*3.1. Model Structure*

The model details are available from the accompanying Supplementary Material and Langarudi et al. [3]. This paper's system dynamics model consists of 15 stocks and 33 flows and simulates from 1969 to 2099. The model is structured into four modules: water, capital, land, and population. All the details of the assumptions and modeling choices, including equations, parameter values, estimations and measurements, and data sources, are reported in the Supplementary Material. These modules interact with each

other within a complex feedback system. The relationships among modules are constructed based on the literature, previous models, or empirical studies (black and blue linkages in Figure 2). The water module consists of surface water, soil moisture of irrigated and non-irrigated land, and groundwater as stocks. The primary physical processes are integrated into the model building, including surface inflow/outflow, river leakage, surface water withdrawal, canal leakage, field percolation, gaining flow from groundwater, and groundwater withdrawal. Income, capital development, irrigated land, and population growth contribute to water demand (Figure 2). Agricultural parameters, such as crop type and soil properties, are aggregated into the model design, addressing their impacts on the hydrologic cycle dynamics and water demand in socio-economic processes. Water availability is driven by precipitation, surface inflow, and groundwater storage. Water availability impacts income, capital development, irrigated land and population in turn. The model successfully passes the confidence building (validation) tests presented in the Supplementary Materials, supporting that this model can provide insights into regional water issues.

**Figure 2.** Causal structure of the model: positive or negative causality is marked as plus or minus; double sides arrows explains that two variables have mutual feedback; blue dash line represents where the policy implements.


Hydroclimate scenarios are introduced in a changing future. Model inputs (precipitation, temperature, and surface water inflow) are acquired from the New Mexico Dynamic Statewide Water Budget (DSWB) model [32]. The DSWB model generates data by using climate projections (Global Circulation Model), including GFDL (Geophysical Fluid Dynamics Laboratory), UKMO (United Kingdom Met Office), and NCAR (National Center for Atmospheric Research). Based on different greenhouse gas emission scenarios, each climate projection offers different drought conditions [26].

The GFDL, UKMO, and NCAR projections are used as inputs because they represent low, moderate and, high emission scenarios. The scenarios are listed in Table 1, where the prediction part of the simulation is divided between 2017–2050 and 2051–2099. The

simulation interval is divided to illustrate the short and long-term results of the climate scenarios.


**Table 1.** Climate inputs for scenario tests with average values of periods 2017–2050 and 2051–2099 (Units of surface inflow are in thousands of acre-feet, KAF).

### 3.2.2. Climate Scenarios as Input

Precipitation has relatively high variability during 2017–2050 and 2051–2099 in the LRG planning region. Compared to historical average annual precipitation, the GFDL projection shows a decreasing trend, UKMO shows an increasing trend, and NCAR is similar to historical conditions.

Projections of temperature have relatively low variability in projections but show a potential increase of 5.1 ◦F in the long-term. All projections show temperature increasing to some degree as time progresses. Compared with historical average annual temperature (61.3 ◦F), GFDL shows a greatly increasing trend, UKMO shows a moderately increasing trend, and NCAR shows a mildly increasing trend.

Projections of surface inflow had high variability and no clear trends over time as all projections showed similar or increased values in the near-term which all decreased in the long-term. Compared with historical average annual surface inflow, the GFDL has similar then decreased flow, UKMO shows increased then slightly decreased flow, and NCAR shows greatly increased then moderately increased.
