is still ongoing, we are updating the Waterapps to be ready to provide soil moisture information for small-holder farmers in Ghana and elsewhere. We expect that next year the CIS embedded with soil moisture module will be ready for operational purposes. We are also discussing with our partners in Senegal to test and implement the soil moisture advisory module there within the WAGRINNOVA project. Let’s give two fingers crossed for the Waterapps team.